WAVEMAN · BACKTEST

Does it actually work?

LIVE

EVENT STUDY · 40,984 EVENTS BENCHMARK Russell 2000 S&P 500

Does insider buying actually make money?

Insider-bought stocks do go up — an average +1.93% over 90 days. That is roughly what the market did anyway. Against a size-matched benchmark the edge is negative to nil. Traded naively and equal-weighted, insider buying is not an edge.

VERDICT · 90D EXCESS · RUSSELL 2000S&P 500

-0.62%-1.31%

Statistically significant, in the wrong direction. t = -4.2-8.8 · win rate 43.8%41.0%.

Raw 90d return+1.93%

Median excess-2.42%-3.36%

Events with full window39,640

The benchmark choice decides the answer.

Insider buying concentrates in small caps. Scored against the S&P 500 the same trades look far worse than against the Russell 2000 — most of the apparent underperformance is the size factor, not the signal. We report both and read the Russell number as the honest one.

vs S&P 500

LARGE CAP · WRONG SIZE BUCKET

-1.31%

vs Russell 2000

SIZE-MATCHED · THE HONEST NUMBER

-0.62%

Every strategy, 90-day excess vs Russell 2000S&P 500

THE FAMOUS FILTERS DO NOT RESCUE IT

Every reported purchasen=39,640

-0.62%-1.31%

Purchases ≥ $250Kn=9,968

-1.49%-2.07%

C-suite ≥ $250Kn=2,818

-2.94%-3.26%

Cluster buys (3+ insiders / 90d)n=14,142

-1.14%-1.83%

SCALE −4% ← 0 → +4%

Strategy breakdown

4 CUTS · 3 HORIZONS

Every reported purchase

Any insider, any size — the widest possible net.

90D EXCESS · RUSSELL 2000S&P 500

-0.62%

-1.31%

HORIZON

EVENTS

RAW

VS SPX

VS R2K

WIN

30d

40,984

+0.42%

-0.34%

-0.08%

45.9%

-1.0

90d

39,640

+1.93%

-1.31%

-0.62%

43.8%

-4.2

180d

37,304

+2.47%

-3.17%

-0.93%

42.1%

-3.3

MEDIAN 90D -2.42% · SIGNIFICANT AT 90D

BY FILING YEAR · 90D EXCESS VS RUSSELL 2000S&P 500

2021-1.68%-5.22%

2022-0.29%+0.16%

2023-0.83%-3.43%

2024+2.51%+0.71%

2025-1.01%+0.16%

2026-5.79%-3.21%

Purchases ≥ $250K

Filter to conviction-sized buys.

90D EXCESS · RUSSELL 2000S&P 500

-1.49%

-2.07%

HORIZON

EVENTS

RAW

VS SPX

VS R2K

WIN

30d

10,382

+0.35%

-0.33%

-0.18%

45.7%

-0.8

90d

9,968

+1.04%

-2.07%

-1.49%

42.4%

-4.4

180d

9,232

+0.89%

-4.37%

-2.47%

39.9%

-3.0

MEDIAN 90D -3.05% · SIGNIFICANT AT 90D

BY FILING YEAR · 90D EXCESS VS RUSSELL 2000S&P 500

2021-3.37%-7.30%

2022-0.76%-0.23%

2023-1.24%-3.88%

2024-0.51%-2.32%

2025-0.51%+0.66%

2026-5.51%-3.00%

C-suite ≥ $250K

CEO / CFO / President / Chair — the best view of the business.

90D EXCESS · RUSSELL 2000S&P 500

-2.94%

-3.26%

HORIZON

EVENTS

RAW

VS SPX

VS R2K

WIN

30d

2,943

+0.02%

-0.70%

-0.55%

45.6%

-2.0

90d

2,818

+0.03%

-3.26%

-2.94%

40.2%

-6.1

180d

2,590

-0.98%

-6.44%

-4.54%

39.0%

-6.0

MEDIAN 90D -4.50% · SIGNIFICANT AT 90D

BY FILING YEAR · 90D EXCESS VS RUSSELL 2000S&P 500

2021-3.17%-5.89%

2022-2.29%-1.70%

2023-1.74%-4.26%

2024+0.94%-0.88%

2025-5.35%-3.99%

2026-6.81%-4.21%

Cluster buys (3+ insiders / 90d)

Several insiders, same name, same window — the literature's favourite.

90D EXCESS · RUSSELL 2000S&P 500

-1.14%

-1.83%

HORIZON

EVENTS

RAW

VS SPX

VS R2K

WIN

30d

14,531

+0.23%

-0.55%

-0.28%

46.5%

-2.2

90d

14,142

+1.52%

-1.83%

-1.14%

45.0%

-5.3

180d

13,354

+2.09%

-3.52%

-1.26%

43.1%

-3.9

MEDIAN 90D -1.91% · SIGNIFICANT AT 90D

BY FILING YEAR · 90D EXCESS VS RUSSELL 2000S&P 500

2021-1.50%-5.63%

2022-1.21%-0.64%

2023-1.87%-4.22%

2024+3.21%+1.56%

2025-2.06%-1.11%

2026-6.61%-3.84%

The famous filters do not rescue it

Cluster buys and C-suite buys are the two setups the literature rates highest. Here cluster buys are roughly flat and C-suite buys are the worst performing group of all — the opposite of the folk wisdom.

It is regime-dependent

2024 was positive across most strategies; 2026 was sharply negative — small caps ran hard that year, so the size-matched bar was unusually high. Any claim that this signal works has to survive the year-by-year cut. This one does not.

Equal weighting is doing real work

Microcaps file far more insider buys than large caps, so an equal-weighted average is dominated by the smallest, least liquid names — before any spread or slippage is charged.

What this is not

This tests one naive rule: buy on publication, hold blind, sell on a fixed clock. It does not test insider data as a feature alongside others, which is how institutional users actually consume it.

Method

Universe

Open-market purchases only (Form 4 code P, non-derivative). Grants, option exercises and sales are excluded.

Entry

The close on the first trading day strictly after the SEC accepted the filing. You cannot trade on a filing you have not seen, and the transaction itself precedes the filing by up to two business days.

Exit

The close 30 / 90 / 180 calendar days later.

Benchmark

SPY and IWM over the identical window. The headline number is excess return.

Incomplete windows

Dropped, not truncated. A half-length window would just record whatever the market did most recently.

No survivorship repair

Delisted and bankrupt names are missing from the price data, which biases every return upward.

BIAS WARNING

Prices cover 76.7% of tickers with insider buys (4,178 of 5,448). Delisted and bankrupt names are missing, which biases every return here upward — the real numbers are somewhat worse than shown.

GENERATED 2026-08-20 · PRICES THROUGH 2026-08-18 · EQUAL-WEIGHTED, NO TRANSACTION COSTS · PAST PERFORMANCE DOES NOT PREDICT FUTURE RESULTS