{ "as_of": "2026-08-08", "source": "Reconciled against the Data Dawgs site, Worker source and inspected upstream projects on 2026-08-08. College Football roadmap ideas added from the 2026-08-08 CFB roadmap.", "tier": "pound", "graded": false, "note": "Two kinds of entry share this inventory. kind \"tool\" (NFL) carries delivery status. kind \"roadmap-idea\" (College Football) carries recommendation plus evidence-backed lifecycle and implementation state. A CFB artifact may be implemented without any MCP tool; candidate_mcp_tools are reservations and remain non-callable unless the independent Worker registry proves otherwise.", "cfb_roadmap": { "as_of": "2026-08-08", "domain": "College Football", "note": "The 12-step ordering is directional, not an irreversible commitment; dependencies ultimately determine the actual build sequence. Recommendations are recommendations, not permanent truths.", "steps": [ { "step": 1, "title": "Foundation / canonical CFB data", "idea_ids": [ "cfb-sportsdataverse", "cfb-cfbd", "cfb-games", "cfb-plays", "cfb-team-game", "cfb-player-game", "cfb-rosters", "cfb-market", "cfb-talent-dataset", "cfb-team-week", "cfb-public-outputs" ] }, { "step": 2, "title": "Ratings registry", "idea_ids": [ "cfb-ratings-registry", "cfb-elo" ] }, { "step": 3, "title": "Model disagreement + receipts", "idea_ids": [ "cfb-disagreement-lab", "cfb-model-receipts" ] }, { "step": 4, "title": "Model diversity / consensus", "idea_ids": [ "cfb-consensus-engine" ] }, { "step": 5, "title": "Fourth-down engine", "idea_ids": [ "cfb-fourth-down" ] }, { "step": 6, "title": "Talent / recruiting / portal", "idea_ids": [ "cfb-portal-flow", "cfb-acquisition-network", "cfb-returning-production", "cfb-talent-composite" ] }, { "step": 7, "title": "Fraud Detector", "idea_ids": [ "cfb-fraud-detector" ] }, { "step": 8, "title": "Opponent-adjusted unit ratings + matchup engine", "idea_ids": [ "cfb-unit-ratings", "cfb-matchup-fingerprints" ] }, { "step": 9, "title": "Season / playoff path simulator", "idea_ids": [ "cfb-season-sim" ] }, { "step": 10, "title": "Coaching decision scorecards", "idea_ids": [ "cfb-coaching-scorecards" ] }, { "step": 11, "title": "Data Dawgs predictive model", "idea_ids": [ "cfb-dd-power-rating" ] }, { "step": 12, "title": "Experimental challenger models", "idea_ids": [ "cfb-challenger-lab" ] } ], "recommendation_classes": { "build": "Strong recommendation: fits the DATA → MODELS/DETERMINISTIC TOOLS → API/MCP → AI → HUMAN DECISION architecture with a plausible path to useful production functionality. Modifiers distinguish foundation, high-priority and decision-engine work.", "lab": "Promising hypothesis worth testing. It must demonstrate incremental predictive or decision value before promotion, and it is explicitly allowed to fail.", "defer": "Interesting, but expected signal, data quality or maintenance economics do not justify building it now. The idea is kept.", "avoid-initially": "Possible eventual value, but current dependency, scraping, legal, reliability, maintenance or architectural concerns make it undesirable right now.", "not-needed": "Redundant with the architecture Data Dawgs is already building." }, "lifecycle": { "statuses": [ "idea", "evaluating", "planned", "building", "live", "deferred", "graveyard", "revived" ], "happy_path": "idea → evaluating → planned → building → live", "alternate_exits": [ "idea/evaluating/live → deferred", "any → graveyard", "graveyard → revived" ], "policy": "Ideas are never silently deleted. Moving to the Graveyard means \"tested or evaluated, and currently not worth active resources\" — the record, its history and its lessons are retained so the Graveyard works as institutional memory, not a trash can.", "graveyard_postmortem_fields": [ "original_hypothesis", "intended_outcome", "recommendation_when_proposed", "why_attempted", "implementation_version", "data_used", "validation_design", "expected_benefit", "actual_result", "performance_vs_baseline", "cost_complexity_maintenance", "known_failure_modes", "reason_retired", "date_retired", "receipts", "lessons_learned", "revival_conditions", "links", "lifecycle_history" ], "revival_question": "What would have to become true for us to dig this Dawg back up?" }, "governance": [ { "id": "cfb-gov-source-provenance", "name": "Source Provenance", "source_heading": "Source Provenance", "principle": "Every important derived CFB artifact should eventually be traceable to its source datasets and processing versions, the way the NFL backbone pins commits and snapshot hashes.", "applies_to": "every CFB idea", "related_ideas": [ "cfb-sportsdataverse", "cfb-cfbd", "cfb-public-outputs" ] }, { "id": "cfb-gov-snapshotting", "name": "Snapshotting / Receipts", "source_heading": "Snapshotting / Receipts", "principle": "Forecasts and model outputs used for grading must be frozen before outcomes occur. This is the same receipts discipline already live for NFL models.", "applies_to": "every CFB forecast or model output", "related_ideas": [ "cfb-model-receipts" ] }, { "id": "cfb-gov-model-cards", "name": "Model Cards", "source_heading": "Model Cards", "principle": "Important models eventually document purpose, target, features, training window, validation design, known limitations, calibration, performance, failure modes, version and retirement status.", "applies_to": "every CFB model promoted past the lab", "related_ideas": [ "cfb-dd-power-rating", "cfb-xrepa", "cfb-challenger-lab" ] }, { "id": "cfb-gov-baseline", "name": "Baseline Requirement", "source_heading": "Baseline Requirement", "principle": "No sophisticated model is promoted merely because it is sophisticated. Everything is benchmarked against the market and simple interpretable baselines.", "applies_to": "every CFB model", "related_ideas": [ "cfb-elo", "cfb-challenger-lab", "cfb-dd-power-rating" ] }, { "id": "cfb-gov-incremental", "name": "Incremental Information Test", "source_heading": "Incremental Information Test", "principle": "New models and features must demonstrate value relative to what Data Dawgs already knows before they earn production resources.", "applies_to": "every CFB model and feature family", "related_ideas": [ "cfb-context-lab", "cfb-availability-lab", "cfb-challenger-lab" ] }, { "id": "cfb-gov-correlation", "name": "Correlation Awareness", "source_heading": "Correlation Awareness", "principle": "Multiple correlated models are not multiple independent pieces of evidence. Consensus construction must measure error correlation before treating agreement as confirmation.", "applies_to": "every multi-model CFB surface", "related_ideas": [ "cfb-consensus-engine", "cfb-disagreement-lab" ] }, { "id": "cfb-gov-uncertainty", "name": "Uncertainty", "source_heading": "Uncertainty", "principle": "Avoid false precision, especially for weak or incomplete datasets such as NIL information. Outputs carry uncertainty and their date.", "applies_to": "every CFB output", "related_ideas": [ "cfb-nil", "cfb-counterfactual", "cfb-fraud-detector" ] } ], "source_headings": [ "SportsDataverse bulk-data ingestion", "CFBD enrichment/model ingestion", "Canonical CFB Games", "Canonical CFB Play-by-Play", "CFB Team-Game Dataset", "CFB Player-Game Dataset", "CFB Rosters / Identity Resolution", "CFB Market Dataset", "CFB Ratings Registry", "CFB Talent Dataset", "CFB Team-Week Analytical Layer", "Compact public CFB outputs", "Full offline-reprocessable upstream/raw pipeline", "Static SQLite / browser-queryable exploratory database", "CFB Model Disagreement Lab", "Model Receipts / Historical Grading", "Model Diversity / Consensus Engine", "Continuous Elo / Glicko Rating", "Opponent-Adjusted Unit Ratings", "Data Dawgs Predictive Power Rating", "Deep/Trajectory Model", "XGBoost / Other ML Challenger Models", "CFBD Model Training Pack / educational modeling resources", "Transfer Portal Flow Network", "Recruiting Acquisition Network", "Player Acquisition Graph", "Returning Production", "Portal Net Flow", "Recruiting Composite / Talent Composite", "Raw 247 / On3 / Rivals Scraping", "NIL Deal Tracking / NIL Valuation", "Fourth-Down Decision Engine", "Fourth-Down Yards-Gained Distribution Model", "CPOE", "xREPA", "Matchup Fingerprints", "Counterfactual Simulator", "Schedule Path Simulator", "Game Leverage Index", "Prediction-Market Integration", "Overrated / Underrated Team Detector", "Coaching Trees", "Coaching Movement / Performance Delta", "Coaching Decision Scorecards", "Weather Effects", "Altitude", "Travel Distance / Time Zones", "Indoor / Outdoor Effects", "Injury / Availability Data", "Practice Participation / Beat-Writer Proxies", "Media / Reddit / Social Sentiment", "ESPN Total QBR Historical Layer", "Deterministic CFB MCP/API Tool Layer", "Local LLM Generated Reports", "Source Provenance", "Snapshotting / Receipts", "Model Cards", "Baseline Requirement", "Incremental Information Test", "Correlation Awareness", "Uncertainty" ] }, "data": [ { "id": "model-scoreboard", "name": "Model Scoreboard / Mean & Conflict", "intended_user_value": "Compare dated nfelo and 538 Classic beliefs game by game.", "existing_website_implementation": "/dawghouse.html#scoreboard", "existing_worker_mcp_implementation": "dd_model_scoreboard", "staged_worker_mcp_implementation": null, "required_data": "Uses the normalized prospective nfelo and 538 Classic receipts published by Data Dawgs.", "human_facing_ui_requirement": "Interactive game table with metric/source states.", "ai_language_requirement": "Explain the equal-weight mean, range, dissent and missing models without claiming a validated consensus blend.", "mcp_api_requirement": "Keep dd_model_scoreboard read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-receipts.json + /data/538-classic.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "disagreement", "name": "Transparent Disagreement Engine", "intended_user_value": "Show mean, median, range, standard deviation and decision-boundary splits.", "existing_website_implementation": "/receipts.html#models + /calculators.html", "existing_worker_mcp_implementation": "dd_summarize_beliefs", "staged_worker_mcp_implementation": null, "required_data": "Any list of probabilities, plus the current scoreboard.", "human_facing_ui_requirement": "Belief-summary lab and per-game descriptive statistics.", "ai_language_requirement": "Descriptive uncertainty only; never call disagreement a predictive edge.", "mcp_api_requirement": "Keep dd_summarize_beliefs read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "538-classic", "name": "538 Classic Elo", "intended_user_value": "Provide a permanent simple benchmark/control.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_elo_game", "staged_worker_mcp_implementation": null, "required_data": "Pinned official reference, completed canonical game history and the 2026 schedule snapshot.", "human_facing_ui_requirement": "Prospective scoreboard column plus one-game probability calculator.", "ai_language_requirement": "Permanent simple benchmark; reproduced historically and still ungraded prospectively.", "mcp_api_requirement": "Keep dd_elo_game read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/538-classic.json + /data/model-receipts.json + /data/538-classic-methodology.md", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "nfelo", "name": "nfelo Power Ratings", "intended_user_value": "Use a dated public model output with Data Dawgs validation.", "existing_website_implementation": "/nfelo.html + /dawghouse.html#scoreboard", "existing_worker_mcp_implementation": "dd_analyze_matchup", "staged_worker_mcp_implementation": null, "required_data": "2026-08-06 nfelo snapshot and existing historical backtest.", "human_facing_ui_requirement": "Existing nfelo page plus Pound scoreboard.", "ai_language_requirement": "State that the backtest overlaps optimization and has not established out-of-sample skill.", "mcp_api_requirement": "Keep dd_analyze_matchup read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/nfelo.json + /data/models.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "nfeloml", "name": "nfeloml EP / EPA / WP", "intended_user_value": "Offer deterministic play-level inference primitives.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "MIT package and canonical play inputs.", "human_facing_ui_requirement": "Contract and provenance only.", "ai_language_requirement": "Do not present upstream calibration as Data Dawgs-verified.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/upstream-models.json", "status": "backend-blocked", "exact_blocker": "Static GitHub Pages cannot run the Python/XGBoost package; a precompute job or Worker-compatible port is required.", "minimum_path_to_completion": "Run the MIT package in a scheduled precompute job with schema checks and publish static output, or build and validate a Worker-compatible inference port.", "kind": "tool", "domain": "NFL" }, { "id": "nfeloqb", "name": "QB Context", "intended_user_value": "Measure starter/replacement and passing-context deltas.