Upcoming pre-registered reads — the decision calendar
Progress bars = resolved events toward each locked read (n of 12). ≈when for fill pools is a rough cadence estimate recomputed at every build; calendar rows come from their pre-registrations. Every read licenses only what its pre-reg says; nothing adopts automatically.
Every pre-registered crypto experiment — verdict & one-line finding
Sharpe = full-sample net @10bps where applicable. Verdicts locked at commit; the discipline (pre-register, run once, report honestly) is why 16 rejections are as valuable as 3 confirmations. Names never published.
Completed forward months — final abnormal (book minus TOPIX-Large bucket, %)
Backtest reference — exp 421 (2016-2026, net hedged)
What this tab is
Rank the book by
LONG leg — every position, best to worst
SHORT leg — every position, best to worst
Both legs on one axis — cumulative share of positions at or below a given return (ECDF)
Quartile summary — where the middle half of each leg sits
Signal vs outcome — formation rank against what actually happened
Mean realized return by signal quintile — Q1 = strongest score at formation
What this tab is
What this tab is
Measurement harnesses that decide future deployments — health and aggregate readings ONLY. Book contents (names, directions, weights) are never published; each recorder's decision gates were pre-registered before its first datapoint (research/298, /301, /284 §3b).