New to these reports? Start here
- Dotted-underlined words have a plain-English definition — hover or tap them. Every term is also on the glossary page.
- "Null" means we found nothing, not that something broke. Most reports here are negative results, on purpose — knowing an idea doesn't work is the point.
- Two questions get asked separately. First, is the effect real? Second, is it already priced into the betting odds? An effect can be completely real and still useless to bet on.
- A "calibration" row is a self-check. It runs the same method on something already known to be true. If that fails, the whole report is unreliable — so it's reported alongside the findings.
- If a confidence interval includes zero, the real effect might be nothing at all, so no claim gets made.
NFL survivor pools: which strategy earns the most pot equity? (2026-10-24)
survivor_sim.py (strategies fixed in its docstring before running) →
survivor_sim_results.json, survivor_sim_rows.csv, survivor_actual_paths.csv.
2010-2025, 16 seasons, 1,000 simulated seasons each, outcomes drawn from the
devigged closing moneyline. Rules: one team a week straight up, no reuse,
loss or tie eliminates, last alive wins, simultaneous final eliminations split.
Pot equity is a multiple of the entry fee (1.0 = break-even, zero rake).
| Field | Pool | Greedy (biggest favorite) | Plan ahead (F) | F + avoid popular (C50) | F + avoid popular (C100) |
|---|---|---|---|---|---|
| mild herding (top pick ~42%) | 50 | 1.26 | 1.49 | 1.59 | 1.66 |
| mild herding | 500 | 1.09 | 1.58 | 1.76 | 1.84 |
| heavy herding (top pick ~64%) | 50 | 0.41 | 1.29 | 2.41 | 2.78 |
| heavy herding | 500 | 0.26 | 1.48 | 3.92 | 5.67 |
Season-level: C100 beat greedy in 14 of 16 seasons, F in 12 of 16 (mild field, N=500). F minus greedy: +0.49 (SE 0.16) at mild/500; +1.21 (SE 0.36) at heavy/500.
Reading
- Greedy is what the public does, and it is roughly break-even at best. When the field herds onto one favorite, copying it is badly negative: when that team loses you go out with everyone, and when it wins you share with everyone.
- Planning ahead is the bulk of the edge that needs no pick-% data: saving strong teams for weeks they're big favorites.
- Avoiding popular picks adds more, the more the field herds. The edge comes from surviving with fewer others, not from surviving longer: on what really happened, greedy lasted 4.4 weeks on average, C100 3.1.
Caveats (large)
- ⚠ Superseded on the field: see FINDINGS_SURVIVOR_REAL.md. The real ESPN field herds less (top pick ~29%, beta ≈ 2) and picks worse teams; against it greedy ties the other strategies and every disciplined strategy earns 2-3x.
- The field is modeled, not measured. Rivals pick softmax(beta * logit p) over their unused teams and never plan ahead. No survivor pick-share data was used. Real players save teams somewhat, which would shrink F's edge.
- C uses pick shares from the same rival model the sim plays against, i.e. a perfect pick-% forecast. Real forecasts are noisier, so C's numbers are upper bounds. Weekly public survivor pick % is published by several sites; check terms before collecting any.
- 16 seasons; season-to-season spread is wide (C100 0.96-3.55 at mild/500).
- Pools that allow buybacks, two picks a week late in the season, or pick-the-loser variants are different games; not modeled.
On this page
Related
Survivor strategies vs the REAL ESPN field, 2020-2025 (2026-10-24) $2,500 across 24 small confidence pools (2026-10-24) Do pick'em players get gutsier late, or when behind? (2026-10-19) How well does the public pick? — field baseline, 2014–2025 Winning a 30-Person Confidence Pool Testing "always pick the favorite" — and why a Monte Carlo can't do it Season Prize vs Weekly Prize: Shoot for the Season Best Pick'em Strategy for 2026: Pick Every FavoriteSource
backtests/pickem/FINDINGS_SURVIVOR.md
updated 2026-10-24 18:18