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "Lawful dated output plus starter mapping.", "human_facing_ui_requirement": "Inventory card; no invented QB adjustment.", "ai_language_requirement": "Context is optional and unvalidated for DFS or game forecasting.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/upstream-models.json", "status": "data-blocked", "exact_blocker": "No verified reusable license and no canonical Data Dawgs adapter.", "minimum_path_to_completion": "Obtain a reusable license, permission, or lawful dated output; then build the starter mapping, adapter, and prospective validation.", "kind": "tool", "domain": "NFL" }, { "id": "wepa", "name": "Weighted EPA", "intended_user_value": "Compare context-weighted EPA with ordinary EPA.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "Lawful methodology or output plus canonical play data.", "human_facing_ui_requirement": "Inventory card and explicit ordinary-EPA baseline.", "ai_language_requirement": "Any improvement claim waits on fixed out-of-sample tests.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/upstream-models.json", "status": "data-blocked", "exact_blocker": "No verified reusable license and no Data Dawgs implementation.", "minimum_path_to_completion": "Obtain a reusable methodology or licensed output, implement the weighting against canonical play data, and test it against ordinary EPA out of sample.", "kind": "tool", "domain": "NFL" }, { "id": "srs", "name": "SRS / Win-Total Ratings", "intended_user_value": "Estimate team strength, priors and schedule difficulty.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "Timestamped lawful win-total market inputs and schedule.", "human_facing_ui_requirement": "Inventory card.", "ai_language_requirement": "Market inputs must carry source, time and vig handling.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/upstream-models.json", "status": "data-blocked", "exact_blocker": "No verified reusable license or licensed market feed.", "minimum_path_to_completion": "Secure a lawful win-total feed and reusable methodology, then publish versioned preseason and weekly ratings from canonical schedules.", "kind": "tool", "domain": "NFL" }, { "id": "hfa", "name": "Home-Field Adjustment Tracker", "intended_user_value": "Inspect the effect of a transparent HFA assumption.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_elo_game", "staged_worker_mcp_implementation": null, "required_data": "User-supplied ratings and HFA; site defaults are dated.", "human_facing_ui_requirement": "Interactive Elo/margin calculator.", "ai_language_requirement": "Labels inputs modelled and does not claim the default is current truth.", "mcp_api_requirement": "Keep dd_elo_game read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/models.json", "status": "frontend-only", "exact_blocker": "Historical HFA tracking feed is not built.", "minimum_path_to_completion": "Build a canonical historical-game input, estimate HFA by declared period/context, and publish a dated series with uncertainty.", "kind": "tool", "domain": "NFL" }, { "id": "units", "name": "Unit Ratings", "intended_user_value": "Separate pass, rush and special-teams strengths.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "Lawful unit outputs and canonical play data.", "human_facing_ui_requirement": "Inventory card.", "ai_language_requirement": "Upstream performance numbers remain upstream claims.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/upstream-models.json", "status": "data-blocked", "exact_blocker": "Research/educational language is not a reusable software license.", "minimum_path_to_completion": "Obtain a reusable license, permission, or lawful output; normalize unit ratings to canonical teams and replicate performance claims independently.", "kind": "tool", "domain": "NFL" }, { "id": "translation", "name": "Probability / Spread Translation", "intended_user_value": "Translate win probability into expected margin and cover estimates.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_translate_probability", "staged_worker_mcp_implementation": null, "required_data": "User probability and dated Data Dawgs normal residual SD.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "Calls the result MODELLED and names the normal, independent, no-key-number assumption.", "mcp_api_requirement": "Keep dd_translate_probability read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/models.json + /data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "market", "name": "Market Normalizer", "intended_user_value": "Normalize American odds, implied probability, hold and devig probabilities.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_devig_market", "staged_worker_mcp_implementation": null, "required_data": "User-entered prices.", "human_facing_ui_requirement": "Odds and hold/vig calculators.", "ai_language_requirement": "No sportsbook feed, closing-line claim or vendor attribution.", "mcp_api_requirement": "Keep dd_devig_market read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "cover-ev", "name": "Cover Probability / EV", "intended_user_value": "Turn a declared probability and price into break-even and expected value.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_calculate_bet_ev", "staged_worker_mcp_implementation": null, "required_data": "User probability and price, or the modelled normal translator.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "Distinguishes the user/model probability from the price-implied probability.", "mcp_api_requirement": "Keep dd_calculate_bet_ev read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "odds", "name": "Odds Converter", "intended_user_value": "Convert American odds to decimal odds and implied probability.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_convert_odds", "staged_worker_mcp_implementation": null, "required_data": "User-entered odds.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "Deterministic conversion; no predictive claim.", "mcp_api_requirement": "Keep dd_convert_odds read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "parlay", "name": "Parlay Calculator", "intended_user_value": "Compound declared leg prices.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_price_parlay", "staged_worker_mcp_implementation": null, "required_data": "User-entered American odds.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "Multiplies declared prices; does not claim a correlation-aware joint probability.", "mcp_api_requirement": "Keep dd_price_parlay read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "hedge", "name": "Hedge Calculator", "intended_user_value": "Size the opposite side for equal net outcome.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_calculate_hedge", "staged_worker_mcp_implementation": null, "required_data": "Original stake/price and hedge price.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "Arithmetic, not advice; ignores limits, tax and execution risk.", "mcp_api_requirement": "Keep dd_calculate_hedge read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "passer", "name": "NFL Passer Rating", "intended_user_value": "Reproduce the public deterministic NFL passer-rating formula.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_nfl_passer_rating", "staged_worker_mcp_implementation": null, "required_data": "Attempts, completions, yards, touchdowns and interceptions.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "Descriptive statistic, not a forecast or QBR substitute.", "mcp_api_requirement": "Keep dd_nfl_passer_rating read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "grader", "name": "Forecast Grader", "intended_user_value": "Compute Brier score and log loss for one declared forecast.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": "dd_score_forecast", "staged_worker_mcp_implementation": null, "required_data": "Probability and binary outcome.", "human_facing_ui_requirement": "Interactive calculator.", "ai_language_requirement": "One observation is not model validation; sample size stays visible.", "mcp_api_requirement": "Keep dd_score_forecast read-only and align its contract with the public JSON.", "machine_readable_requirement": "/data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "receipts", "name": "Immutable Multi-Model Receipts", "intended_user_value": "Make prospective forecasts auditable after outcomes are known.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "Normalized forecasts, input snapshots and market state before kickoff.", "human_facing_ui_requirement": "Contract and link to the existing 2026 receipt ledger.", "ai_language_requirement": "Never mix backtests and prospective forecasts.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/model-receipts.json + /data/model-contracts.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "id": "regimes", "name": "Regime Analysis", "intended_user_value": "Evaluate performance by declared football conditions.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "Adequate prospective comparable samples and predeclared slices.", "human_facing_ui_requirement": "Inventory card only.", "ai_language_requirement": "Post-hoc slices are exploratory; sample sizes and multiple-comparison risk must be shown.", "mcp_api_requirement": "Add a read-only Worker tool only after the required deterministic data or runtime is validated.", "machine_readable_requirement": "/data/model-contracts.json", "status": "data-blocked", "exact_blocker": "Two-model prospective receipts exist, but no 2026 outcomes have occurred or been graded yet.", "minimum_path_to_completion": "Accumulate adequate prospective samples, preregister slices and minimum counts, and mark post-hoc analysis exploratory.", "kind": "tool", "domain": "NFL" }, { "id": "ensembles", "name": "Ensemble Lab", "intended_user_value": "Compare mean, median and later weighted blends.", "existing_website_implementation": "/calculators.html", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "User-entered probabilities or future normalized forecasts.", "human_facing_ui_requirement": "Belief-summary calculator; equal-weight only.", "ai_language_requirement": "Complex methods wait until they beat equal weight on held-out data.", "mcp_api_requirement": "Expose the staged deterministic browser function through a read-only Worker tool using the published calculator contract.", "machine_readable_requirement": "/data/model-contracts.json", "status": "frontend-only", "exact_blocker": "Weighted ensembles need prospective error history.", "minimum_path_to_completion": "Accumulate prospective model errors, then benchmark mean, median and candidate weighting methods against the market and best individual model.", "kind": "tool", "domain": "NFL" }, { "id": "nfl-data", "name": "NFL Data Backbone", "intended_user_value": "Create validated, reproducible canonical snapshots for every model.", "existing_website_implementation": "/dawghouse.html#inventory", "existing_worker_mcp_implementation": null, "staged_worker_mcp_implementation": null, "required_data": "nfelodcm, GitHub Actions, schema/freshness gates and snapshot retention.", "human_facing_ui_requirement": "Inventory and provenance contract.", "ai_language_requirement": "Fail closed on schema drift and suspicious row-count changes.", "mcp_api_requirement": "Static public schedule; no MCP calculation is required.", "machine_readable_requirement": "/data/nfl-schedule.json + /data/model-receipts.json + /data/upstream-models.json", "status": "complete", "exact_blocker": null, "minimum_path_to_completion": "No known blocker for the deployed scope.", "kind": "tool", "domain": "NFL" }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "live", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Built pinned, fail-closed schedule and market ingestion." }, { "status": "live", "on": "2026-08-08", "note": "Canonical schedule and book-market snapshots published with source-byte hashes." } ], "implemented": true, "delivery_evidence": [ "/scripts/cfb_data_backbone.py", "/scripts/cfb_market.py", "/data/cfb-schedule.json", "/data/cfb-market.json" ], "delivery_note": "Live for the schedule and book-market source families. Play-by-play remains a separate cfb-plays dependency.", "graveyard_ready": true, "priority": "high", "roadmap_step": 1, "recommendation_modifiers": [ "foundation", "high-priority" ], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-cfbd", "cfb-games", "cfb-plays" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-sportsdataverse", "name": "SportsDataverse Bulk-Data Ingestion", "category": "foundation", "recommendation": "build", "source_headings": [ "SportsDataverse bulk-data ingestion" ], "rationale": "SportsDataverse already publishes analysis-ready CFB data. Reproducing the ESPN ingestion and enrichment ecosystem would be undifferentiated maintenance work.", "expected_value": "Primary bulk-data substrate: SportsDataverse → Data Dawgs canonical CFB layer feeding every downstream dataset.", "data_source_requirements": "Pinned SportsDataverse CFB releases with schema, freshness and row-count gates at ingest, mirroring the NFL backbone.", "deterministic_tool_potential": "Indirect — this is infrastructure every deterministic CFB tool sits on.", "validation_requirement": "Ingest gates fail closed on schema drift or suspicious row counts; retained snapshots make every load reproducible.", "risks_limitations": "Upstream schema drift and inherited coverage gaps; the layer must never silently pass upstream errors downstream.