Split-adjusted EV: the objective that actually pays

Survivor pools look like a survival game and pay like a parimutuel market. The prize is a shared pool: whatever cohort ends the contest alive — or, in a wipeout, whichever cohort dies last — splits it. That single fact breaks the strategy most players follow.

Chalk survives the most and earns the least

In Atropos's 500,000 simulated 2026 World Championship seasons, the strategy with the single highest survival probability — always taking the field's consensus safest picks — reached the end more often than anything else. It was also one of the worst investments in the study: it returned 0.36x the entry fee, because when it survived, it survived alongside ~47 co-survivors making the same picks and split the pool ~47 ways.

The best measured leverage configuration — deliberately taking modestly less-safe, less-popular routes — returned 1.76x. Same contest, same prize pool, roughly 5x the return, and the entire difference is the denominator: how many entries you share the ending with.

The objective, stated precisely

Split-adjusted EV = P(reach a paying outcome) × E[your share of the pool in that outcome]. Both factors matter, they trade off against each other, and the second one is invisible unless you simulate the whole field jointly — every opponent's picks, every week, both endings. Ownership percentages alone don't capture it, because splits depend on which opponents are still alive when the ending arrives, not on week-one popularity.

The correlation trap

The same logic applies to your own entries. Two entries on identical routes are perfectly correlated — they survive together and split together — so the second adds a fraction of the first's value. Correct multi-entry play spreads entries across deliberately diversified routes, which is a portfolio construction problem, not a picking problem.