Questions & answers

Everything is parsed directly from official government sources: SEC EDGAR (Form 4, Form 144, 13F, 13D/G, XBRL financials, 8-K), congressional STOCK Act disclosures, USASpending federal contracts, FEC committee filings, CFPB complaints and NRC reactor data. No third-party resellers, no analyst estimates. If we can't compute it from a filing or a price series, we don't show it.

Form 4 insider trades are polled from EDGAR continuously and appear within minutes of acceptance. 13F fund holdings load per filing the day they are filed, not on the SEC's quarterly bulk schedule. Congressional trades follow the STOCK Act's own cadence: members have up to 45 days to disclose, so that lag is the law's, not ours.

Our price history covers active U.S.-listed tickers from 2005. Delisted names drop out of the price set, which flatters any backtest run only on today's tickers, our backtest pages say this survivorship caveat explicitly instead of hiding it.

Every night we take a random sample of our own rows (insider trades, 13D stakes, Form 144 notices, fund holdings, financial statement facts, CEO pay, trial statuses and prices) and re-read the ORIGINAL filing or source for each one. Last night (2026-10-27) a random sample of 271 figures was re-read against the original filing or source, across 15 data sets: 271 agreed and 0 did not. The desk shows the same count. A mismatch pages us the same morning. Internal checks run alongside, but only a re-read of the source catches a parser that misreads it.

Since the SEC's 2023 rule, Form 4 filings carry a checkbox marking trades made under a pre-scheduled 10b5-1 plan. A scheduled sale says little; a discretionary trade says more. We show the flag on filings from August 2026 onward, older filings are marked unknown rather than guessed.

Three or more distinct insiders buying the same stock inside the window shown on the card. Multiple independent wallets agreeing is the setup most cited in the academic literature on insider trades.

Traded naively, no. Our own backtest over tens of thousands of events shows a negative edge against a small-cap benchmark (the Russell 2000), and we publish that result on the Backtest page. The signal is an input, not a strategy.

Point-in-time honesty: pick a past date and the fundamentals table shows only what had been FILED by that date, later restatements excluded. Most sites backtest against today's cleaned-up data, which quietly inflates results; we built the whole database so it can't.

Usually a definition, not an error. Example: we count only real purchases (Form 4 code P) as insider buying and report option exercises and conversions separately, sites that fold those in show a bigger number for the same company. Where a comparison matters, the page says what is counted.

Yes, everything you can read is free: all datasets, the screener, company pages and the backtester. A free account adds a watchlist (up to 10 tickers), 1 saved screen with alerts in the daily email, and CSV export with a free account: 25 rows a file, 3 files every 30 days. Pro is $22 a month after a 7-day free trial, cancel any time; it removes those limits and adds real-time Smart Alerts. Nothing you already saved is deleted if it is over a limit.

Monthly dollar-cost averaging on dividend-adjusted total-return prices from 2005, benchmarked against the identical cash flows in SPY. The method, including what it does NOT model, is written on the page itself.

Yes. Every JSON endpoint behind the site is open without a key at 240 requests a minute, 100 rows a call; the reference is at /api. A Pro key, created on your account page, raises that to 600 a minute and 50,000 a day, 500 rows a call, with CSV export and no monthly cap.

No. WaveMan surfaces public filings and computed statistics. Nothing on the platform is a recommendation to buy or sell anything.