Historical results by insider signal

Every signal we compute is tested against what the stock did next, across every company and every month we cover. The whole registry is published here, including the signals that do not work and the ones with too little data to judge. Nothing is cherry-picked.

The main figure is the historical share of events with positive excess returns at the selected horizon. Below is the median market-adjusted return. Cells with too few events or too short a history are marked as insufficient.

Backtest hit rateMedian market-adjusted return

Compare each signal with all purchases and the routine-purchase control at the same horizon. These are descriptive groups, not randomized experiments. Differences do not establish causation or show what an investor could have earned after disclosure.

AMF (FR)

Signal	Events	21 days	63 days	126 days	252 days
All purchases (baseline)	14,982	50%0%	49%0%	46%-2%	42%-6%
Cluster buy	5,213	53%0%	54%2%	49%0%	44%-4%
Unusually large buy	1,101	53%0%	49%0%	47%-1%	45%-5%
First buy after 12 quiet months	2,974	52%0%	51%0%	48%-2%	43%-5%
Routine buy (control)	780	53%0%	57%2%	57%2%	45%-2%
High-conviction buy (score 70+)	912	54%1%	54%2%	46%-3%	37%-11%
Cluster sell	1,972	44%-1%	48%-1%	52%1%	49%-2%

SEC (US)

Signal	Events	21 days	63 days	126 days	252 days
All purchases (baseline)	449,767	53%1%	49%0%	45%-3%	40%-10%
Discretionary buy (US, off-plan)	50,521	50%0%	45%-2%	41%-6%	35%-16%
Routine buy (control)	35,847	48%0%	46%-1%

Snapshot of 2026-10-01, methodology v1.1-2026-07. Market-adjusted returns (BHAR) start at the transaction date, not the filing date. They can include performance before public disclosure and do not measure returns achievable by following a published filing. Trading costs are excluded; the mean is winsorised to limit vendor price artifacts.