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 1, "recommendation_modifiers": [ "foundation", "high-priority" ], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-sportsdataverse", "cfb-ratings-registry", "cfb-talent-dataset" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-cfbd", "name": "CFBD Enrichment / Model Ingestion", "category": "foundation", "recommendation": "build", "source_headings": [ "CFBD enrichment/model ingestion" ], "rationale": "CollegeFootballData.com is an enrichment and model feed — SP+, FPI, SRS, recruiting, talent, returning production, schedules — not the sole center of the architecture.", "expected_value": "CFBD → Data Dawgs canonical CFB layer: ratings and talent enrichments the bulk substrate does not carry.", "data_source_requirements": "CFBD API access with keys held outside the public repo; dated pulls; per-feed schema checks.", "deterministic_tool_potential": "Indirect — feeds the ratings registry and talent datasets.", "validation_requirement": "Every pull dated and snapshotted; feed-level gates; upstream numbers stay labelled as upstream claims until independently validated.", "risks_limitations": "API dependency, rate limits and upstream revisions; enrichment fields can change definitions between seasons.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "live", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Canonicalized 2025 FBS-involved schedule and result facts." }, { "status": "live", "on": "2026-08-08", "note": "Published 934 validated games with unique IDs and a reproducible snapshot hash." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-schedule.json", "/scripts/cfb_data_backbone.py", "/tests/test_cfb_data_backbone.py", "/work/mcp-block.js", "/work/test-mcp.mjs" ], "delivery_note": "Live for the completed 2025 season with a bounded production MCP game finder. The upstream 2026 schedule has not been published yet.", "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-sportsdataverse" ], "candidate_mcp_tools": [ "dd_find_cfb_games" ], "components": [], "related_ideas": [ "cfb-plays", "cfb-market" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-games", "name": "Canonical CFB Games", "category": "foundation", "recommendation": "build", "source_headings": [ "Canonical CFB Games" ], "rationale": "A canonical cfb_games dataset is the spine every schedule, rating, receipt and simulator joins against.", "expected_value": "Durable game IDs, results and kickoff context for all downstream CFB work.", "data_source_requirements": "SportsDataverse schedules/results reconciled with CFBD reference data.", "deterministic_tool_potential": "Yes — bounded canonical game reads and downstream reference joins.", "validation_requirement": "Unique game IDs, season/week coverage gates and snapshot hashes, like /data/nfl-schedule.json.", "risks_limitations": "FBS/FCS boundary cases, cancellations and rescheduling need explicit rules.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-sportsdataverse", "cfb-games" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-team-week", "cfb-fourth-down" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-plays", "name": "Canonical CFB Play-by-Play", "category": "foundation", "recommendation": "build", "source_headings": [ "Canonical CFB Play-by-Play" ], "rationale": "A canonical cfb_plays layer carries the useful play-level fields Data Dawgs needs without forcing downstream applications to understand the entire upstream schema.", "expected_value": "The substrate for EPA, success rate, fourth-down modelling, CPOE, xREPA and unit ratings.", "data_source_requirements": "SportsDataverse play-by-play with EPA/WP fields; documented field selection and derivation notes.", "deterministic_tool_potential": "Indirect — every play-level deterministic tool reads it.", "validation_requirement": "Row-count and coverage gates per season; spot-checks against published aggregates before anything downstream ships.", "risks_limitations": "CFB play-by-play is messier than NFL data: garbage-time definitions, clock quirks and missing fields need explicit handling.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "building", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Published the results-only foundation: two mirrored rows per canonical game. Play-derived aggregates remain unavailable." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-team-game.json", "/data/cfb-games-latest.json", "/scripts/cfb_team_results.py", "/tests/test_cfb_team_results.py", "/work/mcp-block.js" ], "delivery_note": "Partially implemented with observed schedule/result facts, a compact latest-completed-game row per FBS team, and a bounded exact-team game-history reader live over MCP. EPA, success, explosiveness, havoc and garbage-time fields still require canonical play-by-play.", "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-plays", "cfb-games" ], "candidate_mcp_tools": [ "dd_find_cfb_team_games" ], "components": [], "related_ideas": [ "cfb-team-week", "cfb-player-game" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-team-game", "name": "Canonical CFB Team-Game Dataset", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Team-Game Dataset" ], "rationale": "Canonical cfb_team_game: one row per team per game, aggregated once, consistently, instead of separately inside every tool.", "expected_value": "Shared team-game aggregates for ratings, the fraud detector and the team-week layer.", "data_source_requirements": "Derived from canonical plays and games.", "deterministic_tool_potential": "Yes — bounded team-perspective game history is a deterministic read over the aggregate layer.", "validation_requirement": "Aggregates reconcile with the play layer; derivations documented and reproducible.", "risks_limitations": "Aggregation choices (garbage time, overtime) must be declared, not implicit.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-plays", "cfb-rosters" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-cpoe", "cfb-xrepa", "cfb-qbr" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-player-game", "name": "Canonical CFB Player-Game Dataset", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Player-Game Dataset" ], "rationale": "Canonical cfb_player_game gives player-level analysis one consistent home instead of ad-hoc extraction per question.", "expected_value": "Player-game rows for CPOE, xREPA, QBR comparisons and portal/talent follow-up analysis.", "data_source_requirements": "Derived from canonical plays plus roster identity resolution.", "deterministic_tool_potential": "Indirect — aggregate layer.", "validation_requirement": "Player IDs resolve through the roster layer; totals reconcile with team-game rows.", "risks_limitations": "Player attribution in CFB play-by-play is incomplete for some seasons and stat types.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-sportsdataverse", "cfb-cfbd" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-portal-flow", "cfb-acquisition-network", "cfb-player-game" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-rosters", "name": "CFB Rosters / Identity Resolution", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Rosters / Identity Resolution" ], "rationale": "Canonical cfb_rosters with durable IDs and cross-source identity handling. Portal, recruiting and player analysis all die without it.", "expected_value": "Durable player identity across SportsDataverse, CFBD, recruiting classes and portal moves.", "data_source_requirements": "SportsDataverse and CFBD rosters; documented matching rules; manual override table for hard cases.", "deterministic_tool_potential": "Indirect — identity layer.", "validation_requirement": "Match-rate reporting per source pair; ambiguous matches flagged rather than silently merged.", "risks_limitations": "Name collisions, transfers and roster churn make identity the highest-touch dataset in the foundation.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "building", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Published book-identified historical prices; observation time remains unknown." }, { "status": "building", "on": "2026-08-08", "note": "Activated the immutable 24-hour pregame capture cron and public receipt export; first observations depend on eligible 2026 events entering the window." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-market.json", "/scripts/cfb_market.py", "/tests/test_cfb_market.py", "/work/mcp-block.js", "/work/test-mcp.mjs" ], "delivery_note": "Partially implemented. Historical book prices are live with explicit unknown timing and a no-CLV MCP reader. The timestamped 24-hour prospective collector and public export are active on the production Worker; first observations depend on eligible 2026 events entering the capture window.", "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-games" ], "candidate_mcp_tools": [ "dd_find_cfb_historical_market" ], "components": [], "related_ideas": [ "cfb-ratings-registry", "cfb-model-receipts", "cfb-prediction-markets" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-market", "name": "CFB Market Dataset", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Market Dataset" ], "rationale": "Canonical spreads, totals and relevant market observations. The market is the permanent benchmark every model must beat.", "expected_value": "Market baselines for grading; closing lines retained wherever possible for model evaluation.", "data_source_requirements": "A lawful, dated market feed with book and timestamp provenance — the same gate the NFL side already enforces.", "deterministic_tool_potential": "Yes — bounded historical market reads, with timing limitations enforced in the response.", "validation_requirement": "Every observation carries book, timestamp and line; closing-line retention is checked, not assumed.", "risks_limitations": "Historical closing lines with provenance are the hardest part; unprovenanced market numbers stay out, as on the NFL side.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION", "MARKET" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "live", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Defined a source-locked normalization contract over the shipped Elo baseline." }, { "status": "live", "on": "2026-08-08", "note": "Published 136 teams in one dated registry; unsupported outputs remain null and consensus remains explicitly unbuilt." