Methods and code
The scripts behind every headline figure, with the data each study used; gaps are disclosed in each study. There is no name on this site; the scripts are what you check instead.
| Study | Script | Data | Licence |
|---|---|---|---|
| Disposition effect (PGR/PLR) | reproduce.py | Jin, Li & Zhu replication files, PLoS ONE 16(2), 2021 — CC0, Harvard Dataverse | MIT |
| Insider-buying clusters (Form 4) | form4/ | SEC Form 4 bulk files, 81 quarters, 2006–2026 Q1. Not yet published: the cut-point sweep, regime pools and contiguous-window enumeration (disclosed in the study). | MIT |
| Portfolio-driven disposition effect (pre-registered replication; headline corrected 17 Oct 2026) | pdde/ | The same CC0 panel as the disposition-effect study (Harvard Dataverse file 4053809) | MIT |
| Volatility-scaled momentum (pre-registered; negative result) | volmom/ | Kenneth French data library (momentum factor and market data); S&P 500 membership history (fja05680/sp500, MIT); Yahoo monthly closes | MIT |
| Swing-trading strategies (pre-registered; negative result) | To be published with the study; the SHA-256 of every pre-registration, data snapshot and output file is listed in the study | Free vendor-adjusted daily ETF prices, snapshot 16 Oct 2026 | — |
| Combined strategy study: swing, investing, overnight, day trading (pre-registered; negative result) | To be published with the study; the SHA-256 of every pre-registration, addendum, data manifest and output file is listed in the study | Kenneth French data library; FRED DGS10; free vendor-adjusted daily ETF prices; Alpaca SIP one-minute bars for SPY and QQQ, 2016–2026 | — |
| Sample trader report | samples/sample_trader_report.py | The same CC0 panel as the disposition-effect study (Harvard Dataverse file 4053809) | MIT |
| Measurement script for a firm’s own records | partner/ | A five-column extract: one row per position per day. Standard-library Python, no network calls; nothing leaves the machine it runs on. | PolyForm Noncommercial 1.0.0 from v2 (v1 was MIT). Free for non-commercial use; businesses need a licence, included in every engagement. See the partner/ folder. |
Each study whose script is published reproduces from a clean checkout. If it does not, that is a bug, and it belongs on the Corrections page — write to donttrade.research@gmail.com.
Study write-ups and charts are licensed CC BY-NC 4.0. The numbers in them are facts and may be quoted freely, with a link. A figure may be described as “measured by donttrade under a fixed-scope agreement” only when it comes from a signed engagement; no one may describe it as audited, certified or verified.
Open Behavioural Measurement Spec (coming).
Get the next study by email. One email per finished study; free, no advice.
Or follow the Atom feed of new studies and corrections.