Elo's Hoard · Machine Console

Every pick count is solved to its own return, so they are checked separately — a fault in one table would otherwise hide inside the average. σ is how many standard errors the observed return sits from the designed one, using the exact per-play variance the solver computed for that table — not a guess. Two is unremarkable, four is worth investigating.

This is the real audit. A fair 20-of-80 draw against a fixed set of spots has an exactly known shape — the hypergeometric — and nothing in the game is allowed to bend it. Solid bars are what happened, faint bars are what must happen. Chi-square is the whole comparison in one number; a p-value below 0.01 on real volume means the draw is not dealing straight.

Guests have no name and no row, though their plays are in the ledger above.