Running the season 20,000 times…
Used
0 teams
Format
Simulation
20,000 seasons
Your path
| Wk | Pick | Matchup | Line | Win | Public on it | Still alive |
|---|
The whole board
Every team, every remaining week, colored by win probability. Your path is outlined. Hover a cell for the line.
Underdog
Heavy favorite
Your path
Used
How the call is made
A survivor pool is not about surviving. It is about being the last one standing, or one of very few. That means two things matter every week: how likely your team is to win, and how much of the pool goes down with you if it loses. This tool models both.
1. Lines, not opinions
Win probabilities come from the betting market. For the current week, moneylines from Pinnacle, DraftKings, FanDuel, BetMGM, Caesars and BetRivers are de-vigged and averaged, with Pinnacle weighted double because it takes sharp action. For future weeks, SurvivorGrid's consensus look-ahead spreads are converted with a logistic curve calibrated to historical straight-up results (a touchdown favorite wins about 76% of the time). Data refreshes three times a day.
2. Path search
Each team can be used once, so the season is an assignment problem: 18 weeks to 32 teams. The Hungarian algorithm solves it exactly. Maximizing the product of win probabilities is the same as minimizing the sum of −log p, which is the cost the solver sees. A leverage term, λ · log(field survival), is added at three settings (0, ½, 1) to also generate paths that lean on less popular teams. Every team playing this week is forced into slot one, and the rest of the season is re-solved around it.
3. Pool simulation
Each candidate path is played through 20,000 simulated seasons. Every game is a coin weighted by its market probability. The rest of the pool picks teams in proportion to public pick shares, so when a popular favorite loses, most of the field goes with it. If you are the last entry alive you take the pot; if the season ends with several survivors you split it; if everyone left loses in the same week, that week's entrants split. Pool equity is your average share of the pot across those seasons. It is the number that ranks the options.
What is modeled, and what is not
- Public pick shares are real for the current week (Yahoo, ESPN and OFP pools via SurvivorGrid). For future weeks they are projected with a softmax over win probability, fit each build to this week's actual behavior.
- The field is assumed not to be constrained by teams it already used. In practice the public burns its favorites early, which makes late-season contrarian picks slightly more valuable than shown.
- Ties are treated as losses, which is the rule in most pools.
- Two-lives pools are simulated with a strike count for you and for every entry in the field; an entry is out on its second loss. Enter how many entries are already on their last life and the model accounts for a field that thins more slowly.
- Your leagues, picks and locks are saved in this browser. Sign in with an email link to sync them across devices; each account only sees its own leagues. The Copy link button carries a league's full setup for sharing.
- One entry per person. Multi-entry strategy is a different problem.
Sources
- SurvivorGrid for look-ahead spreads and public pick percentages.
- The Odds API for current-week moneylines across books.
- Source code: Python pipeline and tests, JavaScript engine, no framework.
For entertainment. Lines move; check them before you lock in.