⚽ Tomulator
Overview
β—ˆDashboardπŸ“…Today's gamesβ—ŽState of play
Money
πŸ’·Bets & strategies
Explore
⊞Leagues & seasonsπŸ“Metricsβ–€Data sources
Engine room
🎲SimulationsπŸ§ͺLab≣Catalog?How this works

matrix backtest: engine x rule x staking, per-season $1000 replays

← lab Β· 39 engine Γ— rule Γ— staking combos, each season from $1000 Β· ran 7/4/2026

What this is: The tournament: every prediction engine Γ— every betting rule Γ— every staking policy, replayed season by season with each season restarting from $1000. Finds which COMBINATIONS survive, not just which models predict.
Every engine Γ— bet-rule Γ— staking combination replayed season by season, each season restarting from $1000. Progression staking (martingale/fibonacci/paroli) inflates medians and hides ruin β€” always read the worst-season column. Target-chase staking STOPS at its target ($2000 for target_chase_2x), so "median $2000" means "over half the seasons doubled then quit betting", not "typically doubles" β€” the mean is the honest expectation there. Poisson engines carry only 4 stored seasons; 21-season GBM cells are the trustworthy ones.
EngineRuleStakingMedian endMean endWorst seasonWinning
Poisson sim #17 β†’draws onlyflat 10$1087$1069$8032/4
Poisson sim #17 β†’underdogs 3plusflat 10$999$864$1762/4
GBM model #16 β†’favorites sub 1 8kelly quarter cap2$996$963$68410/21
GBM model #16 β†’favorites sub 1 8flat 10$994$977$76410/21
Poisson sim #17 β†’favorites sub 1 8flat 10$986$985$9581/4
GBM model #16 β†’underdogs 3plusflat 10$986$862$10710/21
Poisson sim #17 β†’favorites sub 1 8kelly quarter cap2$984$978$9291/4
Poisson sim #17 β†’favorites sub 1 8kelly half cap5$967$950$8272/4
Poisson sim #17 β†’draws onlykelly quarter cap2$965$1034$8212/4
GBM model #16 β†’favorites sub 1 8kelly half cap5$935$910$4128/21
GBM model #16 β†’draws onlyflat 10$926$911$4645/21
Poisson sim #17 β†’draws onlykelly half cap5$918$1106$6632/4
GBM model #16 β†’edge 10pctflat 10$887$879$2636/21
GBM model #16 β†’edge 5pctflat 10$885$850$2686/21
GBM model #16 β†’totals only edge3flat 10$860$853$6351/7
GBM model #16 β†’edge 2pctflat 10$859$819$1127/21
GBM model #16 β†’draws onlykelly quarter cap2$822$849$4855/21
GBM model #16 β†’underdogs 3pluskelly quarter cap2$791$873$3318/21
GBM model #16 β†’totals only edge3kelly quarter cap2$783$760$5071/7
Poisson sim #17 β†’edge 5pctflat 10$759$757$4801/4
GBM model #16 β†’edge 10pctkelly quarter cap2$754$842$3054/21
Poisson sim #17 β†’edge 2pctflat 10$742$774$3471/4
Poisson sim #17 β†’edge 10pctflat 10$704$723$5760/4
GBM model #16 β†’edge 2pctkelly quarter cap2$686$777$3304/21
GBM model #16 β†’edge 5pctkelly quarter cap2$655$785$3325/21
Poisson sim #17 β†’underdogs 3pluskelly quarter cap2$643$652$4950/4
GBM model #16 β†’draws onlykelly half cap5$601$658$1963/21
Poisson sim #17 β†’edge 10pctkelly quarter cap2$581$546$3340/4
Poisson sim #17 β†’edge 5pctkelly quarter cap2$569$536$3140/4
Poisson sim #17 β†’edge 2pctkelly quarter cap2$567$531$3220/4

39 cells total Β· leaderboard shows the top by median season end.

Reading the columnswhat each number actually means

Median endmedian season-ending bankroll (start $1000). CAUTION: target-chase staking stops at its target, so when most seasons hit the stop the median pins to exactly the target β€” read the mean beside it
Mean endaverage season-ending bankroll β€” the honest expectation; losing seasons drag it down where the median can't see them
Worst seasonthe ruin check β€” progression staking looks great until this column
Winningseasons that ended above $1000 / seasons replayed
Stakinghow bet size was chosen; same rule + different staking = wildly different fates
Spec Β· the reproducible recipe
{
  "kind": "matrix_backtest",
  "rules": [
    "edge_2pct",
    "edge_5pct",
    "edge_10pct",
    "underdogs_3plus",
    "favorites_sub_1_8",
    "draws_only",
    "totals_only_edge3"
  ],
  "run_ids": [
    16,
    17
  ],
  "staking": [
    "flat_10",
    "kelly_quarter_cap2",
    "kelly_half_cap5"
  ]
}