Class WalkForwardAnalyzer

java.lang.Object

com.quantfinlib.backtest.validation.WalkForwardAnalyzer

public final class WalkForwardAnalyzer
extends Object

Walk-forward analysis. The series is split into rolling train/test windows; on each fold the parameter grid is optimised on the train window only and the winner trades the following test window. Test segments are stitched into one continuous out-of-sample equity curve.

Continuity across folds. The book is carried from one test window into the next (Backtester.runSegment(TradingStrategy, BarSeries, BacktestConfig, int, Backtester.State, int, boolean)): a position open at a fold boundary stays open and is managed by the next fold's strategy, as it would be live. Only the last fold liquidates at the end of the data. Trade indices are absolute indices into series.

Warm indicators. Each test window is run on the slice [trainFrom, testTo) with trading starting at the test boundary, so indicators enter the test window warm and nothing after testTo is visible to the strategy's init.

Efficiency is objective(stitched OOS curve) / objective(stitched IS curve), where the IS curve chains each fold winner's in-sample per-bar returns. A ratio of whole-curve objectives, rather than a sum of per-fold objectives (which for a Sharpe is not a Sharpe, and is dominated by short noisy windows). NaN when the IS objective is not positive: a ratio of two failures is not a pass. The IS runs are cold (they start on the train window's first bar) while OOS runs are warm; IS and OOS trade counts are reported so the asymmetry is visible.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final record	WalkForwardAnalyzer.Fold
static final class	WalkForwardAnalyzer.WalkForwardResult

Method Summary

Modifier and Type	Method	Description
static WalkForwardAnalyzer.WalkForwardResult	analyze(BarSeries series, ParameterGrid grid, StrategyFactory factory, BacktestConfig config, int trainBars, int testBars, ToDoubleFunction<PerformanceMetrics> objective)	Runs a walk-forward optimisation over the parameter grid.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

analyze

public static WalkForwardAnalyzer.WalkForwardResult analyze(BarSeries series,
 ParameterGrid grid,
 StrategyFactory factory,
 BacktestConfig config,
 int trainBars,
 int testBars,
 ToDoubleFunction<PerformanceMetrics> objective)

Runs a walk-forward optimisation over the parameter grid.

Parameters:

trainBars - bars in each optimisation window

testBars - bars in each out-of-sample window; the window rolls forward by this amount per fold