Duplicate Poker & GTO Precision Arena — $150 - Sell Project

Online poker has a measurement problem — a winning player needs six figures of hands before results separate them from a losing one.This is a working implementation of the fix. It deals the identical deck to every table in a pod, then scores each player against the cohort who held those exact cards in that exact seat. Card luck cancels out. On top sits a decision-grading layer that scores every action for EV loss and produces a skill rating.Three parts are genuinely non-trivial and are why this is worth buying rather than rebuilding:Provably-fair dealing. Commit → beacon → reveal. The server publishes a hash of its secret seed before any card is dealt, mixes in a public randomness beacon so it can't grind favourable seeds, then reveals the seed afterwards. Players recompute every deal and verify nothing was manipulated. There's a working verifier and a test confirming a tampered seed is rejected.Cohort scoring. Per-deal scoring against same-cards/same-seat cohorts with a variance floor, outlier clamping, and a confidence weight that shrinks scores from small cohorts. Those guards are the difference between a scoring system that works and one that blows up on low-variance deals.Clairvoyance Index. A statistical collusion detector. Duplicate format has a specific hole: since every table holds identical cards, a colluder at another table can hand you opponents' hole cards. The detector runs an OLS regression of a player's behaviour against hidden opponent strength, controlling for legitimate signals. A player who can't see hole cards has no lawful way to move that coefficient off zero. It separates a clairvoyant agent from an honest one, and refuses to score at all below its evidence threshold rather than guessing.Runs single-player out of the box — the cohort is simulated, so it's fully demonstrable without a player base. One command, no external services.