NFL Survivor Pup

2026 schedule from nflverse/nfldata · ratings from nfelo · published game forecasts and team ratings mirrored daily · — of 272 games have a published nfelo forecast. Others use nfelo-based season projections.

Season heatmap

Every team, every remaining week, shaded by win probability. marks the optimal path — the assignment of teams to weeks that maximises your chance of running the whole table. is a team you have already spent.

Show only path teams Hide spent teams

Pick popularity

Week

2026 default: modelled ownership. Week 1 blends ETR/Splash projections with ESPN public picks. Other weeks use the chalk model. These are not observed submissions for your pool. Sources and blend method. Optional supplied picks change your board; the public receipt continues to use the declared modelled default. An estimated share cannot establish a measured contrarian edge.

This week's board

Ranked on closed-form equity, not on the simulation. Equity = your win probability divided by how crowded the pool still is if your team wins — so a 70% team nobody owns can beat a 78% team half the field is on.

What the chips mean, and where the lines are drawn

These four are display conventions, not findings. Each is a threshold someone picked on a number the board already shows, so you can see the same thing at a glance. Nothing below is a claim that the cut-off is where the edge changes — move them in your head if you disagree.

Heavy chalk — ownership above: Share of the field on this team. Modelled ownership carries a tilde on the board and this chip inherits that: with no posted data it is a guess about a guess.
Duplication drag — above of your fellow survivors: Of everyone still alive if your team wins, the fraction on your exact team: E × own ÷ survivorsIfWin. Read straight out of the equity denominator rather than computed again, so it can never disagree with the equity column next to it.
Meaningful future cost — above: How much of the best remaining full-season path you give up by spending this team now. The re-solve behind it is the same one the backup cards use. Expect this one to be common early. With a full board of unspent teams almost every good team is wanted by some later week, so most rows clear the line in week 1 and fewer do by week 12. That is the board reporting a real property of an untouched season, not the threshold being set too low — but if it is firing on everything and telling you nothing, it is a threshold, and you may want it higher.
Clean use — below: No other week needs this team, so spending it costs you effectively nothing later.

The optimal path

This is the ceiling, not a plan. The path solve is a Hungarian assignment over log win probability — it is the single best allocation of teams to weeks given today's numbers, double-pick weeks included. Every week that actually gets played moves the ratings and invalidates the tail of it. Use it for the one thing it is good for: telling you which weeks a team is genuinely needed for, so you know what burning it costs.

Ownership is measured against a shrinking field, and that is all it is. A future week's Owned figure is that team's share of the entries projected to still be alive that week — the pool shrinks by Σ own × win% each week, exactly the term the equity calculation already uses, so the two can never drift apart. What it does not do is condition next week's pick distribution on which teams the survivors have already burned. The field here is simulated as a count, not as individual rosters, so it can scale the denominator and nothing else: it cannot say "the entries still alive mostly spent Kansas City in week 3, so they cannot be on it now". That is roster-level conditional ownership, it is a real upgrade, and it is not what this is. Treat future ownership as directional.

Season simulation

Monte Carlo over the rest of the season, with the field modelled as a surviving count. Every candidate is scored against the same drawn outcomes, so the comparison is paired.

What this does not know

Injuries, holdouts, weather, coaching changes. This is a daily mirror, not a live feed; ratings move only after games are played and lines can move after capture.

Future weeks may not have published nfelo forecasts yet. Those games use nfelo’s pre-market formula with current team and quarterback strength and game-specific home-field adjustments, labeled “nfelo-based season projection.” It is not nfelo’s finished game forecast. Published forecasts pass through unchanged; Data Dawgs applies no additional Vegas weighting. Future team and QB strength are held at today’s values; forecasts refresh once daily.

Game outcomes are simulated as independent. Real upset weeks cluster. If anything this makes the board understate how much fading a heavily-owned team is worth.

The field is a count, not 200 rosters. Entries are not tracked individually, so the model cannot see that the field burns its good teams early and gets cornered late. This understates late-season field collapse.

Prospective weekly receipts. The receipts page shows captured claims and completed grades. The public 2026 default uses modelled ownership without weekly data entry. Each week's recommendation is evaluated independently; this is not one entry's tracked season. Calibration remains unproven until enough results exist.

Automation policy · updated September 9, 2026

Forecasts and their inputs refresh once daily at 14:45 UTC (10:45 a.m. EDT / 9:45 a.m. EST). New final scores for unchanged, already-scheduled games can be accepted automatically after validation. Kickoff changes, added or removed games, and score corrections require review. Published claims are preserved. Full policy and limits.

Every rule here changes the simulation, not the display. Saved to this browser and applied to the board the moment you switch back.

Pool shape

Entries in the pool At week 1. Drives how much fading the chalk is worth — at 20 entries it barely matters, at 5,000 it is the whole game. Lives per entry 1 = single elimination. 2 = one strike allowed. More lives make an early contrarian swing much cheaper. Strikes already taken Your own losses so far this season. Current week The week you are picking now.

Rules that change the maths

Teams may be reused Off in almost every pool. On, the future-value term disappears entirely and the board collapses to "pick the biggest favourite every week." Buybacks allowed Eliminated entries can re-enter. Inflates the field and pushes correct play back toward survival. Buybacks permitted through week Ignored unless buybacks are on. Share of eliminated entries who buy back 0 to 1. A guess, and the model is only as good as it. Ignored unless buybacks are on. Double-pick weeks Weeks where the pool requires two picks. A list, a range, or 12+ for "from week 12 on" — because real formats are not always a suffix. Blank = never. Two picks a week is now modelled exactly in the path, the cards and the simulation. Week 18 Week 18 is the motivation lottery: locked seeds rest starters and the lines stop meaning what they mean in September. There is no discount slider here on purpose — a number for "how much less a locked team tries" would be a guess with nothing behind it. Leaving the week unspent says the same thing without inventing one. What the board optimises for Changes the ORDER of this week's board and the backup cards. The season path is maximum-product under both, because optimising a prize share across a whole season is a different objective function and needs its own design, not a toggle. If more than one entry survives Splitting is the common rule and is what the equity numbers assume.

Model settings

Win probabilities: nfeloPublished nfelo game forecasts, with no extra market blend. Future games without a published forecast use nfelo’s pre-market formula with current team and QB strength and game-specific home-field adjustments.

Field chalk exponent Used only when neither supplied picks nor a dated ownership blend covers the selected week. Higher = the field piles onto favourites harder. Week 1 uses the sourced blend instead. Monte Carlo draws The season simulation only. The weekly ranking is closed-form and carries no sampling error at all.

Teams you have already spent

Click to toggle. Spent teams are removed from every future week in the optimal-path solve.

Reset

Wipes every setting on this page back to its default, including spent teams and any pasted pick popularity.