Annualized Sharpe Ratio (sharpeRatio × √tradesPerYear). Higher is better.
Average sum of pnlPercentage across consecutive losing streaks. Null if no loss streak. Closer to 0 is better.
Average sum of pnlPercentage across consecutive winning streaks. Null if no win streak.
Average trade duration in minutes ((closeTimestamp - pendingAt) / 60_000).
Average fall PNL percentage across all trades (_fall.pnlPercentage). Closer to 0 is better.
Average loss percentage on losing trades
Average duration in minutes of losing trades.
Average peak PNL percentage across all trades (_peak.pnlPercentage). Higher is better.
Average PNL per trade
Average profit percentage on winning trades
Average duration in minutes of winning trades.
Fraction of up-moves among decisive close-to-close moves. 0..1. Higher = buyers more frequent.
Share of upward absolute movement in total close-to-close movement. 0..1.
Calmar Ratio (totalPnl / maxDrawdown). Higher is better.
Certainty Ratio (avgWin / |avgLoss|). Higher is better.
Expectancy: (winRate * avgWin) - (lossRate * avgLoss)
Expected yearly returns (geometric, capped at ±MAX_EXPECTED_YEARLY_RETURNS). Higher is better.
Number of losing trades
Maximum drawdown percentage (largest peak-to-trough decline)
Minimum fall PNL percentage observed across all trades (worst worst-case). Closer to 0 is better.
Maximum consecutive losing trades
Maximum consecutive winning trades
Median pnlPercentage — robust to outliers; reveals distribution skew when paired with avgPnl.
Median |close[i] - close[i-1]| / close[i-1] across trade closes, in %. Robust to outliers.
Maximum peak PNL percentage observed across all trades (best best-case). Higher is better.
buyerStrength - sellerStrength ∈ [-1, 1]. Positive = bullish bias on magnitude.
Profit factor: sum of wins / sum of losses
Recovery Factor (totalPnl / maxDrawdown). Higher is better.
Fraction of down-moves among decisive moves. 0..1. Equals 1 - buyerPressure.
Share of downward absolute movement in total close-to-close movement. 0..1.
Risk-adjusted return per trade (Sharpe Ratio = avgPnl / stdDev)
Sortino Ratio (avgPnl / downside deviation — RMS of losing trades only). Higher is better.
Standard deviation of PNL
Trading pair symbol (e.g., "BTCUSDT")
Total profit/loss percentage across all closed trades
Total number of closed trades
Observed trade frequency extrapolated to one year (signals × 365 / calendarSpanDays).
Bivariate trend classification (slope × R²).
R² of the log-price regression, in [0, 1].
Log-price regression slope, in %/day.
Number of winning trades
Win rate percentage
Portfolio heatmap statistics for a single symbol. Aggregated metrics across all strategies for one trading pair.