Transparency
← Back
Transparency
← Back
This page exposes the underlying probability engine: every event we map, its leakage score, the tier classification that drives its payout adjustment, and the latest result of the adversarial bot simulation. Updated automatically as new pitches come in. If the simulation ever fails, paid contests stay locked.
| Tier | Event | Composite | Lift | Ratio | Vol | Acc | Worst context |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
Lift = max P(event | context) − baseline. Ratio = max / baseline. Volatility = stddev across contexts. Accessibility = how easily a bettor can see the leaky context (0 = post-pitch only, 1 = on every scoreboard).
| Dice | Event | Bet | True odds | Observed | Base | Adjusted | Penalty |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
| Bet | Context | Player EV / $1 | Reason |
|---|---|---|---|
| Loading… | |||
A bet is disabled in a specific context only — same bet remains available in other contexts where the expected value is in the house's favor. This is what makes the system regulatory-grade: every accepted bet has been mathematically verified to have negative-EV for the player.
Each pitch fires a single dice roll based on the per-pitch event (ball, called strike, swinging strike, foul, in-play hit, etc.). Cumulative or context-dependent events (strikeout, walk, sac fly, stolen base) are explicitly excluded — the system only uses events observable from a single pitch.
Payouts use the formula Adjusted = Fair × (1 − HouseEdge) × (1 − LeakagePenalty), where the leakage penalty is derived from how predictable the event is given visible game context. Bets that would still be positive-EV after that adjustment are disabled in the specific contexts where the leak occurs.
Three adversarial bots — Threshold Hunter, Optimal Timing, Aggressive Exploiter — replay 50,000+ historical pitches against the live mapping. Pass condition: all three end with negative cumulative expected value. If any bot wins, paid contests stay locked.