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-ratings.json", "/scripts/cfb_ratings_registry.py", "/tests/test_cfb_ratings_registry.py", "/work/mcp-block.js" ], "delivery_note": "The registry contract is live with one independent, retrodictive Elo system. Bounded registry metadata and ranking/conference readers are live over MCP. CFBD systems and market context remain future additions; one system is not a consensus.", "graveyard_ready": true, "priority": "high", "roadmap_step": 2, "recommendation_modifiers": [ "foundation", "high-priority" ], "dependencies": [ "cfb-games", "cfb-elo" ], "candidate_mcp_tools": [ "dd_get_cfb_rating_system", "dd_rank_cfb_teams" ], "components": [], "related_ideas": [ "cfb-cfbd", "cfb-market", "cfb-disagreement-lab", "cfb-consensus-engine" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-ratings-registry", "name": "CFB Ratings Registry", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Ratings Registry" ], "rationale": "Canonical cfb_ratings stores independent rating and forecast systems consistently — SP+, FPI, SRS, Elo, Data Dawgs EPA ratings, the betting market and eventual Data Dawgs predictive models.", "expected_value": "Normalized expected margin, win probability, predicted total where applicable, and team strength — the comparable outputs everything downstream reads.", "data_source_requirements": "CFBD ratings feeds, the canonical market dataset and internally computed ratings, all dated per system per week.", "deterministic_tool_potential": "Yes — registry reads are deterministic lookups.", "validation_requirement": "Each system documents its normalization into comparable outputs; unconverted or unavailable fields stay null rather than being invented.", "risks_limitations": "Normalizing heterogeneous systems into comparable outputs is where silent errors live; the NFL model-contracts approach applies.", "tags": [ "CFB", "FOUNDATION", "MODEL", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-cfbd", "cfb-rosters" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-returning-production", "cfb-portal-flow", "cfb-talent-composite" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-talent-dataset", "name": "CFB Talent Dataset", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Talent Dataset" ], "rationale": "Canonical talent, recruiting, returning-production and portal features in one dated dataset instead of scattered pulls.", "expected_value": "The feature substrate for the talent/portal ideas and preseason priors.", "data_source_requirements": "CFBD recruiting, talent, returning production and portal feeds keyed to canonical rosters.", "deterministic_tool_potential": "Indirect — feature layer.", "validation_requirement": "Feature families dated and versioned; coverage per team per season reported.", "risks_limitations": "Portal data is noisier than recruiting composites; coverage varies by season.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION", "TALENT" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "building", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Published results-only team-period history and an exact compact latest row per team with opponent, venue, overall record, non-authoritative regular-season conference record and scoring facts. Advanced analytical metrics remain gated." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-team-week.json", "/data/cfb-team-week-latest.json", "/scripts/cfb_team_results.py", "/tests/test_cfb_team_results.py", "/work/mcp-block.js" ], "delivery_note": "Partially implemented. The live surfaces contain observed results, season-to-date scoring/record facts, an explicitly non-authoritative regular-season conference record and a compact exact latest-period row per team, with bounded history and latest readers live over MCP. The latest reader can compare conference records arithmetically but publishes no official rank or tiebreaker result; play, opponent-adjusted and market-performance metrics remain unavailable.", "graveyard_ready": true, "priority": "high", "roadmap_step": 1, "recommendation_modifiers": [ "foundation", "high-priority" ], "dependencies": [ "cfb-plays", "cfb-team-game", "cfb-market" ], "candidate_mcp_tools": [ "dd_find_cfb_team_periods" ], "components": [], "related_ideas": [ "cfb-unit-ratings", "cfb-fraud-detector", "cfb-ratings-registry" ], "governance": [ "cfb-gov-uncertainty" ], "revisit_conditions": null, "notes": null, "id": "cfb-team-week", "name": "CFB Team-Week Analytical Layer", "category": "foundation", "recommendation": "build", "source_headings": [ "CFB Team-Week Analytical Layer" ], "rationale": "The canonical cfb_team_week layer is the standardized analytical substrate for downstream CFB tools.", "expected_value": "One dated row per team per week carrying the candidate metric set: EPA/play, offensive/defensive EPA, dropback and rush EPA, success rate, explosive rate, havoc, early-down EPA, late-down success, garbage-time-filtered metrics, rolling 3- and 5-game metrics, home/away splits, opponent-adjusted versions, strength of schedule, market performance and closing-line error.", "data_source_requirements": "Canonical plays, team-game aggregates and the market dataset.", "deterministic_tool_potential": "Yes — bounded exact-team period history is a deterministic read.", "validation_requirement": "Metric definitions documented once and spot-checked against published sources; opponent adjustment validated against known cases before anything downstream trusts it.", "risks_limitations": "Opponent adjustment and garbage-time filters are modelling choices, and must be labelled as such rather than presented as observation.", "tags": [ "CFB", "FOUNDATION", "MODEL", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "building", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Published compact 136-team rating profiles, 136 latest completed FBS team games and 230 results-only latest-period team rows, each locked to its exact source snapshot." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-teams.json", "/data/cfb-games-latest.json", "/data/cfb-team-week-latest.json", "/scripts/cfb_team_profiles.py", "/scripts/cfb_team_results.py", "/tests/test_cfb_team_profiles.py", "/tests/test_cfb_team_results.py" ], "delivery_note": "Partially implemented with compact rating profiles, exact latest completed FBS team games and exact latest-period team rows carrying overall and non-authoritative conference records; both latest surfaces are readable over MCP as the latest-per-team scope of the team-game and team-period readers. Richer team-week metrics and model grades remain future outputs.", "graveyard_ready": true, "priority": "standard", "roadmap_step": 1, "recommendation_modifiers": [ "foundation" ], "dependencies": [ "cfb-games", "cfb-team-week", "cfb-ratings-registry" ], "candidate_mcp_tools": [ "dd_find_cfb_team_games", "dd_find_cfb_team_periods" ], "components": [], "related_ideas": [ "cfb-sqlite-explorer", "cfb-mcp-layer" ], "governance": [ "cfb-gov-source-provenance" ], "revisit_conditions": null, "notes": null, "id": "cfb-public-outputs", "name": "Compact Public CFB Outputs", "category": "delivery", "recommendation": "build", "source_headings": [ "Compact public CFB outputs" ], "rationale": "Publish compact derived outputs rather than dumping massive play-by-play datasets into the website repository — the same envelope-and-manifest discipline as the existing /data/ contract.", "expected_value": "A possible /data/cfb/ family: manifest.json, teams.json, games-latest.json, team-week-latest.json, ratings-latest.json, disagreement-latest.json, model-grades.json.", "data_source_requirements": "The canonical CFB layers above; every file carries as_of, source and a manifest entry.", "deterministic_tool_potential": "Yes — static surfaces are the cheapest deterministic reads.", "validation_requirement": "Same validator contract as existing /data/: envelope fields mandatory, manifest agreement enforced.", "risks_limitations": "Derived outputs drift from their sources unless rebuild discipline is enforced, exactly as AGENTS.md warns for the NFL data.", "tags": [ "CFB", "DATA-PIPE", "FOUNDATION", "API" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [ "defer-until-justified" ], "dependencies": [ "cfb-sportsdataverse" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-public-outputs" ], "governance": [], "revisit_conditions": [ "A real reprocessing need the canonical layer cannot answer", "Upstream reliability degrades", "Storage/maintenance cost drops materially" ], "notes": null, "id": "cfb-raw-pipeline", "name": "Full Offline-Reprocessable Raw Pipeline", "category": "delivery", "recommendation": "lab", "source_headings": [ "Full offline-reprocessable upstream/raw pipeline" ], "rationale": "A raw → enriched → canonical archival pipeline with manifests and hashes is worth preserving as an option, but duplicating SportsDataverse infrastructure merely because we can is exactly the wrong instinct.", "expected_value": "Full reprocessability if upstream sources become unreliable or if a question needs fields the canonical layer dropped.", "data_source_requirements": "Bulk raw archives with storage and versioning costs accepted up front.", "deterministic_tool_potential": "Indirect.", "validation_requirement": "A concrete case where the canonical layer was insufficient — before any build starts.", "risks_limitations": "High storage and maintenance cost for value that SportsDataverse already provides.", "tags": [ "CFB", "DATA-PIPE", "EXPERIMENTAL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [ "cfb-public-outputs" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-public-outputs" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-sqlite-explorer", "name": "Static SQLite / Browser-Queryable Explorer", "category": "delivery", "recommendation": "lab", "source_headings": [ "Static SQLite / browser-queryable exploratory database" ], "rationale": "An OpenCFB-style static SQLite interface would let large datasets be queried without API credentials. Test the actual use case before committing.", "expected_value": "Ad-hoc exploration beyond what compact JSON surfaces answer.", "data_source_requirements": "Canonical CFB layers packed into a versioned static database.", "deterministic_tool_potential": "Yes — client-side SQL over a dated snapshot.", "validation_requirement": "Demonstrated real queries that the compact outputs cannot serve.", "risks_limitations": "Large static payloads against unknown demand; stats.html already shows the weight of heavy in-browser data.", "tags": [ "CFB", "DATA-PIPE", "EXPERIMENTAL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "evaluating", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "evaluating", "on": "2026-08-08", "note": "Published the 2025 Elo-vs-market probe as blocked; timestamped prospective prices are the named unblocker." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-disagreement.json", "/scripts/cfb_disagreement.py", "/tests/test_cfb_disagreement.py", "/work/mcp-block.js", "/work/test-mcp.mjs" ], "delivery_note": "The lab artifact and its aggregate read-only MCP reader are live, but the headline question remains blocked by historical prices without observation timestamps. The tool returns the blocker rather than a winner or blend.", "graveyard_ready": true, "priority": "high", "roadmap_step": 3, "recommendation_modifiers": [ "high-priority" ], "dependencies": [ "cfb-ratings-registry" ], "candidate_mcp_tools": [ "dd_get_cfb_model_disagreement", "find_cfb_disagreements" ], "components": [], "related_ideas": [ "cfb-consensus-engine", "cfb-model-receipts" ], "governance": [ "cfb-gov-correlation" ], "revisit_conditions": null, "notes": null, "id": "cfb-disagreement-lab", "name": "CFB Model Disagreement Lab", "category": "model-validation", "recommendation": "build", "source_headings": [ "CFB Model Disagreement Lab" ], "rationale": "Compare independent models — SP+, FPI, SRS, Elo, DD EPA, DD predictive model, market — for individual games and teams. The purpose is not showing rankings; it is surfacing meaningful disagreement.", "expected_value": "The CFB counterpart of the live NFL scoreboard: where credible systems disagree is where attention belongs.", "data_source_requirements": "The ratings registry with normalized comparable outputs.", "deterministic_tool_potential": "Yes — descriptive disagreement statistics are deterministic.", "validation_requirement": "Descriptive only until graded: disagreement is not a predictive edge, matching the NFL Pound language.", "risks_limitations": "Correlated systems can agree and still be wrong; the display must not read agreement as confirmation.", "tags": [ "CFB", "MODEL", "MODEL-RISK", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "building", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "planned", "on": "2026-08-08", "note": "Prospective 24-hour market observation contract is staged; first model receipts wait on a 2026 schedule and pre-kickoff forecasts." }, { "status": "building", "on": "2026-08-08", "note": "Published an empty append-only ledger and fail-closed append contract. No receipt is fabricated from completed 2025 games." