NFL Survivor Pool Picks

How the survivor pick model works

Every team's chance to win comes from two places. Three quarters of it is the sportsbook moneyline with the vig removed, because sixteen seasons of public survivor picks show the closing line prices survival to within half a point. The last quarter is our drive-level game sim, which is the same engine behind the moneyline and total bets on the Monte Carlo page. Weeks that have not been priced yet fall back to our season power ratings with home field, and get replaced by the market as lines post.

Then every team is priced against the field. The EV number answers one question: if this team wins, what share of the pool is still alive next to you? A 70% team that 5% of the pool is on usually beats an 80% team that 40% of the pool is on, because the second one only moves your equity when it loses. A 1.00 means you gain nothing on the field by surviving.

The cost to burn is the price of using a team now, not a measure of the team. The model plans the next eight pick slots with a max-weight matching, once with the team available and once without, and the drop in the plan's log-probability is the cost. A 0.00 means the plan is the same either way: every later slot that team could fill already has a bigger favorite in it, on today's lines. Treat it as a low cost, not a free pick. Lines move, favorites get hurt, and a team like that is the fallback the plan does not price. Bergman and Imbrogno showed in Operations Research that a rolling eight-week horizon beats both picking week to week and planning the whole season at once.

Splash Sports versus Circa rules

Splash counts Thursday through Monday as one week and thins the field with double-pick weeks, where both teams have to win. The $21 million contest doubles up in weeks 9, 12, 13, 14, 15 and 16, so Thanksgiving and Christmas both fall on double weeks. Circa instead splits the Thanksgiving and Christmas games into standalone legs with a short menu of teams. Both treat a tie as a loss. Pick the contest at the top of the page and the planner rebuilds around those slots.

Saved picks and multiple entries

Lock a pick and it is saved to your account, removed from the board, and factored into every later recommendation. Add entries with different names to run several Splash entries side by side. Picks grade themselves once the final score posts.

The multi-entry optimizer

Ten entries on the same team is one bet, not ten. The optimizer simulates every game of the week thousands of times, jointly, so entries that share a team live or die together and the field's survival moves with the same results. For each entry it knows the teams already burned, so entries with different histories get different menus. It then searches for the allocation that maximizes one of three objectives: the expected share of the pool, a balanced score that pays for a better worst case, or the chance that at least one entry survives.

In a pool of twenty thousand your entries are a rounding error in the denominator, so Max EV usually stacks every entry on the best team. In a pool of a hundred your own survivors crowd the denominator, and the same objective starts to spread. The comparison table shows all three answers plus the all-in baseline, with the 10th and 90th percentile week for each, so you can see exactly what diversifying costs and buys.

Frequently asked questions