ai strategy code reviewer

The problem

AI-generated trading strategies often produce impossible backtest results due to lookahead bias and repainting. These scripts frequently lack the safeguards and execution realism required to survive live market conditions without blowing up an account.

What it does

• Audits code for lookahead bias and repainting that invalidates backtest results.
• Identifies state management bugs where position flags fail to reset or update correctly.
• Verifies position-sizing math to ensure risk calculations and lot sizes are accurate.
• Checks for unrealistic order execution like zero-slippage fills or same-bar entries/exits.
• Flags missing safeguards such as drawdown limits, cooldown periods, and input validation.

Frameworks & tools

Pine Script (TradingView), Python (backtesting.py, Backtrader, vectorbt, zipline), C# (NinjaScript), and MQL4/MQL5.

Why this beats prompting it yourself

Standard LLMs often overlook subtle indexing errors or framework-specific quirks like Pine Script's barmerge behavior. This skill applies a rigid, multi-point audit process specifically designed to catch the "too good to be true" bugs that generic prompts miss.

Use cases

• Reviewing a Pine Script strategy before deploying it to a TradingView alert.
• Auditing a Python backtest for hidden future-data leakage in dataframes.
• Validating the risk management logic of an MQL5 Expert Advisor.
• Converting a backtest-only script into a live-ready execution bot.

Known limitations

Cannot verify external data quality or broker-side execution latency. Assumes standard framework behavior unless custom libraries are fully provided.