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-model-receipts.json", "/scripts/cfb_model_receipts.py", "/tests/test_cfb_model_receipts.py", "/work/mcp-block.js", "/work/test-mcp.mjs" ], "delivery_note": "Receipt infrastructure is implemented, the public ledger is honestly empty, and a production read-only status tool reports that state without manufacturing grades. The timestamped market collector is active; the first model receipt still waits on a 2026 scheduled game and a forecast frozen before kickoff.", "graveyard_ready": true, "priority": "high", "roadmap_step": 3, "recommendation_modifiers": [ "high-priority" ], "dependencies": [ "cfb-ratings-registry", "cfb-market" ], "candidate_mcp_tools": [ "dd_get_cfb_model_receipt_status", "get_cfb_model_grades" ], "components": [], "related_ideas": [ "cfb-disagreement-lab", "cfb-consensus-engine" ], "governance": [ "cfb-gov-snapshotting" ], "revisit_conditions": null, "notes": "Consolidates the receipts idea with the Snapshotting / Receipts governance principle: the principle (cfb-gov-snapshotting) states the rule; this entry is the build that implements it for CFB.", "id": "cfb-model-receipts", "name": "CFB Model Receipts / Historical Grading", "category": "model-validation", "recommendation": "build", "source_headings": [ "Model Receipts / Historical Grading" ], "rationale": "Grade model forecasts historically, with forecast snapshots preserved before outcomes so evaluation is genuinely out-of-sample. Extends the snapshotting/receipts discipline already live for NFL models to CFB.", "expected_value": "Spread MAE and RMSE, moneyline Brier score, calibration, ATS versus closing line, total error, performance by season and week, and performance conditional on large disagreement or on market-versus-efficiency splits.", "data_source_requirements": "The ratings registry, the market dataset with closing lines, and append-only frozen forecast snapshots.", "deterministic_tool_potential": "Yes — grading a frozen ledger is deterministic.", "validation_requirement": "Receipts frozen before kickoff, never mixed with backtests — the exact rule the NFL ledger enforces.", "risks_limitations": "Grading against closing lines depends on the market dataset clearing its provenance gate.", "tags": [ "CFB", "MODEL-RISK", "FORECASTING", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 4, "recommendation_modifiers": [ "high-priority", "decision-engine" ], "dependencies": [ "cfb-model-receipts", "cfb-disagreement-lab" ], "candidate_mcp_tools": [ "get_cfb_consensus" ], "components": [], "related_ideas": [ "cfb-ratings-registry" ], "governance": [ "cfb-gov-correlation", "cfb-gov-incremental" ], "revisit_conditions": null, "notes": "Consolidates the consensus-engine idea with the Correlation Awareness governance principle: cfb-gov-correlation states the rule; this entry is the capability that operationalizes it.", "id": "cfb-consensus-engine", "name": "CFB Model Diversity / Consensus Engine", "category": "model-validation", "recommendation": "build", "source_headings": [ "Model Diversity / Consensus Engine" ], "rationale": "Measure correlations among model errors; never treat agreement between highly correlated models as independent evidence. A diversity-adjusted consensus is built only from demonstrated incremental information.", "expected_value": "A consensus that knows how many independent opinions it actually contains — strongly aligned with model-risk management.", "data_source_requirements": "Graded receipts across enough games to estimate error correlations.", "deterministic_tool_potential": "Yes — a declared weighting over registry entries is deterministic.", "validation_requirement": "Any weighted blend must beat equal weight and the best single model on held-out data before it is called consensus — the standing NFL ensemble rule.", "risks_limitations": "Error-correlation estimates need seasons of graded receipts; until then only descriptive summaries ship.", "tags": [ "CFB", "MODEL-RISK", "DECISION-ENGINE", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "live", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Implemented the deterministic Elo baseline and fixed-parameter backtest." }, { "status": "live", "on": "2026-08-08", "note": "Published ratings, calibration bins, market comparison, non-ranked retrodictive team diagnostics and generated model card." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-elo.json", "/data/cfb-model-cards.json", "/scripts/cfb_elo.py", "/tests/test_cfb_elo.py", "/work/mcp-block.js", "/work/test-mcp.mjs" ], "delivery_note": "The baseline is live and reproducible with a production generated model-card reader plus compact expected-versus-observed team diagnostics. Diagnostics are explicitly non-ranked, retrodictive and ungraded; no luck label or prospective 2026 claim is made.", "graveyard_ready": true, "priority": "standard", "roadmap_step": 2, "recommendation_modifiers": [], "dependencies": [ "cfb-games" ], "candidate_mcp_tools": [ "dd_get_cfb_model_card" ], "components": [], "related_ideas": [ "cfb-ratings-registry", "cfb-challenger-lab" ], "governance": [ "cfb-gov-baseline" ], "revisit_conditions": null, "notes": null, "id": "cfb-elo", "name": "Continuous Elo / Glicko Rating", "category": "model-validation", "recommendation": "build", "source_headings": [ "Continuous Elo / Glicko Rating" ], "rationale": "An independently reproducible continuous rating updated after games — an interpretable challenger and baseline, not presumed to outperform other systems.", "expected_value": "The permanent simple benchmark, playing the role 538 Classic Elo plays on the NFL side.", "data_source_requirements": "Canonical games only.", "deterministic_tool_potential": "Yes — pure recurrence over results.", "validation_requirement": "Reproducible from published parameters; graded through receipts like every other registry system.", "risks_limitations": "Deliberately simple: no talent, injury or market information.", "tags": [ "CFB", "MODEL", "FORECASTING" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 8, "recommendation_modifiers": [ "high-priority" ], "dependencies": [ "cfb-plays", "cfb-team-week" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-matchup-fingerprints", "cfb-ratings-registry" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": null, "id": "cfb-unit-ratings", "name": "Opponent-Adjusted Unit Ratings", "category": "model-validation", "recommendation": "build", "source_headings": [ "Opponent-Adjusted Unit Ratings" ], "rationale": "Ratings at the unit level — pass offense, pass defense, rush offense, rush defense, special teams, pressure, pass protection — instead of relying solely on team-level power ratings, translated where appropriate into game-level forecasts.", "expected_value": "The structure the matchup engine reads; team-level ratings conceal exactly the edges this exposes.", "data_source_requirements": "Canonical plays and the team-week layer with opponent adjustment.", "deterministic_tool_potential": "Yes — published unit ratings are deterministic reads.", "validation_requirement": "Unit ratings must demonstrate incremental game-forecast value over team-level ratings before the translation is trusted.", "risks_limitations": "Unit-level samples are small; opponent adjustment errors compound at this granularity.", "tags": [ "CFB", "MODEL", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "eventual", "roadmap_step": 11, "recommendation_modifiers": [ "eventual" ], "dependencies": [ "cfb-team-week", "cfb-model-receipts", "cfb-ratings-registry", "cfb-talent-dataset" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-challenger-lab", "cfb-elo" ], "governance": [ "cfb-gov-baseline", "cfb-gov-incremental", "cfb-gov-model-cards" ], "revisit_conditions": null, "notes": null, "id": "cfb-dd-power-rating", "name": "Data Dawgs Predictive Power Rating", "category": "model-validation", "recommendation": "build", "source_headings": [ "Data Dawgs Predictive Power Rating" ], "rationale": "The house predictive model — developed only after canonical data, receipts and baseline infrastructure exist, and promoted only on demonstrated incremental out-of-sample value over simpler alternatives.", "expected_value": "A Data Dawgs rating in the registry graded on the same receipts as everyone else.", "data_source_requirements": "The team-week layer, talent features, the market benchmark and a working receipts pipeline.", "deterministic_tool_potential": "Yes — published ratings and projections are deterministic reads.", "validation_requirement": "Out-of-sample receipts against market, SP+, Elo and simple EPA regression baselines; a model card before promotion.", "risks_limitations": "The largest overfitting surface on this roadmap; deliberately sequenced near the end for that reason.", "tags": [ "CFB", "MODEL", "FORECASTING" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 12, "recommendation_modifiers": [], "dependencies": [ "cfb-team-week", "cfb-model-receipts" ], "candidate_mcp_tools": [], "components": [ { "name": "Deep / trajectory model", "source_heading": "Deep/Trajectory Model", "summary": "Forecast future team state/feature vectors and simulate games rather than estimating static team strength. Not promoted because it uses neural networks; it must beat reasonable baselines out-of-sample." }, { "name": "XGBoost / other ML challengers", "source_heading": "XGBoost / Other ML Challenger Models", "summary": "Gradient-boosted and similar challengers as potential ensemble components, benchmarked against market, SP+, Elo, simple EPA regression and other simple interpretable models." } ], "related_ideas": [ "cfb-dd-power-rating", "cfb-elo" ], "governance": [ "cfb-gov-baseline", "cfb-gov-incremental", "cfb-gov-model-cards" ], "revisit_conditions": null, "notes": "Consolidation: the deep/trajectory and XGBoost headings become one challenger laboratory; each hypothesis is preserved as a named component with its own benchmark requirement.", "id": "cfb-challenger-lab", "name": "Challenger Model Laboratory", "category": "model-validation", "recommendation": "lab", "source_headings": [ "Deep/Trajectory Model", "XGBoost / Other ML Challenger Models" ], "rationale": "One laboratory for experimental challenger models with two distinct hypotheses: trajectory simulation of future team state, and ML challengers over engineered features. Complexity earns no special treatment in either.", "expected_value": "Occasional genuine challengers and ensemble components; cheap, well-bounded failures otherwise.", "data_source_requirements": "The team-week layer, talent features and graded receipts for honest benchmarking.", "deterministic_tool_potential": "Only for models that survive the lab and publish frozen outputs.", "validation_requirement": "Every challenger is graded through receipts against the standing baselines; failing challengers go to the Graveyard with a postmortem.", "risks_limitations": "Experimental models are explicitly allowed to fail; the risk is letting one ship before it has beaten a baseline.", "tags": [ "CFB", "MODEL", "EXPERIMENTAL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [ "reference" ], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-dd-power-rating", "cfb-challenger-lab" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-training-pack", "name": "CFBD Model Training Pack / Educational Resources", "category": "model-validation", "recommendation": "lab", "source_headings": [ "CFBD Model Training Pack / educational modeling resources" ], "rationale": "Potential reference or educational resource. Do not create a dependency on paid training material if open tooling already provides the required capability.", "expected_value": "Methodological reference while building house models.", "data_source_requirements": "None beyond the material itself.", "deterministic_tool_potential": null, "validation_requirement": "A concrete capability gap that open tooling does not already cover.", "risks_limitations": "Paid-material dependency for capability that may be free elsewhere.", "tags": [ "CFB", "EXPERIMENTAL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 6, "recommendation_modifiers": [ "high-priority" ], "dependencies": [ "cfb-rosters", "cfb-talent-dataset" ], "candidate_mcp_tools": [], "components": [ { "name": "Portal flow network", "source_heading": "Transfer Portal Flow Network", "summary": "Transfers as a directed talent-flow network: net importers/exporters, where talent flows, feeder programs, who attracts high-value portal talent, ecosystem centralization, and how portal acquisition relates to subsequent performance — with graph statistics and null-model comparisons." }, { "name": "Portal net flow", "source_heading": "Portal Net Flow", "summary": "The scalar view of the same ledger: talent gained versus lost per program, eventually integrated with team strength and counterfactual tools." } ], "related_ideas": [ "cfb-acquisition-network", "cfb-counterfactual", "cfb-talent-composite" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": "Consolidation: Transfer Portal Flow Network and Portal Net Flow merge into one canonical entry; both underlying concepts are preserved as components.", "id": "cfb-portal-flow", "name": "Transfer Portal Flow Network & Net Flow", "category": "talent-portal", "recommendation": "build", "source_headings": [ "Transfer Portal Flow Network", "Portal Net