Walk-Forward Validation

Walk-forward validation framework for trading strategies and ML models. Standard cross-validation (k-fold, random splits) fails catastrophically for financial time series because it introduces lookahead bias and ignores autocorrelation. This skill covers proper time-series validation techniques including rolling and expanding windows, purged cross-validation, combinatorial purged cross-validation (CPCV), and overfit detection metrics.

Why Standard Cross-Validation Fails

Standard k-fold CV assumes data points are independent and identically distributed (IID). Financial time series violate both assumptions:

1. Lookahead bias — Random splits let the model train on future data and predict past data, artificially inflating performance.
2. Autocorrelation — Adjacent observations are correlated. A random split that puts Monday in test and Tuesday in train leaks information.
3. Regime dependence — Markets shift between regimes. A model trained on a bull market and tested on a bull market tells you nothing about bear market performance.
4. Label overlap — If labels are computed over windows (e.g., 24h forward return), adjacent train/test samples share label computation periods, leaking information.

Walk-Forward Framework

Rolling Window (Fixed Train Size)

The train window has a fixed size and slides forward in time. This is preferred when you believe older data is less relevant (common in crypto).

Window 1: [===TRAIN===][=TEST=]
Window 2:    [===TRAIN===][=TEST=]
Window 3:       [===TRAIN===][=TEST=]

Parameters:

• train_size

  : Number of bars/days in the training window

• test_size

  : Number of bars/days in the test window

• step_size

  : How far to advance between folds (often equals

  test_size

  )

Expanding Window (Growing Train)

The train window starts at the beginning and expands forward. This uses all available historical data, which helps when data is scarce.

Window 1: [==TRAIN==][=TEST=]
Window 2: [====TRAIN====][=TEST=]
Window 3: [======TRAIN======][=TEST=]

Parameters:

• min_train_size

  : Minimum training samples before first fold

• test_size

  : Fixed test window size

• step_size

  : How far to advance between folds

Choosing Between Them

Factor	Rolling	Expanding
Data recency	Prioritizes recent data	Uses all history
Regime changes	Better adapts to new regimes	May dilute recent regime
Sample size	Fixed, may be small	Grows over time
Crypto preference	Preferred for < 6mo horizons	Better for regime-stable models

Purging and Embargo

Purging

Remove training samples whose labels overlap with the test set's time range. If a label is computed as the 24h forward return starting at time

t

, any training sample where

t + 24h

extends into the test period must be purged.

def purge_train_indices(
    train_idx: list[int],
    test_start: int,
    label_horizon: int,
    timestamps: list[int],
) -> list[int]:
    """Remove train samples whose label windows overlap test period."""
    test_start_time = timestamps[test_start]
    return [
        i for i in train_idx
        if timestamps[i] + label_horizon < test_start_time
    ]

Embargo

Add a buffer gap between the end of training and start of testing to account for serial correlation that purging alone does not eliminate.

[===TRAIN===][--EMBARGO--][=TEST=]

Typical embargo sizes:

• 1-minute bars: 60–240 bars (1–4 hours)
• 5-minute bars: 12–48 bars (1–4 hours)
• Hourly bars: 6–24 bars (6–24 hours)
• Daily bars: 2–5 bars (2–5 days)
• Crypto rule of thumb: Embargo >= 2x the label computation horizon

Combinatorial Purged Cross-Validation (CPCV)

CPCV (Lopez de Prado, 2018) generates all possible train/test combinations from

N

groups while maintaining temporal ordering. This produces far more test paths than standard walk-forward, enabling statistical tests for overfitting.

Key properties:

• Splits data into

  N

  contiguous groups
• For each combination of

  k

  test groups, the remaining

  N-k

  groups form the training set
• Applies purging and embargo at each train/test boundary
• Produces

  C(N, k)

  backtest paths (e.g., N=6, k=2 gives 15 paths)

See

references/methodology.md

for the full CPCV algorithm and formulas.

