Research
Research
We pre-register every idea, count every try against one multiple-testing bar, and fill simulated orders the way the market would have. Then we publish what failed with the same care as what held up. Losses first.
Report · v1.0 · 15 pages · free
What Doesn’t Work
5,192 strategy variants tested on 5-minute Bitcoin prediction markets, one registry, a bar of t ≥ 4.27, and none cleared it on data they were not designed on. The method, the failures, the nine bugs that nearly fooled us, and what still holds up.
Read the summary and get the PDFResearch notes
No notes published yet. Tick the box when you request the report and the first one comes to you.