Flow" ], "rationale": "The flow network and net-flow metric are two views of one portal ledger: the graph explains structure, the net-flow number feeds team-strength and counterfactual work.", "expected_value": "Structural answers about the portal ecosystem plus a per-program net talent signal.", "data_source_requirements": "Portal entries and destinations keyed to canonical rosters and talent ratings.", "deterministic_tool_potential": "Yes — network statistics and net-flow are deterministic over the ledger.", "validation_requirement": "Null-model comparisons for network claims; net-flow must demonstrate incremental value before entering any forecast.", "risks_limitations": "Portal timing and rating coverage are uneven; per-player value weights are modelled, not observed.", "tags": [ "CFB", "PORTAL", "NETWORK", "TALENT", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 6, "recommendation_modifiers": [], "dependencies": [ "cfb-rosters", "cfb-talent-dataset" ], "candidate_mcp_tools": [], "components": [ { "name": "Recruiting acquisition network", "source_heading": "Recruiting Acquisition Network", "summary": "School ↔ geography/hometown/state recruiting networks: structural footprints, pipelines and their changes through time." }, { "name": "Unified player acquisition graph", "source_heading": "Player Acquisition Graph", "summary": "High-school recruiting and portal acquisition unified into one talent-acquisition view. caleb-chandler/cfb-player-acquisition-network is a methodological reference if still appropriate, not a hard production dependency." } ], "related_ideas": [ "cfb-portal-flow", "cfb-talent-composite" ], "governance": [], "revisit_conditions": null, "notes": "Consolidation: Recruiting Acquisition Network and Player Acquisition Graph merge; the geography network and the unified graph are preserved as components.", "id": "cfb-acquisition-network", "name": "Player Acquisition Network", "category": "talent-portal", "recommendation": "build", "source_headings": [ "Recruiting Acquisition Network", "Player Acquisition Graph" ], "rationale": "The recruiting-geography network and the unified acquisition graph are one capability at two zoom levels; building them separately would duplicate the identity and graph plumbing.", "expected_value": "Where programs source talent, how pipelines shift, and one graph covering both acquisition channels.", "data_source_requirements": "Recruiting classes with geography plus portal moves, keyed to canonical rosters.", "deterministic_tool_potential": "Yes — graph queries over a dated snapshot.", "validation_requirement": "Descriptive structure first; any predictive claim needs the incremental-information test.", "risks_limitations": "Hometown/geography data quality varies; graph aesthetics can outrun analytical content.", "tags": [ "CFB", "RECRUITING", "NETWORK", "TALENT" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 6, "recommendation_modifiers": [ "high-priority" ], "dependencies": [ "cfb-talent-dataset", "cfb-rosters" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-talent-composite", "cfb-counterfactual" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": null, "id": "cfb-returning-production", "name": "Returning Production", "category": "talent-portal", "recommendation": "build", "source_headings": [ "Returning Production" ], "rationale": "Returning production is a first-class preseason and early-season signal, not a footnote inside someone else’s rating.", "expected_value": "A dated returning-production feature per team feeding preseason priors and early-season adjustments.", "data_source_requirements": "CFBD returning production reconciled against canonical rosters and portal moves.", "deterministic_tool_potential": "Yes — published feature values are deterministic reads.", "validation_requirement": "Demonstrated early-season predictive value versus priors without it.", "risks_limitations": "Definition sensitivity: portal-era roster churn changes what \"returning\" means year to year.", "tags": [ "CFB", "TALENT", "FORECASTING", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 6, "recommendation_modifiers": [], "dependencies": [ "cfb-talent-dataset" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-returning-production", "cfb-portal-flow" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": null, "id": "cfb-talent-composite", "name": "Recruiting / Talent Composite", "category": "talent-portal", "recommendation": "build", "source_headings": [ "Recruiting Composite / Talent Composite" ], "rationale": "Maintain talent acquisition as an independent feature family instead of burying it inside a single rating.", "expected_value": "Composite talent features that models can include or exclude explicitly, keeping talent and performance signals separable.", "data_source_requirements": "CFBD recruiting and talent composites, dated per season.", "deterministic_tool_potential": "Yes — feature reads.", "validation_requirement": "Incremental value versus performance-only features before entering any forecast.", "risks_limitations": "Composites embed upstream rating choices Data Dawgs does not control.", "tags": [ "CFB", "TALENT", "RECRUITING" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-talent-composite", "cfb-talent-dataset" ], "governance": [], "revisit_conditions": [ "A lawful structured feed or licensing route appears", "A demonstrated question CFBD data cannot answer", "Vendor terms change" ], "notes": null, "id": "cfb-recruiting-scrape", "name": "Raw 247 / On3 / Rivals Scraping", "category": "talent-portal", "recommendation": "avoid-initially", "source_headings": [ "Raw 247 / On3 / Rivals Scraping" ], "rationale": "Possible incremental information exists, but CFBD already provides cleaned recruiting data. Scraping adds fragility, legal/ToS questions and permanent maintenance for undemonstrated gain.", "expected_value": "Marginal recruiting detail beyond the cleaned composites — currently unproven.", "data_source_requirements": "Would require scraping vendor sites against their terms of service.", "deterministic_tool_potential": null, "validation_requirement": "A demonstrated incremental need that CFBD data cannot satisfy, plus a lawful ingestion route.", "risks_limitations": "Fragile scrapers, ToS exposure and ongoing maintenance — the exact source-risk profile the NFL side declines.", "tags": [ "CFB", "RECRUITING", "SOURCE-RISK" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-portal-flow", "cfb-talent-composite" ], "governance": [ "cfb-gov-uncertainty" ], "revisit_conditions": [ "Verifiable NIL disclosure or reporting standards emerge", "A credible dataset with known coverage appears", "New evidence that even coarse NIL signals carry predictive lift" ], "notes": null, "id": "cfb-nil", "name": "NIL Deal Tracking / Valuation", "category": "talent-portal", "recommendation": "defer", "source_headings": [ "NIL Deal Tracking / NIL Valuation" ], "rationale": "Sparse observations, opaque private deals, inconsistent definitions and self-reported numbers make any precise \"NIL value\" metric false precision today. The hypothesis stays alive; the metric does not ship.", "expected_value": "If defensible data ever exists: a market-structure signal for talent acquisition.", "data_source_requirements": "Currently unavailable at defensible quality; would need consistent, verifiable deal data.", "deterministic_tool_potential": null, "validation_requirement": "Coverage and definition audits before any number is published.", "risks_limitations": "Every input class is unreliable: private, self-reported, inconsistently defined.", "tags": [ "CFB", "TALENT", "MARKET", "SOURCE-RISK" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 5, "recommendation_modifiers": [ "high-priority", "decision-engine" ], "dependencies": [ "cfb-plays", "cfb-team-week" ], "candidate_mcp_tools": [ "get_fourth_down_decision" ], "components": [ { "name": "Decision engine", "source_heading": "Fourth-Down Decision Engine", "summary": "Deterministic game-state answers — \"4th-and-3 at the opponent 41, up four with 8:13 remaining: go, punt or kick?\" — returning the expected result of each choice and the decision advantage, not prose-only advice. Built on the open CFB fourth-down / EPA / WP modeling ecosystem." }, { "name": "Yards-gained distribution model", "source_heading": "Fourth-Down Yards-Gained Distribution Model", "summary": "Supporting model: the existing SportsDataverse multi-class yards-distribution methodology, evaluated for whether it should feed the conversion outcomes inside the engine." } ], "related_ideas": [ "cfb-coaching-scorecards", "cfb-counterfactual" ], "governance": [ "cfb-gov-baseline" ], "revisit_conditions": null, "notes": "Consolidation: the yards-gained distribution model is preserved as a supporting component of the decision engine rather than a separate card.", "id": "cfb-fourth-down", "name": "Fourth-Down Decision Engine", "category": "decision-engine", "recommendation": "build", "source_headings": [ "Fourth-Down Decision Engine", "Fourth-Down Yards-Gained Distribution Model" ], "rationale": "One of the best architectural fits for Data Dawgs: a bounded deterministic calculation an AI can call and a human can check.", "expected_value": "Per-choice expected value and decision advantage for any fourth-down state; the substrate for coaching scorecards.", "data_source_requirements": "Canonical plays with EPA/WP; the open fourth-down modeling ecosystem as scientific substrate.", "deterministic_tool_potential": "Yes — the defining example on this roadmap.", "validation_requirement": "Calibration of the underlying WP/EPA models on held-out CFB seasons; decision outputs benchmarked against the open ecosystem’s published results.", "risks_limitations": "CFB clock rules, pace and talent gaps differ from the NFL; imported assumptions must be re-validated, not assumed.", "tags": [ "CFB", "DECISION-ENGINE", "MODEL", "MCP", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [ "cfb-plays", "cfb-player-game" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-xrepa", "cfb-qbr" ], "governance": [ "cfb-gov-model-cards" ], "revisit_conditions": null, "notes": null, "id": "cfb-cpoe", "name": "CPOE Layer", "category": "decision-engine", "recommendation": "build", "source_headings": [ "CPOE" ], "rationale": "Completion probability over expectation as a player/team analytical layer where useful — a known-good construct ported to CFB data.", "expected_value": "QB and team passing evaluation less polluted by depth-of-target and situation mix.", "data_source_requirements": "Canonical plays with the fields a completion-probability model needs.", "deterministic_tool_potential": "Yes — published CPOE aggregates are deterministic reads.", "validation_requirement": "The completion-probability model gets a model card; CPOE stays descriptive until incremental predictive value is shown.", "risks_limitations": "CFB pass-level data quality varies more than NFL; expected-completion calibration must be checked per season.", "tags": [ "CFB", "MODEL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [ "cfb-plays", "cfb-player-game" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-cpoe" ], "governance": [ "cfb-gov-model-cards" ], "revisit_conditions": null, "notes": null, "id": "cfb-xrepa", "name": "xREPA (Expected Rushing EPA)", "category": "decision-engine", "recommendation": "build", "source_headings": [ "xREPA" ], "rationale": "Expected rushing EPA as an analytical and player-evaluation capability — with a clearly documented model card and definition required from the start.", "expected_value": "Rushing evaluation that separates situation from execution.", "data_source_requirements": "Canonical plays with rushing context fields.", "deterministic_tool_potential": "Yes — published aggregates are deterministic reads.", "validation_requirement": "A documented model card and definition before any number is published; descriptive until incremental value is shown.", "risks_limitations": "Definition drift is the known failure mode for expected-rushing constructs; hence the model-card requirement in the source roadmap.", "tags": [ "CFB", "MODEL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "high", "roadmap_step": 8, "recommendation_modifiers": [ "high-priority", "decision-engine" ], "dependencies": [ "cfb-unit-ratings" ], "candidate_mcp_tools": [ "find_cfb_matchup_edges" ], "components": [], "related_ideas": [ "cfb-fourth-down", "cfb-fraud-detector" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": null, "id": "cfb-matchup-fingerprints", "name": "Matchup Fingerprints", "category": "decision-engine", "recommendation": "build", "source_headings": [ "Matchup