Z-Score Mean Reversion

Percentage change tells you how much price moved (e.g., -5%). Z-score tells you how unusual that move is relative to recent volatility (e.g., -2 SD). A -5% move could be z = -1 (normal) in a volatile stock or z = -3 (extreme) in a stable stock. Z-score normalizes for volatility.

Start with 20 periods for swing trading. This is balanced - responsive enough to catch moves but stable enough to avoid noise. If you hold positions longer, use 50 or 100. Shorter periods (10-15) are for very short-term trading.

Mean reversion is a tendency, not a guarantee. Strong trends can keep z-scores extreme for extended periods because the mean itself is moving (in a rolling calculation). This is why trend filters help - you trade with the trend, not against it.

Extremely negative z-scores indicate rare events. They can happen in crashes or major selloffs. While they represent higher probability reversion eventually, they can extend further before reverting. Use stops to protect against extended moves. Z = -3 is your stop, not a signal to add more.

Z-score uses pure statistics (standard deviations from mean). RSI uses a momentum formula. The key advantage of z-score is its statistical interpretation - a z-score of -2 has a known probability (about 2.3% of values below). RSI 30 doesn't have such precise statistical meaning.

In trending markets, the rolling mean follows price, making z-scores less extreme. Solution: Use a trend filter - only trade z-scores in the trend direction. Or use a longer lookback (50-100) so the mean moves more slowly. Accept that pure mean reversion underperforms in trends.

Close price is standard and works well. Typical price ((H+L+C)/3) can reduce noise by averaging the bar's range. Either works - just be consistent. Close price is simpler and most common.

Use longer z-score (like Z50 or Z100) as a filter and shorter (Z20) for entry. Best signals: Both negative (for longs) or both positive (for shorts). This ensures both tactical and strategic alignment.

High volatility means larger standard deviations, so z-scores stay closer to zero even with big price moves. Adjust by using wider thresholds (±2.5 or ±3.0) in high volatility. Alternatively, use a shorter lookback so SD reflects current volatility.

No. Z-score is mathematically (Price - Mean) / SD. If price is above mean, the numerator is positive, so z-score is positive. If price is below mean, z-score is negative. Z-score directly reflects position relative to mean.

Adapt lookback based on volatility: Shorter in high vol (faster adaptation), longer in low vol (more stability). Adapt thresholds similarly: Wider in high vol, tighter in low vol. Use VIX/VFTSE or ATR percentile to define regime. Test rigorously to avoid overfitting.

Check IV rank before strategy selection. High IV (common at extremes): Use credit spreads to benefit from IV crush. Low IV: Use debit spreads or outright options. Target strike at mean price. Expiry 45-60 DTE for statistical edge to manifest.

Rank signals by |z| magnitude. Apply trend filter to each. Limit sector concentration (max 2-3 per sector). Track average portfolio z-score. Balance long/short exposure when possible. Cap concurrent positions at 8-12 to manage correlation risk.

Mean reversion relies on statistical pricing inefficiencies. Large positions can move markets, especially in smaller stocks. The strategy works best when your trades don't impact price significantly. Scale is limited by liquidity and market impact.

Separate backtest by regime: trending (ADX > 25) vs ranging (ADX < 20), high vol vs low vol. Compare performance metrics across regimes. Expect z-score to excel in ranging/low-vol and underperform in trending/high-vol. Use regime filters to trade only favorable conditions.