Live game · Evolution

Marble Race fairness check

Across 220,141 recorded rounds, Marble Race is statistically consistent with random start positions.

Status (all rounds)Consistent
Rounds220,141
Method2: frequencies

All recorded rounds, judged together with every other check on the monitor.

CheckMeasured (95% range)StatedRoundsStatus
All 720 finishing orders720 outcomes, chi-square 710.5 on 719 degrees of freedom p 0.58 Random start positions 220,112 Consistent
Winning colour6 outcomes, chi-square 5.3 on 5 degrees of freedom p 0.38 Random start positions 220,112 Consistent
First and second place30 outcomes, chi-square 48.9 on 29 degrees of freedom p 0.01 Random start positions 220,112 Consistent
Black winsFair: 1 race in 6 16.81% (16.65 to 16.97) 16.67% 220,112 Consistent
Red winsFair: 1 race in 6 16.59% (16.43 to 16.75) 16.67% 220,112 Consistent
White winsFair: 1 race in 6 16.65% (16.50 to 16.81) 16.67% 220,112 Consistent
Yellow winsFair: 1 race in 6 16.72% (16.57 to 16.88) 16.67% 220,112 Consistent
Blue winsFair: 1 race in 6 16.66% (16.51 to 16.82) 16.67% 220,112 Consistent
Green winsFair: 1 race in 6 16.56% (16.41 to 16.72) 16.67% 220,112 Consistent

How we calculate this

Every figure is compared with what random start positions give: each of the 720 finishing orders, each winning colour and each first-and-second pair. Each check has a 95% range and is judged with every other check on the monitor by the same rules.

Every check gets a two-sided p-value. Because 230 checks run on this monitor at once, the p-values are corrected with the Holm method before a result counts as not consistent (corrected p below 0.01). A result with an uncorrected p below 0.01 is shown as borderline. These thresholds are set in advance and are the same for every game. Full method.

Six marbles start each race from randomised positions, so every finishing order is equally likely: each colour wins 1 race in 6, each first-and-second pair comes up 1 race in 30, and each full order 1 race in 720.

Open the Marble Race live tracker