📊 Strategy Performance Comparison
| Strategy | Total PnL | Win Rate | Max drawdown | Best Day | Worst Day | Sharpe Ratio | Trades |
|---|---|---|---|---|---|---|---|
| Simple | $-9362.25 | 25.5% | -29.58% | $7670.40 | $-10255.97 | -0.13 | 365 |
| Long Short Best Return | $12229.61 | 49.3% | -38.00% | $7670.40 | $-10255.97 | 0.33 | 365 |
| Threshold | $11092.83 | 1.9% | -2.10% | $4649.83 | $-1810.96 | 1.69 | 365 |
| Realistic | $10691.48 | 45.2% | -41.51% | $7544.58 | $-2344.83 | 0.35 | 365 |
Metrics. Worst trade is the worst single-day return (PnL / price at prediction). Max drawdown is the largest fall of the equity curve that compounds those daily returns from 1.0. The Sharpe ratio is the mean over the standard deviation of the daily returns, times the square root of 365. Total PnL, Avg Win and Avg Loss are in dollars per 1 unit of the asset.
Assumptions.
- Simple: long when the model predicts up, in cash otherwise. It never shorts.
- Long Short: long when the model predicts up, short otherwise. Shorting is not possible on Binance spot, and the funding or borrowing costs of futures or margin are not modeled.
- Threshold: the long-short rule, but it stays out when the predicted move is under 1%. The same shorting and funding limits apply.
- Realistic: the long-short rule with a 0.2% fee charged every day, even when the position does not change (conservative), and a loss capped at 2% of the close-to-close move, because no intraday data is stored and a real stop-loss would fill differently (optimistic). The same shorting and funding limits apply.