Class DrawdownAnalytics

java.lang.Object
com.quantfinlib.backtest.DrawdownAnalytics
public final class DrawdownAnalytics extends Object
DRAWDOWN structure — because "max drawdown 18%" hides the number that actually fires clients: how LONG the pain lasted. A strategy that loses 18% and recovers in three weeks and one that spends two years under water have the same max drawdown and completely different survival odds. Redemptions, risk-committee reviews and career risk are all functions of drawdown DURATION, not just depth.

The walk: track the running peak; a drawdown episode opens the first period equity dips below it and closes when equity regains the peak (recovery) or the series ends (still open — recoveryIndex = -1, a fact worth surfacing, not hiding: an open drawdown at the end of a backtest is often the honest state of the strategy today).

  • depth — 1 − trough/peak per episode;
  • duration — periods from the peak to recovery (or to the last bar for an open episode);
  • time under water — the fraction of ALL periods spent below the running peak. A strategy under water 60% of the time is painful to hold even when each individual dip is shallow;
  • episodes — the full chronological list, so callers can take the top-k by depth, histogram durations, or line episodes up against market events.

Complements RiskMetrics.maxDrawdown(double[]), which returns only the max depth; the two agree exactly on it (tested). Equity must be positive throughout — a ratio-of-peak drawdown is meaningless through zero or negative equity. Static, deterministic, research lane.

  • Method Details

    • analyze

      public static DrawdownAnalytics.Result analyze(double[] equity)
      Parameters:
      equity - equity curve, ≥ 2 points, all finite and > 0