{
 "source_page": "/bozo.html",
 "tier": "labs",
 "graded": false,
 "as_of": "2026-08-06",
 "source": "bozo.html ruleset and the Worker-enforced write path, as deployed.",
 "note": "The board is fully open by design — every leg, price and timestamp is visible to everyone and to everyone's bots before they place. That was a decision, not an oversight.",
 "data": {
  "game": "Each member submits ONE leg. Every leg goes on ONE real parlay, funded by last week's bozo. Whoever busts it worst wears it and funds the next ticket.",
  "rules": [
   "Favorites only, inside the league price band.",
   "Every leg must be a real DraftKings selection that is legal in a same-game parlay. Futures are not legal legs — DraftKings will not parlay one with game markets.",
   "No duplicate selections: DraftKings rejects the same selection twice on one parlay, so two players cannot hold it. This is a constraint on the bet, not a league preference, and it is enforced on every league.",
   "No contradicting selections either: two different sides of one market instance can never both cash. Uniqueness does not catch this, because over 45.5 and under 45.5 are distinct selections.",
   "The ticket locks the moment the last leg lands; that moment is the close for the ticket.",
   "Editing a leg resets both your timestamp AND your price.",
   "Only legs that LOST are eligible to be the bozo. Nobody who cashed can wear it."
  ],
  "formats": {
   "standard": "Everyone plays every week. Whoever busts it worst wears it, funds the next ticket, and plays again.",
   "royale": "Bozo Royale — the guillotine. One player is chopped each week, ranked by that week's drawn lever hierarchy. If the ticket cashes nobody goes and the pot carries. RE-DEPLOY: get chopped with one left and you are back on the next ticket automatically — there is no decision to make and no window to miss, which is exactly why it cannot be saved for a convenient week. EVERY player gets exactly one and it always fires: being able to sit out and dodge bozos would be unfair to everyone still playing, so the cost is a price and never an off switch. After that a parachute stays beside your name and the next chop ends your season. Never granted at the final two, or the league could not resolve. The chopped player funds the next ticket either way. Last one standing takes the pot.",
   "immutable": "Format is chosen at league creation and frozen when the league locks its first ticket. Changing a ruleset mid-season retroactively changes who should have been eliminated."
  },
  "parlay_price": "The displayed parlay price is the product of the leg prices. That is price arithmetic, not a correlation-aware joint outcome model — so whenever two legs share an event the ticket is a same-game parlay, DraftKings reprices correlated legs, and the number shown is an UPPER BOUND rather than the payout. It is labelled indicative in that case. DraftKings' real SGP price is not obtainable: it needs their bet-slip pricing endpoint, and DraftKings has no public API.",
  "price_band": {
   "ceiling": -100,
   "floor": -500,
   "unit": "American odds",
   "enforced_by": "Worker, not the page"
  },
  "tiebreak": {
   "mechanism": "Each week the tiebreaker hierarchy is a FRESH RANDOM PERMUTATION of the league's live levers, drawn on the server and written once. Walk the drawn order; the first lever that isolates one person names the bozo; ties fall to the next lever down.",
   "design_rationale": "The randomisation is the design, not decoration: it keeps each lever legible while making the meta unsolvable. A weighted composite was considered and rejected — any blend is just a new deterministic objective someone will solve.",
   "levers": [
    {
     "key": "odds",
     "name": "Shortest Odds",
     "desc": "biggest favorite in the pool"
    },
    {
     "key": "beat",
     "name": "Worst Beat",
     "desc": "finished furthest under its number, in standard deviations"
    },
    {
     "key": "last",
     "name": "Last In",
     "desc": "final leg submitted"
    },
    {
     "key": "clv",
     "name": "Worst CLV",
     "desc": "price moved most against it"
    }
   ]
  },
  "trust_model": {
   "writes": "Every write goes through the Worker, which stamps server time, maps token to player, and validates.",
   "identity": "One-time join tokens. No accounts.",
   "what_is_not_verified": "ENTRY PRICES ARE SELF-REPORTED. Nothing checks what a player typed against a book. Report what was entered, flag what looks off market, never vouch for a number and never accuse anyone. CLOSING prices are different: they are captured at kickoff from DraftKings through a licensed aggregator and stamped with a server clock. So any CLV figure combines one verified number with one unverified one."
