How the R-coefficient works

The R-coefficient measures the historical skill of an insider at picking the right moment to buy their own stock. It is the core metric behind the top-25 ranking.

For each purchase, we look at the return 60 trading days later. If the stock is up, that purchase counts as a win; if down, as a loss. R is the win rate, adjusted for how many trades the insider has made.

A high R does not mean an insider always wins — it means their average win rate is statistically higher than a random guess, given their number of trades.

R = (wins + k·p_global) / (n + k), where:

· wins = purchases with a positive 60-day return; n = total purchases; k = 20 (shrinkage parameter); p_global = global 60-day win rate of all open-market purchases (~58%).

This is Bayesian shrinkage: insiders with few trades are pulled toward the global average, preventing small samples from producing extreme scores.

The window is the last 2 years of purchases (rolling). The ranking shows the most recently active insiders first.

Open-market purchases (transaction code P, acquired/disposed = A) with a price greater than zero, reported on SEC Form 4 filings.

We include both individuals and entities (companies buying back their own stock, funds, holding companies).

Returns are measured against the adjusted close price (splits and dividends adjusted).

Derivative exercises (code M/F), gifts (code G), and non-open-market transactions.

Transactions flagged as price anomalies (deviation >40% from the same-day market close) — likely foreign currency, unadjusted splits or dirty data.

The 10b5-1 scheduled trading flag is not yet exposed (it is on the roadmap).

Form 4 filings must be submitted to the SEC within 2 business days of the trade. We ingest them daily at 06:00 (server time).

Prices are updated daily at 04:00. Everything you see is end-of-day data, never intraday.

Past performance does not guarantee future results. Insider buying is a signal, not a recommendation to buy or sell any security.

Historical prices before recent splits may not be fully adjusted, which can distort old track records.

R measures buying skill, not selling skill, and does not account for portfolio size or context.