Public simulation · real Binance futures prices · no orders, no real money

Does an AI reviewer make a trading system better?

BreakoutBot is a rule-based crypto breakout system with fixed-dollar risk and hard drawdown limits. Two copies of it run side by side on the same prices. One follows its rules alone. The other asks Claude Opus 5.5 to review every full entry first — and skips the ones it vetoes. Everything both of them do is published here, losses included.

the rules alone Without AI Every signal the strategy produces is taken. This is the system as designed — the baseline the AI has to beat. loading… Open the track record → same bot + an AI veto With AI Same code, same settings, same account at the split. Before each full entry the AI reads the candles, the indicators and the bot's own stop and targets, and may veto it. It can remove trades, never add one. Paused on 29 Sep: we found a defect in what the AI was shown. Details below. loading… Open the track record →

With AI is paused — its first four days do not count

From 25 to 29 September the AI was told, at every review, that five positions were open. The With-AI bot actually held none (once, one). The bug was in how we read closed trades back out of the bot's log, not in the model. It vetoed 3 of 9 entries and cited that phantom exposure every time, so these decisions measure our bug, not the AI. They are void and will not count toward the test. The bot was stopped on 29 Sep at 17:22 UTC with a position still open, so its numbers are frozen as they stood.

Where the $23.10 gap came from, to the cent:

−$9.45 from the three vetoes. One avoided a $5.40 loss; two skipped winners worth $14.85.
−$13.23 from two trades only the bot without AI opened, on bars where the AI was never asked. The likeliest cause is data. Each bot fetches its own candles, and we have since measured that a candle read two to three seconds after it closes is often not final yet: in more than half of such reads it still changed afterwards. We have not proven that this is what happened to these two trades, and an earlier veto changing the With-AI bot's state is not ruled out.
−$0.50 in extra probe fees and +$0.08 from fill differences.

Before the reviewer runs again, the input is rebuilt from what the bot actually holds, and a control run with the AI switched off must reproduce the bot without AI trade for trade.

Since the split

Loading…

How the experiment is run

One intended difference. The two bots were split from the same account at the same moment — same balance, same open positions, same risk state. The only designed difference is the AI veto. But each bot fetches its own prices, and a candle read two seconds after it closes is often not final yet, so the paths can also drift apart without the AI being involved. We count that drift separately and do not credit it to the AI or blame it on the AI.
The AI can only say no. It is asked once per confirmed entry. A veto skips that entry; anything else — an allow, a timeout, an outage, a spent budget — lets it through. The With-AI bot can therefore trade less than the one without it, never more.
Decided in advance. The test that decides whether the AI helps was written down before any result existed: a first check at 100 closed positions, a decision at 300, against a random veto of the same rate. Which page is ahead on a given day is not that test — one account path is not evidence.
No look-ahead. The AI only sees candles that had closed when the entry was made. It is not told how anything turned out.
Simulation. Prices are real Binance futures data; entries, fills and balances are simulated. No exchange orders, no real money, no return promises, not financial advice.