What this is: a demo. Each lever's effect is the real measured rate from the 2025 full field. They combine into a raw construction score, which overstates the extremes because the levers aren't independent — so the score is then calibrated against what teams with that score actually did across all 186,832 entries. Raw scores implied top-1% rates well above 100%; the calibrated ceiling is about 30%, which is what the best-constructed rosters really hit.
On ROI and profit odds: expected ROI is the average payout of 2025 entries at each construction score, so it already includes jackpots — a genuine expectation, but a volatile one. The profit odds come from 20,000 simulated seasons that redraw each entry's payout from real 2025 results at its score. Most entries return nothing and a few return thousands, which is why the median season is well below the mean. Calibrated in-sample on one season.
Nine levers are graded: elite-tier RBs, RBs taken early, RB depth, what your QB cost, QBs rostered, tight end plan, QB stack, bye overlap, and value vs ADP (your pick number minus each player’s ADP, summed).
Tier vs. price — keyed per position. Spike supply per player barely moves across QB tiers (1.8× from the top tier to the bottom), so buying a better QB buys no extra ceiling and the only thing that varies is what the pick cost — the QB lever is keyed to price. Spike supply falls 7.7× across RB tiers, 4.7× at TE and 3.8× at WR, so there the tier is the asset and those levers are keyed to 2026 positional rank (Loveland grades as TE3, not by where he’s priced).
Byes are now graded from the official 2026 schedule (Weeks 5–14, none in Week 12). Bye concentration was worth up to +27% in 2025 — weeks are independent, so donating one costs nothing.
How this was built. The coefficients come from my own crawl of all 186,832 entries of the 2025 contest — computed from that data, not estimated. The scoring engine and this interface were built with Claude.
Contact