Overview
Every trade is recorded with full context: the strategy that generated it, the signals that triggered it, the AI reasoning, entry/exit conditions, and resulting PnL. Analytics aggregate this data into actionable insights.Analytics dimensions
- By Strategy
- By Account
- By Symbol
- By Operation
- By Trigger Type
Compare the performance of different prompts and programs:
| Metric | BTC Momentum (Prompt) | RSI Reversion (Program) | Breakout Hunter (Prompt) |
|---|---|---|---|
| Trades | 89 | 145 | 78 |
| Win Rate | 62% | 55% | 68% |
| Total PnL | $2,100 | $1,200 | $1,223 |
| Sharpe | 1.8 | 1.1 | 2.1 |
AI attribution analysis
For prompt-based strategies, Hyperoru analyzes the quality of AI trading decisions:| Metric | Description |
|---|---|
| Decision accuracy | Percentage of AI decisions that resulted in profitable trades |
| Reasoning quality | Correlation between stated confidence and actual outcomes |
| Context utilization | Which template variables most influenced profitable decisions |
| Model comparison | Side-by-side performance when using different LLM models |
| Regime performance | AI accuracy broken down by market regime |
Program analytics
For program strategies, analytics include execution metrics alongside performance:| Metric | Description |
|---|---|
| Execution time | Average and p99 Python execution duration |
| Decision distribution | Breakdown of hold vs long vs short decisions |
| Factor correlation | Which factors most influenced profitable decisions |
| Backtest alignment | How closely live performance matches backtest results |
Trade replay
Replay individual trades to understand the decision context at the time of execution:Market snapshot
Market snapshot
Price data, indicators, and regime classification at the time the decision was made.
Strategy input
Strategy input
The full prompt (with injected variables) sent to the LLM, or the program input data.
AI response / program output
AI response / program output
The raw decision output including reasoning.
Execution details
Execution details
Order placement timestamps, fill prices, slippage, and fees.
Outcome
Outcome
Position lifecycle from entry to exit, including TP/SL triggers and final PnL.
Trade replay is invaluable for debugging underperforming strategies. Review the AI reasoning alongside the market context to identify patterns of errors.