Aether | Market Intelligence

The Evidence Console was sizing overlapping trades off money that had not settled. That is fixed on the live page. The modeled dollars went down. That is what we are publishing. The page is a model of the sealed walk-forward record. It is not a trading account. Every graded trade is a dot. Losses are red. Each dot is that trade’s result in R, a multiple of risk. Those R values come from the sealed books and stay put when the display rules change. The browser turns them into a dollar path from the start capital on the page (these shots use $100,000), the risk per trade (2.0%), the card’s declared leverage band, and the execution stress (as modeled, 0.1% taker per side). Every dollar, every percent per year, and every max drawdown below is that model under those rules. What was wrong. The dollar path compounded trades in the order they were opened, and it applied each result to equity on the open date. That is harmless when trades never overlap. It is wrong when they do. On a 30-day horizon with weekly decision dates, as many as four trades are open at once. The old rule sized week two off week one’s result before week one had settled. The equity base was ahead of reality. All horizons can have seven or eight trades open at once, so the same mistake compounded harder. Later trades were sized off money that had not been made yet. The result of each trade was right. The order the page counted them was not. What is fixed, live. Every trade is sized at entry from equity that has already settled, and the result is recognized when the trade settles. Settlement is the decision date plus the horizon length, the latest it could have settled, because the sealed rows do not carry an exit bar. The new line on the page, “Most trades open at once,” is the check. It is absent on the before shots. The 1-day, 3-day, and 7-day single-horizon views did not move. Decision dates are weekly, so those trades never overlap, and the old rule and the new rule draw the same path. These shots are the All-horizons view, where the overlap is real. Only the 30-day view and All horizons moved. The page now opens on All horizons: the whole record, every horizon lane on the same decision dates, with the risk per trade split across the lanes. A single-horizon landing hid the overlap. The whole record is the first thing a reader should see. Per-horizon views are one click away. Stacking every lane at full risk is not shown as the headline. That drawdown is an upper bound on pain. Full record means bi-directional, longs and shorts, with the designed dynamic sizing: confidence-scaled leverage inside each card’s own band. A trade with no published confidence sizes at the band floor. The sizing map is declared before any result is read. It is not fitted to outcomes. 1x flat is the dashed baseline. Swing / spot is long-only, flat 1x, no leverage. The horizon dropdown is 1d, 3d, 7d, 30d, or All horizons (shared risk budget). True Edge. Full record. All horizons, shared risk budget. Side All. From 2013. Start $100,000. Risk 2.0%. As modeled, 0.1% taker per side. The band runs up to 10x on the highest-confidence calls, across a 13-year record. Before: ending equity $8,509,885,921. 136.0% per year. Return 8,509,785.9%. Max drawdown −43.6%. After: ending equity $1,165,083,864. 102.2% per year. Return 1,164,983.9%. Max drawdown −42.1%. Most trades open at once: 8. The modeled ending equity is about 86% lower. $1,165,083,864 is the model under those rules. Unchanged on True Edge: expectancy 0.237. Total R 615.2. 2,600 trades. Win rate 59.7%. Average sized leverage 4.87x. First image is the page before. Second image is the page after. The old record stays up beside the new one. This is not a claim that the page is finished.