RSI(2) Strategy by Larry Connors: Short-Term Mean Reversion

In 2008, Larry Connors published research showing that a 2-period RSI — far shorter than the standard 14-period setting — produced remarkably consistent short-term trading signals. The insight was elegant: by using just two periods, the RSI becomes hypersensitive to recent price action, identifying moments of brief but extreme selling pressure that almost always snap back. In equities, the system demonstrated a win rate exceeding 75%. The question for crypto traders: does this mean-reversion edge survive in a market known for its brutal, sustained moves?

The Original Connors Rules

The system is deceptively simple:

Setup: Price must be above the 200-day moving average (confirms long-term uptrend — you only buy dips in an uptrend).

Entry: RSI(2) drops below 5. Buy at the close.

Exit: RSI(2) rises above 65. Sell at the close.

No stop loss in the original system — the 200 MA filter provides structural protection. Connors argued that stops on mean-reversion trades get hit at exactly the wrong moment, just before the snapback.

Why RSI(2) Works

A 2-period RSI reflects only the last two candles. Two consecutive down days push RSI(2) toward zero regardless of the broader trend. This captures panic selling — moments when short-term traders dump positions into what is still a structurally sound uptrend. The 200 MA filter ensures you're only buying these dips during genuine uptrends, not catching falling knives in bear markets.

The mathematics favor the system: extreme RSI(2) readings below 5 occur after sharp 2-3 day selloffs that, by definition, create short-term oversold conditions. In uptrending markets, these conditions resolve within 1-5 days as buyers step in — hence the high win rate.

Adapting for Crypto

The critical adaptation is adding a stop loss. In equities, a stock in an uptrend rarely drops more than 5-8% from a 2-day selloff. In crypto, a 15-20% selloff in two days is common even during bull markets. Without a stop, a single trade can erase months of profits.

Backtest Results on BTC

The adapted version generates nearly twice the signals with only a modest win rate reduction. The stop-loss version sacrifices some returns but reduces maximum drawdown from -18% to -12% — a worthwhile tradeoff for most traders.

Advanced Variations

Cumulative RSI(2): Instead of a single reading below 10, require that the 3-day sum of RSI(2) is below 20. This filters for sustained selling pressure rather than a single bad day. Win rate improves to 68% with slightly fewer signals.

Connors RSI (CRSI): Connors later developed a composite indicator combining RSI(3), Up/Down streak length, and Rate of Change percentile. CRSI below 10 provides more nuanced signals but introduces complexity without significantly improving results on BTC.

Volatility filter: Only take signals when ATR(14) is within 1 standard deviation of its 100-day mean. This avoids entries during volatility spikes when even mean-reversion setups can fail spectacularly. Reduces trades by 20% but improves profit factor by 0.3.

Key lesson: RSI(2) is a short-term tactical tool. It's not a portfolio strategy. Allocate a fixed portion (15-25%) of your trading capital to RSI(2) trades and run it alongside a core trend-following system. The two approaches are naturally complementary — one profits from trends, the other from temporary dislocations within trends.

FAQ