Walk-Forward Analysis

Walk-forward analysis (WFA) was developed to address the most fundamental flaw in traditional backtesting: that a strategy optimized on the entire historical dataset is almost certainly overfit to that data. WFA simulates the realistic scenario where a trader periodically re-optimizes their strategy as new data arrives.

**How walk-forward analysis works:** 1. **Define windows:** Choose an in-sample (IS) window (e.g., 12 months) and out-of-sample (OOS) window (e.g., 3 months) 2. **First pass:** Optimize strategy parameters on months 1–12; test on months 13–15 (OOS) 3. **Roll forward:** Optimize on months 4–15; test on months 16–18 (OOS) 4. **Continue:** Each cycle optimizes on IS data and tests on the next OOS period 5. **Concatenate OOS results:** Stitch together all OOS performance to create the walk-forward report 6. **WFA ratio:** OOS performance / IS performance. Above 0.5 (50%+) is generally considered acceptable.

**Why WFA is more reliable than simple backtesting:** In simple backtesting, the strategy "knows" the future implicitly because parameters were tuned to match it. In WFA, each OOS window is genuinely unseen during optimization — it accurately simulates live trading.

**Anchored vs. rolling walk-forward:** - **Rolling (moving window):** IS window has fixed length, moves forward — equally weights all data, adapts to recent conditions - **Anchored (expanding window):** IS window grows from a fixed start point — gives more weight to longer-term patterns, slower to adapt

**Crypto-specific challenges:** Crypto market regimes change more rapidly than traditional markets. A 12-month IS window optimized entirely in a bull market may produce parameters poorly suited to the following bear market. Using shorter IS windows (6 months) with 1–2 month OOS windows may be more appropriate for crypto.