Fingerprints" ], "rationale": "Overall ratings conceal matchup-specific structure. Represent strengths and weaknesses across pass offense × pass defense, rush offense × rush defense, pressure × pass protection, explosiveness × explosive prevention, early downs × early downs, red zone × red zone, and havoc × ball security.", "expected_value": "Named, quantified matchup edges for any pairing — the analytical answer to \"where is this game actually won\".", "data_source_requirements": "Opponent-adjusted unit ratings across the fingerprint dimensions.", "deterministic_tool_potential": "Yes — edge-finding over published unit ratings is deterministic.", "validation_requirement": "Edges must demonstrate forecast value beyond team-level ratings before being presented as predictive rather than descriptive.", "risks_limitations": "Depends entirely on unit-rating quality; noisy units produce confident-looking false edges.", "tags": [ "CFB", "DECISION-ENGINE", "MCP", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [ "decision-engine" ], "dependencies": [ "cfb-ratings-registry", "cfb-talent-composite", "cfb-returning-production" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-season-sim", "cfb-portal-flow" ], "governance": [ "cfb-gov-uncertainty" ], "revisit_conditions": null, "notes": null, "id": "cfb-counterfactual", "name": "Counterfactual Simulator", "category": "decision-engine", "recommendation": "build", "source_headings": [ "Counterfactual Simulator" ], "rationale": "Controlled what-if scenarios: different returning production, altered portal additions/losses, injuries/availability where reliable, different game outcomes, different team-strength assumptions — with uncertainty explicit.", "expected_value": "Disciplined counterfactuals instead of vibes: what actually changes if an assumption changes.", "data_source_requirements": "The ratings registry plus the talent/portal feature families being perturbed.", "deterministic_tool_potential": "Yes — declared perturbation in, distribution out, reproducibly.", "validation_requirement": "Every scenario labelled simulation; uncertainty bands mandatory; no counterfactual presented as observation.", "risks_limitations": "Counterfactuals inherit every modelling assumption underneath them; false precision is the standing hazard.", "tags": [ "CFB", "SIMULATION", "DECISION-ENGINE" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "building", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "building", "on": "2026-08-08", "note": "Implemented an exact caller-supplied hypothetical schedule-path distribution plus per-game win-threshold leverage over the dated Elo registry; actual schedules, official standings, tiebreakers and playoff selection remain unbuilt." } ], "implemented": true, "delivery_evidence": [ "/work/mcp-block.js", "/work/test-mcp.mjs", "/data/cfb-ratings.json" ], "delivery_note": "Partially implemented through live dd_project_cfb_schedule_path: exact Poisson-binomial win-count distributions and optional per-game win-threshold leverage for up to 20 caller-supplied matchups using fixed end-of-2025 Elo. Threshold leverage is not conference/playoff leverage; this is not an actual 2026 schedule, conference simulation or playoff model.", "graveyard_ready": true, "priority": "high", "roadmap_step": 9, "recommendation_modifiers": [ "high-priority", "decision-engine" ], "dependencies": [ "cfb-ratings-registry", "cfb-games" ], "candidate_mcp_tools": [ "dd_project_cfb_schedule_path" ], "components": [ { "name": "Season Monte Carlo", "source_heading": "Schedule Path Simulator", "summary": "Monte Carlo the remainder of a season: expected record, conference championship probability, playoff probability, seed distribution and conditional paths — \"what happens to Ohio State’s playoff path if they lose to Oregon?\"" }, { "name": "Game leverage index", "source_heading": "Game Leverage Index", "summary": "Derived capability of the same simulator: which remaining games swing conference/playoff outcomes most, computed by differencing conditional simulations." } ], "related_ideas": [ "cfb-counterfactual", "cfb-prediction-markets" ], "governance": [ "cfb-gov-uncertainty" ], "revisit_conditions": null, "notes": "Consolidation: Game Leverage Index is preserved as a derived component of the simulator rather than a separate card.", "id": "cfb-season-sim", "name": "Schedule Path Simulator", "category": "season-simulation", "recommendation": "build", "source_headings": [ "Schedule Path Simulator", "Game Leverage Index" ], "rationale": "The leverage index is a derived output of the path simulator, not a second engine; one simulation core serves both.", "expected_value": "Playoff odds, seed distributions, conditional paths and per-game leverage from a single reproducible simulator.", "data_source_requirements": "The ratings registry for game probabilities plus canonical schedules and tiebreak/selection rules.", "deterministic_tool_potential": "Yes — seeded simulation with declared inputs is reproducible.", "validation_requirement": "Simulated probabilities carry their input ratings and date; selection-committee modelling is labelled as modelled, not rule-derived.", "risks_limitations": "Playoff selection is a committee, not a formula; that model layer carries irreducible uncertainty.", "tags": [ "CFB", "SIMULATION", "DECISION-ENGINE", "MCP", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [ "add-when-available" ], "dependencies": [ "cfb-market", "cfb-ratings-registry" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-consensus-engine" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-prediction-markets", "name": "Prediction-Market Integration", "category": "season-simulation", "recommendation": "build", "source_headings": [ "Prediction-Market Integration" ], "rationale": "Prediction-market prices as another independent consensus signal — when sufficiently liquid and reliable CFB contracts exist. Market probability is a signal, not truth.", "expected_value": "An additional independent belief in the registry, useful precisely because its errors correlate differently.", "data_source_requirements": "Liquid CFB contracts with dated, attributable prices.", "deterministic_tool_potential": "Yes — dated price reads.", "validation_requirement": "Liquidity thresholds declared before inclusion; thin markets excluded rather than smoothed.", "risks_limitations": "CFB prediction-market liquidity is currently thin and uneven across games.", "tags": [ "CFB", "MARKET", "FORECASTING" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "evaluating", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." }, { "status": "evaluating", "on": "2026-08-08", "note": "Published a results-only record-versus-scoring divergence baseline with one-score records; no predictive labels are permitted yet." }, { "status": "evaluating", "on": "2026-08-08", "note": "A leakage-resistant chronological holdout found a small incremental signal beyond Elo; the one-season retrodictive result keeps the detector evaluating and does not permit team labels." } ], "implemented": true, "delivery_evidence": [ "/data/cfb-record-divergence.json", "/data/cfb-record-divergence-validation.json", "/scripts/cfb_record_divergence.py", "/scripts/cfb_record_divergence_validation.py", "/tests/test_cfb_record_divergence.py", "/tests/test_cfb_record_divergence_validation.py", "/work/mcp-block.js", "/work/test-mcp.mjs" ], "delivery_note": "The descriptive baseline now has a one-season, pregame-only chronological holdout: one Elo-offset coefficient improved Brier by 0.001123 and log loss by 0.002277 over 233 untouched games. A production read-only MCP explorer exposes the dated gaps and aggregate receipt without labels. This remains retrodictive, not market-adjusted or prospective, and does not authorize overrated/underrated team labels.", "graveyard_ready": true, "priority": "high", "roadmap_step": 7, "recommendation_modifiers": [ "high-priority" ], "dependencies": [ "cfb-team-week", "cfb-ratings-registry", "cfb-market" ], "candidate_mcp_tools": [ "dd_find_cfb_record_divergence", "find_cfb_overrated_teams", "find_cfb_underrated_teams" ], "components": [], "related_ideas": [ "cfb-disagreement-lab", "cfb-matchup-fingerprints" ], "governance": [ "cfb-gov-uncertainty" ], "revisit_conditions": null, "notes": null, "id": "cfb-fraud-detector", "name": "Overrated / Underrated Team Detector", "category": "fraud-detector", "recommendation": "build", "source_headings": [ "Overrated / Underrated Team Detector" ], "rationale": "Identify gaps between observed results and underlying performance from record, point differential, EPA, success rate, postgame win expectancy, turnover variance, fourth-down variance, opponent strength, close-game results and market/SP+/FPI/Data Dawgs ratings. The calculation quantifies divergence; \"fraud\" is a brand, not a method.", "expected_value": "Expected record, luck index, underlying performance rank, market over/undervaluation and uncertainty — one of the strongest potential weekly CFB products.", "data_source_requirements": "The team-week layer, ratings registry and market dataset.", "deterministic_tool_potential": "Yes — declared divergence formulas over published inputs.", "validation_requirement": "Divergence must demonstrate forward predictive value (does \"fraud\" regress?) before any claim beyond description; outputs carry uncertainty.", "risks_limitations": "Luck metrics are seductive and noisy at CFB sample sizes; the label must never outrun the calculation.", "tags": [ "CFB", "MODEL", "FORECASTING", "MCP", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [ "cfb-games" ], "candidate_mcp_tools": [], "components": [ { "name": "Coaching trees", "source_heading": "Coaching Trees", "summary": "Coaching movement, ancestry, coordinator relationships, scheme inheritance and tenure as structural data. Interesting; predictive/decision value must be demonstrated." }, { "name": "Movement / performance delta", "source_heading": "Coaching Movement / Performance Delta", "summary": "Performance changes associated with coaching and coordinator transitions, controlling for obvious confounders (talent, schedule, roster churn)." } ], "related_ideas": [ "cfb-coaching-scorecards" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": "Consolidation: Coaching Trees and Coaching Movement / Performance Delta merge into one lab; both hypotheses are preserved as components.", "id": "cfb-coaching-lab", "name": "Coaching Networks & Transition Effects Lab", "category": "coaching", "recommendation": "lab", "source_headings": [ "Coaching Trees", "Coaching Movement / Performance Delta" ], "rationale": "The tree is the structure and the performance delta is the test of whether that structure predicts anything; one lab covers both, sharing the same coaching-history dataset.", "expected_value": "If the delta analysis shows signal, a coaching-transition feature; if not, a documented negative result.", "data_source_requirements": "A coaching/coordinator history dataset keyed to canonical teams and seasons — itself a build.", "deterministic_tool_potential": "Only if promoted: tree queries and transition features would be deterministic reads.", "validation_requirement": "Transition effects must survive confounder controls and the incremental-information test before entering any forecast.", "risks_limitations": "Coaching changes correlate with everything else changing at once; attribution is genuinely hard.", "tags": [ "CFB", "COACHING", "NETWORK", "EXPERIMENTAL" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": 10, "recommendation_modifiers": [], "dependencies": [ "cfb-fourth-down", "cfb-plays" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-coaching-lab" ], "governance": [ "cfb-gov-baseline" ], "revisit_conditions": null, "notes": null, "id": "cfb-coaching-scorecards", "name": "Coaching Decision Scorecards", "category": "coaching", "recommendation": "build", "source_headings": [ "Coaching Decision Scorecards" ], "rationale": "Separate coaching decision quality from realized outcomes across fourth-down decisions, timeout usage, two-point decisions, clock management, aggression and decision EPA/WP. Do not punish a mathematically correct decision because it happened to fail.", "expected_value": "Season-long decision-quality grading per staff — philosophically the most Data Dawgs product on this roadmap.", "data_source_requirements": "Canonical plays plus the fourth-down engine’s per-choice expected values.", "deterministic_tool_potential": "Yes — scorecards are deterministic aggregation over graded decisions.", "validation_requirement": "The underlying decision models must be calibrated first; scorecards