2026 So Far: +38.35R, and What Nine Months Taught Us About Risk

Nine months and 257 trades into 2026, SkyAnalyst is up 38.35R. Through Oct 1, 2026, the system has banked +38.35R YTD since the Jan 12 inception. A $100,000 simulated account risking 2% per trade sits at $176,703 on static sizing, from 148 winners against 109 losers, a 57.59% win rate. Every one of those trades is listed in the table below, including the ones we would rather not have taken. The shape of the year is simple to state. The desk went live on January 12 with three trades and spent the first quarter proving that a four-agent process could produce an edge at all. From April through July it accumulated steadily and reached +31.43R by the end of July. August took 2.67R back. September added +9.59R and carried the year past its July high. What the headline does not show is that the two most important things we learned this year both came out of losing stretches, and both were about the same question: at what resolution should a trading desk look at itself? August answered it at the level of a single book. September started to answer it at the level of the whole portfolio. This article is the record of both.

Act I: Proving the process works at all

The first quarter was about whether the system could produce an edge before it was asked to produce a large one. January was a soft launch, three trades from January 12, all winners, +3.02R. February was the first month with a real sample, +6.64R, and the first month where several instrument traders were working the same tape and reasoning about it differently. March was the largest month of the quarter by trade count and finished close to flat.

Flat was the right outcome to learn from at that stage. A break-even month with risk held constant and no single trade larger than its plan is evidence that the risk architecture holds when the tape is not paying. Toward the end of March we consolidated execution onto one primary engine, the one whose analysis had held the most consistent discipline across the quarter's regime changes, and from that point the desk ran on a single, settled process.

Act II: The accumulation

April through July is where the year was made. The desk was not doing anything new in those four months. It was doing the same thing repeatedly, in markets that paid for it: pullback entries in trending tape, continuation reads on the indices, and a Risk Agent that sized every position the same way regardless of how the last one went.

May alone added +8.33R. June produced the best single trade of the year, a NAS100 short off a bounce rejection on June 5 that paid +3.26R on the TP1 baseline, and it is still the best trade on the board nine months in. By July 31 the desk stood at +31.43R across 176 trades, 104 winners against 72 losers.

That number is the anchor of our published record. Every figure in this article builds forward from it, and none of it gets re-derived when the underlying data moves.

Act III: Two losing stretches, two lessons

August reversed the climb. Forty-four trades, 21 winners, -2.67R, and the deepest drawdown of the year, a little over 11R from the late-July high to the mid-August low before a winning run pulled most of it back. The loss was concentrated in one place, a GBPUSD long book that took ten trades, won two and gave back 6.18R, and under the reporting we used at the time that book did not exist as a separate object. On August 27 we split every instrument trader into independent long and short books, and the problem became a line item. We covered it in full in the August monthly recap.

September was the answer to August on the scoreboard and the continuation of it underneath. Thirty-seven trades, 23 winners, +9.59R, and 32 of the 37 entries were shorts. The same directional read showed up on US30, NAS100, US500, EURUSD, GBPUSD and USDCAD. On the days it held, it paid on four instruments in a session. On September 14 it stopped out four shorts in about an hour, and on September 24 it stopped out three more.

The first lesson of the year was that a pooled instrument line can hide a failing direction. The second is that a book of independent-looking trades can be one bet with several tickets. Both were learned in public, and both are in the table below.