Knowledge · Strategies · RSI-2 Mean Reversion (Connors)

RSI-2 Mean Reversion (Connors)

Larry Connors — 'Street Smarts' (1996), 'Short Term Trading Strategies That Work' (2008)
Mean Reversion Evidence: Strong swing equitiesequity_indicesetfs
5/10
Relevance for Botty
In an uptrend (price > 200-MA): go long when RSI-2 < 10, exit when the close rises above the 5-day MA. Very high win rate, short trades.
Core concept
After short pullbacks within an existing uptrend, there is a high probability of mean-reversion back up. RSI with a period of 2 is extremely sensitive and pinpoints exactly these brief oversold states. A trend filter via the 200-MA keeps the strategy out of bear markets.
Relevance for Botty
Relevance Score 5/10
BTC is not a typical mean-reversion market — deep oversold phases often keep running instead of reversing. Connors' logic would need to be adapted: a higher trend threshold (e.g. price > EMA-200 AND RSI-14 > 50) and a shorter timeframe (4h or 1h RSI-2). Botty's existing strategies are more trend-following, but an RSI-2 pullback filter could be useful as an add-on for long entries.
Rules

Entry

  • Trend filter: price > 200-day SMA
  • Long entry: 2-period RSI < 10 (or < 5 for even rarer signals)
  • Enter at the close or at the next open

Exit

  • Close > 5-day SMA (classic Connors exit)
  • Alternative: RSI-2 > 70
  • No stop-loss in the original version — the trend filter is the safety net
Parameter
NameTyp. valueDescription
rsi_period 2 Extremely short RSI
oversold_level 10 (oder 5) Buy zone
trend_ma 200 Filter for the higher-level trend
exit_ma 5 Exit trigger
Pros & Cons

Pros

  • Very high win rate (psychologically comfortable)
  • Short holding periods — little tail risk
  • Clear rules, fully mechanical
  • Tested across 25+ years and various regimes

Cons

  • Misses large trend moves — exits too early
  • Weaker on crypto/forex (not enough mean-reversion character)
  • Dangerous without a stop-loss when the trend filter fails (e.g. flash crash)
  • Long-only under the original rules — no short setups
Typical performance
cagr
~8-12% over 25 years on SPY
notes
Works better at the index level than on single stocks. Weaker on crypto (stronger trends overshoot more).
avg hold
3-7 days
win rate
70-80% on US equities/ETFs
max drawdown
~15-20%
profit factor
~2.0-2.5
Bot suitability
Perfect for daily-frequency bots on indices/ETFs. For crypto the rules have to be adapted.
Background

Core idea

Larry Connors took the well-known RSI indicator and turned two dials to the extreme:

  1. Period set to 2 instead of the classic 14 → extremely sensitive to short-term price moves
  2. Threshold set to 10 instead of 30 → only truly extreme oversold states trigger

The result is an indicator that, in strong uptrends, pinpoints exactly the brief 1-3 day pullbacks after which mean-reversion sets in with very high probability.

The rules

SPY > SMA(200)       ← trend filter (only trade in bull markets)
RSI(2) < 10          ← oversold entry
→ Long
Close > SMA(5)       ← exit as soon as price stabilizes

Why it works

The trend filter is the key. Without it the strategy tries to buy dips in bear markets → catastrophic drawdowns. With it, it only buys in structurally bullish markets, where even short pullbacks are typically absorbed.

Psychologically: when SPY closes red for 3 days within an uptrend, market participants start to panic-sell. That is exactly the 'blood in the streets' moment where long-term buyers step in cheaply. Connors' RSI-2 captures this moment mechanically.

Evidence

Backtests 1993-2025 on SPY: - Win rate ~75-80% - Average holding period 3-5 days - CAGR ~8-12% (less than buy-and-hold, but with far less time in the market) - Max drawdown ~15-20% vs. ~55% buy-and-hold during the GFC

Limits

  • Does not work on assets without structural mean-reversion (commodities, FX).
  • Problematic on crypto: BTC can close red 7 days in a row and then simply keep falling — its mean-reversion character is markedly weaker.
  • No stop-loss: there is none in the original form — if the trend filter fails (flash crash, gap-down), the loss is potentially large.

Relevance for Botty

BTC on the daily is not a Connors market. But: on higher timeframes during bull phases (price well above the EMA-200, low vol), an RSI-2 entry could serve as a supplementary pullback filter for Botty long strategies. Not recommended as a stand-alone strategy.