  },
  "closing_prices": {
   "captured": true,
   "book": "draftkings",
   "via": "SportsGameOdds (licensed aggregator). DraftKings has no public odds API — no developer portal and no key programme — so a DraftKings price can only be obtained through a reseller. The book that set the price and the route it travelled are recorded in separate fields.",
   "when": "A cron fires in the minutes before kickoff, per event, and writes closeObservedAt from the server clock.",
   "both_sides": "Both sides of the market are captured, because without the opposite side there is no de-vig and an expected-win baseline is wrong by the whole hold — roughly two probability points.",
   "immutable": "A close is written once and never revised. Grading a week later cannot rewrite its prices.",
   "nulls": "No capture means closePrice stays null with a closeUnavailableReason attached. It is never back-filled from the entry price: that would record a fabricated zero movement rather than a missing reading, and drag every average toward it.",
   "coverage_gap": "Every leg goes on a real DraftKings bet slip, so every market exists and every market closes — including player props. What can fail is resolving the free text a player typed onto the odds source's market identifier, which needs the stat, the player and the number all to line up. When one of them does not, the row records which. \"Other\" legs are the genuine exception: they describe an arbitrary game market with no stat, player or number to match on, and stay uncaptured. A missing close is a matching failure or an \"other\" leg — never evidence about a player."
  },
  "known_soft_spots": [
   "Bozo odds and leg-win numbers are SIMULATION output, not observation. Say so when quoting them.",
   "The simulation draws every leg independently. Two legs on the same game are not independent.",
   "CLV is computable only for a leg that has a captured close AND both sides of both prices. Say it is unmeasured for any leg that does not, and never average the two together.", "In the weekly simulation only, a leg whose market CAN be captured but whose close has not landed yet draws a modelled CLV resampled from that player's own past legs, or exactly even when they have no history. A prop or \"other\" leg never does: no close will ever arrive for it, so a modelled one would be invented evidence rather than a prediction. This is simulation input, never stored, and never mixed into a reported mean.", "When no live lever separates the survivors the bozo is drawn uniformly at random from them. It is not the first key in insertion order, which handed every unresolvable tie to the earliest submitter while the verdict said chance decided it.",
   "The Worst Beat lever has no margin to work with on a prop or an \"other\" leg — they are binary, so the leg is given the margin a comparable margin leg would have had: the EXPECTED miss of a leg priced the same way, given that it lost. That is the inverse Mills ratio phi(k)/(1-p) where k = invNorm(p), and p is the de-vigged close if one exists, else the de-vigged entry, else raw implied. This holds in BOTH formats. Busting a -400 prop scores about 1.40 SD and busting a -145 coin flip about 0.90, so the chalkier prop is still the worse beat. Do NOT use invNorm(p) itself: that is the THRESHOLD, the point where the leg sits exactly on its line, and a margin leg that loses is drawn from above it — scoring a binary leg there made a busted prop read as a milder miss than any lost spread priced the same way.",
   "A spread or total priced past about -145 is off market. A moneyline has no internal cross-check at all.",
   "The demo league (demo-2026) is SIMULATED. Every leg, price, close and result in it is fabricated, it uses the real player names on purpose, and it counts toward nothing. The Royale demo (demo-royale) was deleted; Bozo Royale remains a live format."
  ]
 },
 "tier_meaning": "Pup — live and useful, not yet validated. It may compute real answers and still have open questions about calibration, assumptions, data quality or edge. Everything starts here.",
 "built": "2026-09-03",
 "canonical_url": "https://datadawgs216.com/data/bozo-rules.json"
}