Decision journal · NFAI engine
Every recommendation, what was done with it, and the realized outcome. The accountability record: whether following the engine helped, measured, not narrated after the fact.
Decisions
2
Followed
2
Ignored
0
Overrode
0
Follow rate
100.0%
Forward return from each decision's reference price to the latest close on July 31, 2026. The engine only reduces exposure, so a call is right when the name fell after. Equal-weighted, not dollar P&L.
Captured alpha
+7.8%
Engine hit rate
100.0%
Left on the table
0.0%
Saved by overriding
0.0%
Graded
2 of 2
8 realized trades: 1 win, expectancy -0.64R/trade. Provisional at 8 of 10 — no class verdict yet (stop 7×-1.01R, manual 1×+1.90R). Expectancy is the mean realized return per unit of planned risk (R). A class needs 5 realized trades before its read is trusted; muted rows are still provisional.
| Exit class | Trades | Win rate | Expectancy | Payoff | Realized |
|---|---|---|---|---|---|
| Stop-out | 7 | 0% | -1.01R | — | −$7,944 |
| Discretionary exitprov | 1 | 100% | +1.90R | — | +$1,003 |
| Allprov | 8 | 13% | -0.64R | 1.89× | −$6,941 |
Read from the realized closed-trade ledger, segmented by how each position ended. Stops cutting losers at ~−1R is the system working as designed; the signal to watch is whether discretionary exits earn a positive expectancy as the sample grows.
PLMRTrimVindicated −3.9%FollowedTrimmed 40.02 of 186 PLMR units at 139.55 (+21.9% on the lot, +$1,002.90 realized) on the engine's over-extension and single-name concentration call. Reduced PLMR from the book's largest position toward target weight while keeping the core.
PLMRRaise stopVindicated −3.9%FollowedRaised the PLMR stop from 101.33 to 114.49 (breakeven) on the remaining 145.98 units, so a winner that ran +21% can no longer round-trip below the entry. Locks the risk to zero on the held lot.
Each decision is logged with the symbol's price at that moment, then graded by its forward return to the latest close. A reduce call is right when the name fell after; following it captures that edge, ignoring it forfeits it. Outcomes are equal-weighted realized returns, not dollar P&L, and a single decision is noise. The signal is the pattern across many, which sharpens as the record grows. Not investment advice.
