← All guides
Analysis
Multiple testing: seeing patterns everywhere
Look at enough statistics and some of them will look remarkable. That is not evidence — it is arithmetic. Multiple testing is the name for the trap, and correction is the way out.
The trap
A 6/49 game has 49 numbers, 1,225 pairs, and 18,424 triples. If each one is 'tested' for unusualness, chance alone guarantees that several clear any reasonable threshold. The most-drawn number, the rarest pair, the longest gap — all are expected to be extreme somewhere.
The correction
When testing many hypotheses at once, the significance bar must be raised (e.g. Bonferroni: divide α by the number of tests). The prediction lab applies this before flagging anything, and the reality check always compares observed values to the range produced by simulated random history.