LLM-Directed Crypto Futures Trading with Performance-Based Risk Controls
Summary
This workflow describes automated cryptocurrency perpetual-futures trading that delegates per-asset decisions to a large language model. An hourly process gathers market indicators, funding rates, positions, account state, and historical trading performance, then requests structured signals that include direction, leverage, risk amount, and stop and target levels. A separate frequent monitoring loop checks open positions and closes them according to either fixed-percentage exits or the targets specified at entry. Trades and status information are stored for later decisions and dashboard display.
Risk controls include temporarily freezing assets after a run of recent losses, adjusting risk allocation according to historical results, considering long and short performance separately, and calculating order size from risk, stop distance, and contract value. The source also contains extensive model instructions, but the supplied material does not report audited live or backtest performance. Results would depend on the model’s decisions, exchange execution, fees, slippage, data quality, and implementation details; the described workflow is not evidence that its adaptive rules improve returns.
Key ideas
- An hourly workflow supplies market, account, position, and historical performance data to an LLM for per-asset decisions.
- A separate frequent loop monitors open positions and applies configured exit rules.
- Recent consecutive losses can freeze trading for an asset for a defined cooling period.
- Risk allocation and directional bias are adjusted using historical asset and side-specific performance.
- Position quantity is calculated using risk amount, stop distance, and contract value, but no performance evidence is provided.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.