Stock Market · Trading Systems

Mean reversion systems

Mean reversion is the opposite philosophy of trend following: prices that move far from their average tend to revert back. Buy oversold conditions, sell overbought conditions, profit from the snap-back.

Classic mean-reversion rules

Entry: Buy when RSI(2) < 10 AND price > 200-day SMA (oversold pullback in uptrend)

Exit: Sell when RSI(2) > 70 OR after 5 days

Stop: 3% below entry

Performance characteristics

• Win rate: HIGH (60-80%)
• Average win: SMALL
• Lots of small wins, occasional medium-sized losses
• Works best in range-bound, low-volatility markets
• Devastating in trending markets (you keep buying as the trend continues lower)

Trend following vs mean reversion

Trendlow win rate, big wins

Mean reversionhigh win rate, small wins

[compare:Trend follow=35-45% win, 3-5x R:R|Mean revert=60-80% win, small R:R]

Both have the same expected return mathematically, but VERY different psychological profiles. Most retail traders prefer mean reversion because winning often feels better.

> The danger: mean reversion looks like a beautiful, consistent strategy until the day it doesn't revert. One trend can wipe out months of small wins.

Indian markets context

Bank Nifty has historically been a great mean-reversion vehicle. Indices revert more reliably than individual stocks. Mean reversion strategies on individual small/mid-caps often fail because trends in these names run longer.

Tools for mean reversion

• RSI (2-day or 14-day)
• Bollinger Bands (price touching lower band)
• Z-score of returns
• Distance from moving average

When mean reversion fails catastrophically

• 2008 financial crisis: indices kept falling, no reversion
• March 2020 COVID crash: 40% decline in weeks, no reversion
• Strong trending bull markets (no oversold conditions to buy)

This is what a regime filter is for. An oversold reading means something different in a range-bound market than in a confirmed downtrend, in the first it's a stretched rubber band, in the second it's just the trend continuing. The filter is what tells the two apart before the signal is acted on.

Takeaway. Mean reversion wins often (60-80%) and wins small, which makes it work in range-bound markets and fail catastrophically in trending ones. The same oversold reading means opposite things in those two regimes, which is why the regime filter does more work here than the signal itself.