TRADING ENGINE
A scanner with a Claude veto on Bybit futures, built up over four levels from a grid proof-of-concept to a reflection loop. The question: can an LLM add edge on top of the scanner? I ran a 90-day backtest to find out.
Evolution
Four levels, each shipped end-to-end: Grid PoC, Scanner, LLM Gate, Reflection Loop.
Scanner
Three setup detectors, running on 4h candles. Score 70+ gets escalated.
Every 4 hours the scanner walks the Top-20 futures market. Setups crossing score 70 get handed to the LLM gate.
LLM Gate
Claude approves or rejects — with specific reasons, not generic caution.
Claude sees the setup, the market context, and recent trade history — and returns a go / no_go with reasoning.
The Decision Layer
The gate never looked for trades — it approved or rejected setups the scanner had already qualified. The system prompt set the default to go rather than caution, and said so plainly.
You are an aggressive crypto trading analyst. Your job:
find reasons to TAKE trades, not avoid them.
The default answer should be "go" unless you find a
SPECIFIC reason to reject.
Decision framework:
- Score 80+: say "go" unless news directly contradicts
- Score 70-79: say "go" if sentiment is neutral or supportive
- Only say "no_go" for SPECIFIC reasons: breaking news,
extreme crowding (funding > 0.05%), or a clear catalyst
against direction.Return · backtest
Profit Factor
Max Drawdown
Win Rate
Reflection
Every closed trade became an atomic note, and the bot re-read its own reflections on the next call.
A trade closes. Claude writes a post-mortem. The learning becomes an atomic note — same shape as the Second Brain vault.
What I Settled On
- →The thesis held, with conditions
Over the 90-day backtest the LLM did add edge — but only when the scanner did the heavy lifting and a default of "go" beat "cautious."
- →Claude judges, it doesn't hunt
It only judged setups the scanner already qualified, never hunted for its own. The worst case was a missed trade, not a bad one.
- →Bybit over Binance
Lower fees on perpetuals and a futures-first API kept the execution layer simple. Taker fee 0.055% was cheap enough that short holds stayed profitable.
- →Post-mortems as atomic notes
Each closed trade produced one learning and one reflection, structured for retrieval. The LLM re-read them on the next decision, so they worked as context, not logs.