Calibration Forge

Before trusting a detector, make it earn its alarm.

Each method receives fresh fair histories to measure false alarms, then fresh rigged histories matched to the pattern it claims to detect. Calibration and evaluation simulations never overlap.

100 synthetic draws

Jackpot Triple Play6/46 era · since May 2025100 draws

Frequency χ²

static weighted balls

False alarms 7.8%95% interval 5.5%–10.8%Detection power 74.8%95% interval 70.3%–78.8%

Rank-stream KS

static weighted balls

False alarms 6.5%95% interval 4.5%–9.4%Detection power 53.0%95% interval 48.1%–57.8%

Locked hot strategy

learnable static bias

False alarms 11.0%95% interval 8.3%–14.4%Detection power 37.0%95% interval 32.4%–41.8%

Global e-process

persistent marginal bias

False alarms 6.8%95% interval 4.7%–9.6%Detection power 37.8%95% interval 33.1%–42.6%

Set-kernel MMD

abrupt regime change

False alarms 5.3%95% interval 3.5%–7.9%Detection power 14.5%95% interval 11.4%–18.3%

LZ rank stream

repeated draw motif

False alarms 12.3%95% interval 9.4%–15.8%Detection power 79.3%95% interval 75.0%–82.9%

Independent null calibration and power simulationcalibration-bench-v2seed 20260827400 calibration + 400 evaluation histories per cell

Gold is useful detection power. The outlined meter is false-positive rate; its target is 5%.

400 synthetic draws

Jackpot Triple Play6/46 era · since May 2025400 draws

Frequency χ²

static weighted balls

False alarms 3.0%95% interval 1.7%–5.2%Detection power 100.0%95% interval 99.0%–100.0%

Rank-stream KS

static weighted balls

False alarms 6.8%95% interval 4.7%–9.6%Detection power 97.5%95% interval 95.5%–98.6%

Locked hot strategy

learnable static bias

False alarms 3.5%95% interval 2.1%–5.8%Detection power 76.3%95% interval 71.8%–80.2%

Global e-process

persistent marginal bias

False alarms 10.5%95% interval 7.9%–13.9%Detection power 99.3%95% interval 97.8%–99.7%

Set-kernel MMD

abrupt regime change

False alarms 5.5%95% interval 3.7%–8.2%Detection power 14.0%95% interval 10.9%–17.7%

LZ rank stream

repeated draw motif

False alarms 5.5%95% interval 3.7%–8.2%Detection power 100.0%

Gold is useful detection power. The outlined meter is false-positive rate; its target is 5%.

Every rate here is a proportion out of 400 seeded runs, so it carries a Wilson 95% interval and the interval is what should be read. At fifty runs — the figure this page used to quote — a nominal five per cent false-alarm rate had an interval running from about one per cent to fourteen, and a badly miscalibrated test looked exactly like a perfect one.

The maths, in plain language

The story behind it

Before trusting a detector, make it earn its alarm.

100 synthetic draws

Frequency χ²

Rank-stream KS

Locked hot strategy

Global e-process

Set-kernel MMD

LZ rank stream

400 synthetic draws

Frequency χ²

Rank-stream KS

Locked hot strategy

Global e-process

Set-kernel MMD

LZ rank stream

Calibration before interpretation

The fire alarm gets tested too