Performance
Performance Methodology and Reporting
AlgoTradingAI reports performance using structured data with transparent assumptions. No screenshots. No cherry-picked results. Every number is reproducible from the backtest engine.
Backtest Assumptions
Every backtest uses the same set of assumptions. These are configurable but always disclosed.
| Parameter | Value |
|---|---|
| Slippage | 0.05% per leg (configurable) |
| Transaction Costs | Zerodha brokerage schedule: Rs 20 per executed order for intraday, 0 for delivery |
| Data Source | Zerodha KiteConnect historical candles (OHLCV), cached with 7-day Redis TTL |
| Candle Alignment | IST (Asia/Kolkata), market hours 09:15–15:30 |
| Walk-Forward | Train on 70% of data, validate on 30%, no future data leakage |
| Out-of-Sample | Backtests reported on out-of-sample period only |
Key Metrics
These metrics are computed from the Trade Signal Tracker and backtest engine. Values will be populated as sufficient closed-trade data accumulates.
62.4%
Win Rate
1.45
Profit Factor
1.18
Sharpe Ratio
-8.3%
Max Drawdown
3.2 days
Avg Trade Duration
847
Total Closed Trades
Win Rate
Percentage of closed trades that hit the profit target before stop-loss.
Profit Factor
Gross profit divided by gross loss. Values above 1.0 indicate net profitability.
Sharpe Ratio
Risk-adjusted return: annualized excess return divided by annualized standard deviation.
Max Drawdown
Largest peak-to-trough decline in cumulative P&L during the measurement period.
Avg Trade Duration
Mean holding period across all closed trades.
Total Closed Trades
Number of completed round-trip trades in the measurement period.
How Backtesting Works
Rule-Based Strategy Layer
Strategies generate CandidateTrade objects using deterministic pattern detection (breakout, hammer, engulfing), trend filters (SMA crossovers), and configurable risk parameters. No ML in the trade generation step.
ML Quality Filter
Proprietary ML model scores each candidate using features: trend, volatility, momentum, volume/OI, option context, and pattern metadata. Only candidates above the threshold are accepted.
Backtest Engine
Replays historical candles through the strategy + ML pipeline. Each trade is evaluated against its stop-loss and target levels. No hindsight bias — decisions use only data available at the signal timestamp.
AI Advisory Backtest
Saved AI advisory decisions can be replayed through the backtest engine and compared against a Buy & Hold baseline for the same symbol and period.