Whenever we change the model or an execution rule, we record the date and its R impact here, every time. The figures in each entry are as recomputed on that date; the headline figures are recomputed daily, so the two differ slightly.
Project O Nasdaq update (not a change to the existing method, but a separate principle combined with it). Portfolio 711.3 to 739.6R, max drawdown 46.3 to 42.2R. Applied without a shadow period, and adopted although it fell slightly short of one pass criterion (the random comparison).
Project O (not a change to the existing method, but a separate principle combined with it), a change to the gold model, data correction, 1R $500. Portfolio 426.9 to 711.3R, max drawdown 33.7 to 46.3R. The shadow period was one to two days, and none for bitcoin.
Cost calculation fixed - some trades were missing their fees; the history was re-computed. Bitcoin moved from 124.8 to 124.6R (win rate 32.1% to 26.7%: trades that turn into small losses after fees now count toward the win rate) and the portfolio from 427.6 to 427.3R - against us.
We recomputed the dollar simulations on whole contracts, the same basis the app uses. $60,000 of own capital, 5-year typical: +$54,483 to +$50,133; small-capital default path, 3-year typical net profit: +$14,301 to +$16,319.
We updated the model to make it more robust. The re-computation moved the portfolio from 482.4 to 428.1R (max drawdown 27.0 to 33.7R). Effective September 27.
The open trade excluded on 9/21 (the 9/20 Bitcoin long) closed on 9/22 and entered the record as realised +46.6R - Bitcoin 100.1→145.5R. Neither the method nor the counting rule changed; the 9/21 rule counts only finished trades, and this one finished. Its exit card was first scored on the price at send time (46.77R) and then corrected to the 5-minute close (46.64R); the card shows the correction and the original value.
Aggregation fix - trades that have not finished are no longer counted in the performance figures. The Bitcoin exit rule resolves over the 24 hours after entry, and when the daily re-computation ran before those 24 hours were up, it scored the open trade as if it had closed at the price at that moment. The trading method itself is unchanged; only the aggregation was corrected. The re-computation moved Bitcoin back from 141.2 to 100.1R (reward-to-risk 7.05 to 4.99) - the value it had held for ten days; the previous display included roughly 41R of unrealised profit from a position that was still open. It moved in our favour this time, but it can move against us by the same amount, so it is excluded. Nasdaq and gold are unaffected because their sessions close within four hours (171.6R and 166.6R unchanged).
Execution rule added (gold). The re-computation (8da05766) moved gold from 156.4 to 168.1R (max drawdown 34.4 to 25.8R, 576 to 548 trades) and the portfolio from 426.5 to 438.1R (max drawdown 31.2 to 27.0R). The full-history max drawdown of a Nasdaq-plus-gold account fell from 45.1 to 36.5R.
Execution rule added (Nasdaq). The re-computation moved Nasdaq from 178.6 to 169.9R (max drawdown 39.4 to 20.4R, 908 to 747 trades) and the portfolio from 437.4 to 428.8R (max drawdown 52.8 to 31.2R) - less cumulative R and a shallower drawdown.
Early-close and holiday sessions are no longer traded; the backtest now excludes those days too, matching live policy. Nasdaq moved from 173.7 to 178.6R (8 early-close trades worth +2.5R dropped, plus the knock-on re-computation) and gold from 153.4 to 158.7R (3 trades worth -5.3R dropped) - in our favour.
Bitcoin execution consolidated to a single session - during the initial live window (July 11-14) two copies of the sender ran twice a day. That is why the public ledger shows entries outside today's schedule in that window.