A LOOK BACK · MAY–AUG 2026
ACrypto lets an AI place the trades. After the first stretch — about 11k decision cycles and 154 finished trades — I sat down with the logs. This is the short version: what went well, what got weird, and what I am not ready to claim yet.
Short version: It made money, survived a nasty drawdown, and the plumbing got a lot better. Sustainability is not yet proven — interesting sample, not a victory lap.
Where the account ended up
The ugliest stretch along the way
May–Aug 2026
A bit more wins than losses
01 / THE RIDE
If you only glance at one chart, make it this one. Trade-by-trade scorekeeping told a rosier story than the account curve — fees, partials, and bookkeeping gaps still need tidying.
02 / FAQ-ISH
So far, yes-ish — but I would not call it proven. Finished trades looked pretty good (profit factor 2.83), while the account itself only climbed about 14%. Those two stories do not fully line up yet, so I am not treating the nicer number as gospel.
Not pure coin-flip luck. Dropping the best few trades still left the sample in the green, and 11 of 15 weeks were profitable. That said, a handful of symbols and a few big winners did a lot of the heavy lifting — so luck (or concentration) still matters.
The system got sturdier for sure: better risk rules, sizing, exits, and fewer silly failures. The P&L path was less tidy. The best stretch was an earlier round of risk tightening, not the latest “manage everything actively” version.
Less prompt tinkering, more boring-but-important stuff: cleaner trade books, clearer version tracking, and checks that hold up outside this sample. Until then, the honest take is: promising, not proven.
03 / BY MONTH
| Month | Trades | Return | Win rate | PF |
|---|---|---|---|---|
| 2026-05 | 44 | +51.1% | 56.8% | 5.60 |
| 2026-06 | 40 | +34.1% | 62.5% | 2.27 |
| 2026-07 | 38 | +17.8% | 34.2% | 1.84 |
| 2026-08 | 32 | +6.9% | 84.4% | 11.07 |
Later months won more often, but each win got smaller. Feels less like “alpha exploding” and more like “take profits quicker, sleep a little better.”
04 / VERSIONS
LLM calling more of the shots, wider coin list, simpler stop/take-profit habits.
26 trades · +13.8% · PF 4.11
Best stretch in this sample — clearer entries, less churn, nicer expectancy.
21 trades · +39.6% · PF 6.32
Less long bias, more willing to fade when the market was ugly.
37 trades · +31.9% · PF 2.22
More engineer energy on when and how much to trade. Small sample, mixed results.
10 trades · +5.0% · PF 1.93
More support/resistance and ATR context. Also the soft patch in this review.
17 trades · −5.9% · PF 0.55
Fear/greed, open interest, BTC daily context. Fewer trades, chunkier winners.
12 trades · +20.1% · PF 8.30
Partial closes, trailing, faster risk loops. Lots of little wins, smaller average size.
31 trades · +5.6% · PF 9.16
These eras follow nearby code changes, not a perfect deploy diary. Markets changed too — so the best window is not automatically “credit the commit.”