inherit and must disclose their assumptions.", "risks_limitations": "Decision EPA/WP models embed assumptions coaches may rationally reject (roster, injury, weather context the model cannot see).", "tags": [ "CFB", "COACHING", "DECISION-ENGINE" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [ "cfb-games", "cfb-team-week" ], "candidate_mcp_tools": [], "components": [ { "name": "Weather effects", "source_heading": "Weather Effects", "summary": "Test whether detailed weather features provide stable incremental forecast value." }, { "name": "Altitude", "source_heading": "Altitude", "summary": "Test the effect instead of assuming its magnitude." }, { "name": "Travel distance / time zones", "source_heading": "Travel Distance / Time Zones", "summary": "Test incremental value after controlling for team strength and venue." }, { "name": "Indoor / outdoor effects", "source_heading": "Indoor / Outdoor Effects", "summary": "Potential interaction with weather and team style." } ], "related_ideas": [ "cfb-availability-lab" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": "Consolidation: Weather, Altitude, Travel/Time Zones and Indoor/Outdoor merge into one contextual-effects lab; each hypothesis is preserved as a separately testable component.", "id": "cfb-context-lab", "name": "Contextual Effects Lab (Weather / Altitude / Travel / Venue)", "category": "context-signals", "recommendation": "lab", "source_headings": [ "Weather Effects", "Altitude", "Travel Distance / Time Zones", "Indoor / Outdoor Effects" ], "rationale": "Four environmental hypotheses sharing one experimental design: engineer the feature, control for team strength and venue, and run the same incremental-information test. A single lab prevents four copies of the same harness.", "expected_value": "Whichever contextual features genuinely survive testing enter forecasts; the rest become documented negative results.", "data_source_requirements": "Venue metadata, historical weather, travel/time-zone derivations joined to canonical games.", "deterministic_tool_potential": "Only for features that graduate.", "validation_requirement": "Each component is separately tested against the no-context baseline; stable out-of-sample lift required for promotion.", "risks_limitations": "Contextual effects are small, collinear and easy to overfit; that is exactly why they are lab work.", "tags": [ "CFB", "EXPERIMENTAL", "FORECASTING" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [ "high-potential" ], "dependencies": [ "cfb-rosters" ], "candidate_mcp_tools": [], "components": [ { "name": "Structured injury / availability data", "source_heading": "Injury / Availability Data", "recommendation": "lab", "summary": "Potentially valuable, but data quality and reliable historical snapshots are major concerns. Explore trustworthy structured sources before relying on anything scraped." }, { "name": "Practice participation / beat-writer proxies", "source_heading": "Practice Participation / Beat-Writer Proxies", "recommendation": "defer", "summary": "Potential signal, but unstructured-data maintenance and historical reproducibility could be expensive. Higher source risk than structured feeds — evidence of incremental value required before productionization." } ], "related_ideas": [ "cfb-context-lab", "cfb-counterfactual" ], "governance": [ "cfb-gov-incremental", "cfb-gov-snapshotting" ], "revisit_conditions": null, "notes": "Consolidation: the two availability headings merge into one lab while their differing source-risk recommendations are preserved per component (structured: lab; beat-writer proxies: defer).", "id": "cfb-availability-lab", "name": "Injury & Availability Signals Lab", "category": "context-signals", "recommendation": "lab", "source_headings": [ "Injury / Availability Data", "Practice Participation / Beat-Writer Proxies" ], "rationale": "Availability is one signal family with two source classes carrying different risk: structured feeds are worth exploring now; beat-writer proxies are deferred until the structured route is exhausted and value is demonstrated.", "expected_value": "If historical snapshots prove reliable: one of the few context signals with plausible material forecast value.", "data_source_requirements": "Structured availability sources with reproducible historical snapshots; snapshots frozen at forecast time, like receipts.", "deterministic_tool_potential": "Only for signals that graduate.", "validation_requirement": "Historical reproducibility audit first — an availability signal that cannot be reconstructed as-of game time cannot be validated at all.", "risks_limitations": "CFB injury reporting is inconsistent by conference and era; retroactive data quietly leaks outcomes.", "tags": [ "CFB", "EXPERIMENTAL", "SOURCE-RISK" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-coaching-lab" ], "governance": [], "revisit_conditions": [ "A cheap reproducible corpus becomes available", "Evidence emerges that sentiment leads rather than follows results", "A specific bounded question (e.g. coaching pressure) shows measurable lift" ], "notes": null, "id": "cfb-sentiment", "name": "Media / Reddit / Social Sentiment", "category": "alternative-signals", "recommendation": "defer", "source_headings": [ "Media / Reddit / Social Sentiment" ], "rationale": "A potential experimental signal for coaching pressure or recruiting momentum, but likely noisy, endogenous, hard to reproduce, expensive to maintain and prone to narrative contamination.", "expected_value": "Speculative; retained as an idea, not a current priority.", "data_source_requirements": "Social/media firehoses with archival for reproducibility — expensive before any value is shown.", "deterministic_tool_potential": null, "validation_requirement": "Reproducible historical corpus plus demonstrated lift over public information already in the registry.", "risks_limitations": "Sentiment often just restates record and market — endogeneity is the default outcome.", "tags": [ "CFB", "EXPERIMENTAL", "SOURCE-RISK" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [ "cfb-player-game" ], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-cpoe" ], "governance": [ "cfb-gov-incremental" ], "revisit_conditions": null, "notes": null, "id": "cfb-qbr", "name": "ESPN Total QBR Historical Layer", "category": "alternative-signals", "recommendation": "lab", "source_headings": [ "ESPN Total QBR Historical Layer" ], "rationale": "Investigate the incremental analytical value and a permitted, reliable ingestion route. Do not duplicate existing open functionality unnecessarily.", "expected_value": "A QB evaluation layer to test against CPOE/xREPA — kept only if it adds information.", "data_source_requirements": "A permitted, reliable QBR ingestion route with dated snapshots.", "deterministic_tool_potential": "Only if adopted: historical reads.", "validation_requirement": "Incremental value over the in-house passing layers before any dependency is created.", "risks_limitations": "Proprietary metric with an opaque method; ingestion permission and stability are unverified.", "tags": [ "CFB", "MODEL", "SOURCE-RISK" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "live", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Captured as the read-only deterministic interface for validated CFB capabilities." }, { "status": "building", "on": "2026-08-08", "note": "Implemented and evaluated canonical game, results-only team-game, compact latest-game, team-period and compact latest-period, and historical market lookup, generated model-card governance, rating-system discovery, bounded single-system rankings, team profile, comparison, hypothetical Elo matchup, exact hypothetical schedule-path, descriptive record-divergence, blocked model-disagreement and prospective-receipt status tools locally; all remain staged and non-callable in production." }, { "status": "live", "on": "2026-08-08", "note": "Deployed sixteen validated, bounded, read-only CFB tools to the production toto Worker as version 4de01c9d-cff6-4234-9be1-3b93df6f49e7." } ], "implemented": true, "delivery_evidence": [ "/work/mcp-block.js", "/work/test-mcp.mjs", "/data/cfb-ratings.json", "/data/cfb-record-divergence.json", "/data/cfb-record-divergence-validation.json", "/data/cfb-market.json", "/data/cfb-model-cards.json" ], "delivery_note": "Sixteen source-backed CFB tools are live in production, including canonical game, results-only team-game, compact latest-game, team-period and compact latest-period, no-CLV historical market lookup, generated model-card governance, rating-system discovery, bounded single-system rankings, exact matchup and schedule-path calculations, plus no-label divergence, blocked-study and prospective-receipt status readers. The production registry now contains 42 live tools total.", "graveyard_ready": true, "priority": "high", "roadmap_step": null, "recommendation_modifiers": [ "foundation", "high-priority" ], "dependencies": [ "cfb-games", "cfb-market", "cfb-ratings-registry", "cfb-disagreement-lab", "cfb-consensus-engine", "cfb-fourth-down", "cfb-matchup-fingerprints", "cfb-season-sim", "cfb-fraud-detector", "cfb-model-receipts", "cfb-team-week" ], "candidate_mcp_tools": [ "dd_find_cfb_games", "dd_find_cfb_team_games", "dd_find_cfb_team_periods", "dd_find_cfb_historical_market", "dd_get_cfb_model_card", "dd_get_cfb_rating_system", "dd_rank_cfb_teams", "dd_cfb_team_profile", "dd_compare_cfb_teams", "dd_project_cfb_matchup", "dd_project_cfb_schedule_path", "dd_find_cfb_record_divergence", "dd_get_cfb_model_disagreement", "dd_get_cfb_model_receipt_status", "get_cfb_model_grades", "find_cfb_disagreements", "get_cfb_consensus", "get_fourth_down_decision", "find_cfb_matchup_edges", "find_cfb_overrated_teams", "find_cfb_underrated_teams" ], "components": [], "related_ideas": [ "cfb-public-outputs" ], "governance": [], "revisit_conditions": null, "notes": null, "id": "cfb-mcp-layer", "name": "Deterministic CFB MCP/API Tool Layer", "category": "ai-interface", "recommendation": "build", "source_headings": [ "Deterministic CFB MCP/API Tool Layer" ], "rationale": "The production product layer: read-only tools that call deterministic Data Dawgs computation, with the AI doing orchestration and explanation — fourteen CFB tools now run alongside the rest of the registry, and future names remain gated on their underlying capability.", "expected_value": "AI-callable CFB answers backed by receipts instead of vibes.", "data_source_requirements": "Each candidate tool depends on its underlying capability existing and being validated first.", "deterministic_tool_potential": "Yes — this entry is the tool layer itself.", "validation_requirement": "A candidate tool goes live only after its underlying computation is built, tested and deployed — the fourteen live readers passed that gate; remaining reserved names stay non-callable.", "risks_limitations": "The temptation is registering tool names before the computation earns them; the Pound exists to prevent exactly that.", "tags": [ "CFB", "MCP", "API", "FOUNDATION", "HIGH-PRIORITY" ] }, { "kind": "roadmap-idea", "domain": "College Football", "lifecycle_status": "idea", "lifecycle_history": [ { "status": "idea", "on": "2026-08-08", "note": "Entered The Pound from the 2026-08-08 CFB roadmap." } ], "implemented": false, "delivery_evidence": [], "delivery_note": null, "graveyard_ready": true, "priority": "standard", "roadmap_step": null, "recommendation_modifiers": [], "dependencies": [], "candidate_mcp_tools": [], "components": [], "related_ideas": [ "cfb-mcp-layer" ], "governance": [], "revisit_conditions": [ "A requirement emerges that hosted models cannot serve (offline operation, private data that cannot leave the machine)", "The existing language layer becomes unavailable" ], "notes": null, "id": "cfb-llm-reports", "name": "Local LLM Generated Reports", "category": "ai-interface", "recommendation": "not-needed", "source_headings": [ "Local LLM Generated Reports" ], "rationale": "A separate Ollama/report-writing architecture duplicates the language layer Claude/GPT/Toto already provide. A deterministic analytical capability is more valuable than another generated summary.", "expected_value": "None over the existing architecture.", "data_source_requirements": "None.", "deterministic_tool_potential": null, "validation_requirement": "Not applicable while redundant.", "risks_limitations": "Builds and maintains a second AI stack to produce prose the current stack already produces.", "tags": [ "CFB", "EXPERIMENTAL" ] } ], "tier_meaning": "The DawgHouse (formerly The Pound) — shelved, with the reason attached.", "built": "2026-08-28", "canonical_url": "https://datadawgs216.com/data/pound-tools.json" }