Overfit Detection

Deflated Sharpe Ratio (DSR)

The observed Sharpe ratio must be adjusted for:

• Number of strategies tested (multiple testing)
• Non-normality of returns (skewness, kurtosis)
• Length of the backtest

import numpy as np
from scipy.stats import norm

def deflated_sharpe_ratio(
    observed_sr: float,
    num_trials: int,
    backtest_length: int,
    skewness: float = 0.0,
    kurtosis: float = 3.0,
) -> float:
    """Compute the probability that observed SR > 0 after deflation.

    Args:
        observed_sr: Annualized Sharpe ratio of the selected strategy.
        num_trials: Number of strategies tested (including discarded ones).
        backtest_length: Number of return observations.
        skewness: Skewness of returns.
        kurtosis: Excess kurtosis of returns.

    Returns:
        p-value (probability SR is genuinely > 0).
    """
    sr_std = np.sqrt(
        (1 - skewness * observed_sr + (kurtosis - 1) / 4 * observed_sr**2)
        / (backtest_length - 1)
    )
    # Expected max SR under null (Euler-Mascheroni approximation)
    euler_mascheroni = 0.5772156649
    expected_max_sr = norm.ppf(1 - 1 / num_trials) * (
        1 - euler_mascheroni
    ) + euler_mascheroni * norm.ppf(1 - 1 / (num_trials * np.e))
    dsr = norm.cdf((observed_sr - expected_max_sr) / sr_std)
    return dsr

A DSR below 0.95 suggests the observed performance is likely due to overfitting across the trials tested.

Probability of Backtest Overfitting (PBO)

PBO uses CPCV to measure the fraction of backtest paths where the in-sample optimal strategy underperforms the median out-of-sample. A PBO above 0.50 indicates more-likely-than-not overfitting.

See

references/overfit_detection.md

for complete derivations and implementation details.

Crypto-Specific Considerations

1. Shorter windows: Crypto regimes change faster than equities. A 90-day rolling window may be more appropriate than 252 days.
2. 24/7 markets: No weekends or holidays to account for, but funding rate resets (every 8h on perps) create microstructure effects.
3. Survivorship bias: Many tokens delist. Validation must include delisted tokens or at minimum acknowledge this limitation.
4. Liquidity regime shifts: A token's liquidity profile can change dramatically (new CEX listing, liquidity mining end). Train/test splits should ideally not straddle major liquidity events.
5. Data availability: Many tokens have < 1 year of data. Expanding windows with small

   min_train_size

   may be necessary.

Practical Window Sizes for Crypto

Strategy Timeframe	Train Window	Test Window	Embargo
Scalping (1-5min)	3-7 days	1 day	2-4 hours
Intraday (15min-1h)	14-30 days	3-7 days	12-24 hours
Swing (4h-daily)	30-90 days	7-14 days	2-5 days
Position (daily-weekly)	90-180 days	30 days	5-10 days

Quick Start

from walk_forward import WalkForwardValidator, WalkForwardConfig

config = WalkForwardConfig(
    train_size=90,
    test_size=14,
    step_size=14,
    window_type="rolling",
    embargo_size=3,
    purge_horizon=1,
)

validator = WalkForwardValidator(config)
for fold in validator.split(price_data):
    model.fit(fold.train_X, fold.train_y)
    predictions = model.predict(fold.test_X)
    fold.record_performance(predictions, fold.test_y)

results = validator.aggregate_results()
print(f"OOS Sharpe: {results.oos_sharpe:.3f}")
print(f"Train/Test Sharpe ratio: {results.sharpe_ratio_ratio:.2f}")

Files

References

• references/methodology.md

  — Walk-forward theory, window types, purging, embargo, CPCV algorithm with formulas

• references/overfit_detection.md

  — Deflated Sharpe ratio, probability of backtest overfitting, multiple testing corrections

• references/practical_guide.md

  — Window size selection for crypto, regime considerations, common validation mistakes

Scripts

• scripts/walk_forward.py

  — Walk-forward validation engine with rolling and expanding windows;

  --demo

  mode with synthetic data

• scripts/overfit_detector.py

  — Deflated Sharpe ratio and PBO computation;

  --demo

  mode with synthetic backtest results