Mean reversionRSI-2 Mean Reversion
Larry Connors' short-term pullback system: buy deep 2-period RSI dips inside an established uptrend and exit fast on the snap-back.
Swing tradingIntermediateDaily (the classic); works on 1h for faster variants
The idea
RSI-2 is a counter-intuitive, short-term mean-reversion system popularised by Larry Connors, and it inverts the usual momentum logic: in a healthy uptrend, sharp one-to-three-day pullbacks tend to snap back, so you buy fear rather than strength. The 2-period RSI is an extremely fast oscillator that spikes to extremes on tiny moves, flagging when a name has pulled back hard enough to be stretched. Crucially the setup only fires above the 200-day average, so you are always buying a dip inside an established uptrend, never trying to catch a falling knife. The holding period is short — often one to four days — because the edge is the reversion, not the trend. It is a high-win-rate, small-average-win approach, the mirror image of trend following.
The setup
Plot a 200-day simple moving average as the trend gate and a 5-day average as the exit trigger, then add a 2-period RSI in a pane below. The market must be above the 200-day average for any long to be considered. You are now waiting, doing nothing on the vast majority of days, until a fast pullback drives the 2-period RSI to a genuine extreme. Because RSI-2 whipsaws violently by design, the surrounding rules — the trend filter and the fast exit — are what convert its noise into an edge.
Entry
When price is above the 200-day average and the 2-period RSI closes below 10, buy on that close. Connors' research found that lower thresholds improve the edge, so many traders use RSI below 5, or scale in: a first unit under 10 and a second under 5, averaging into deeper weakness. There is no waiting for confirmation — the entire premise is buying while it still feels wrong, into the down-close. This is emotionally hard, which is precisely why the mechanical rule exists.
Exit and targets
Exits are deliberately quick because reversion is fast and fickle. The standard rule is to sell when price closes back above its 5-day moving average, which typically happens within a few days of the entry. An alternative is to exit when the 2-period RSI rebounds above 70, capturing the snap-back directly. Either way, do not overstay — turning a quick reversion trade into a would-be trend hold is the fastest way to give the profit back. The average winner is small; the system's edge is winning often, so banking the reversion mechanically is essential.
Risk management
RSI-2 uses a counter-intuitive stop philosophy: because you are buying into weakness that often gets briefly worse before it reverts, a tight technical stop gets shaken out constantly and destroys the edge. Instead use a wide catastrophe stop — say 8 to 10 percent — purely to cap disaster, and control risk mainly through small position size and diversification across several uncorrelated setups. The real risk control is the 200-day trend filter, which keeps you out of the falling markets where mean reversion fails hardest. Accept occasional larger losers as the cost of a high win rate, and never let one position dominate the book.
Best timeframes and markets
The system was built and tested on the daily chart of liquid, mean-reverting instruments — broad index ETFs like SPY and QQQ are the canonical vehicles because indices revert more reliably than single stocks. It also adapts to intraday charts (for example a 2-period RSI on the hourly) for faster variants, though noise rises. It works poorly on strongly trending momentum names that simply keep running, and on illiquid stocks prone to gaps. Reversion strategies as a class also suffer in high-volatility bear phases, which is exactly what the trend filter is there to avoid.
Common mistakes and variations
The most common failure is dropping the 200-day trend filter and buying oversold readings in a downtrend, which turns the system into a knife-catcher. Variations include using RSI-3 or the CumulativeRSI for slightly smoother signals, adding a maximum-holding-period time stop, or combining several ETFs into a rotational basket to smooth the equity curve. Some traders replace the 5-day-average exit with a fixed one-to-four-day time exit. All variations keep the two pillars intact: trade only with the long-term trend, and exit the reversion fast.
A worked example
SPY is trading at 430, comfortably above its 200-day average at 405, so longs are enabled. A three-day pullback drops it to 418 and the 2-period RSI plunges to 8 — deep oversold inside an uptrend. You buy the close at 418 with a wide 8% catastrophe stop at ~385 and a small position size so that stop is only a 1% account risk. Two days later SPY rebounds and closes at 425, back above its 5-day average, triggering the exit for a quick +1.7% gain. The win is modest, but this same pattern recurs dozens of times a year, and the trend filter keeps the rare losers from compounding.