Composite & famousConnors RSI · CRSI
Larry Connors' three-part composite that supercharges RSI for short-term mean reversion.
Works best in ranging marketsEngine-computed on a fixed sample series
What it is
Connors RSI (CRSI), developed by Larry Connors and his research group, is a composite momentum oscillator built specifically for short-term mean-reversion trading. It blends three separate measurements into one 0-to-100 reading to capture different facets of how overextended a move has become: the raw momentum of price, the persistence of the current up- or down-streak, and how extreme today's move is versus recent history. By combining them, it aims to be a more robust overbought and oversold gauge than a plain RSI for very short holding periods. The question it answers is whether a market has become stretched enough, across several dimensions at once, that a snap-back is likely. It is designed for quick, counter-trend entries in liquid markets, typically held only a few days.
How it is calculated
Connors RSI is the equal-weighted average of three components. The first is a short-period RSI of price, with a default length of 3. The second is an RSI of the streak, meaning the number of consecutive days price has closed up (counted as a positive run) or down (a negative run), with a default length of 2, which measures how persistent the current directional streak is. The third is the Percent Rank of the single-day Rate of Change over a lookback of 100, that is, what percentile today's one-day return falls into versus the last hundred days. The final value is CRSI = (RSI(3) of price + RSI(2) of the streak + PercentRank(100) of the 1-day ROC) / 3, producing a 0-to-100 oscillator. The default parameter set is written as (3, 2, 100).
Reading it, step by step
Being a 0-to-100 oscillator built for mean reversion, CRSI is read at its extremes: readings below about 10 to 15 flag a deeply oversold condition ripe for a bounce, and readings above about 85 to 90 flag an overbought condition ripe for a pullback. Because it fuses three stretched-ness measures, an extreme reading means price momentum, streak persistence, and the size of the latest move all agree the market is overextended, which Connors' research found gives sharper mean-reversion edges than RSI alone. It is deliberately fast and jumpy, so it reaches extremes often and is meant to be traded frequently over short horizons. Middle readings around 50 carry little information. The signal is the extreme itself, usually acted on the same or next day, not a slow crossover.
Best timeframes and settings
The canonical settings are (3, 2, 100), tuned by Connors for daily-chart, short-term equity mean reversion, and most users keep them. It is overwhelmingly a daily-timeframe tool applied to liquid stocks, ETFs, and indices, where short-term pullbacks within larger uptrends tend to revert. Lengthening the price RSI or the streak RSI slows and steadies the oscillator but blunts the short-term edge it was built for, while the 100-day percent-rank lookback sets how much history the latest move is judged against. It generally is not used on long timeframes, since its whole premise is short-horizon reversion. The trade-off is that its speed is the point: it is meant to be responsive and frequently at extremes, so smoothing it defeats the purpose.
When and where to use it
Use Connors RSI for systematic, short-term mean-reversion trades, especially buying oversold dips inside a confirmed longer-term uptrend and exiting quickly on reversion. It is most effective on liquid stocks, ETFs, and index products on daily charts, where Connors' original research was conducted, and it works best when a higher-timeframe trend filter keeps you buying dips in uptrends rather than catching falling knives. It is not built for trend-following or long holds, and it performs poorly as a standalone tool in strongly trending markets where it stays overbought or oversold. Reach for it when your edge is the tendency of short-term extremes to revert, and pair it with a trend filter to avoid fighting a genuine breakdown.
Strategies that use it
Oversold pullback buy: in a market above its long-term moving average, such as the 200-day, buy when CRSI drops below 10 to 15, and exit when it rises back above 50 to 70 or after a fixed few-day hold, a classic Connors mean-reversion template. Overbought fade: in a downtrend or for exits, sell or trim when CRSI pushes above 85 to 90. Scaled entries: some traders add to a position in tranches as CRSI makes progressively lower extreme readings, betting on reversion. Time-based exit: because the edge is short-lived, many implementations exit after a set number of days regardless of CRSI, treating the extreme purely as the entry trigger. In all cases a trend filter governs which direction of extreme to trade.
Combining it with other indicators
CRSI is almost always paired with a long-term trend filter, such as a 200-day moving average, so you only buy oversold readings in uptrends and only fade overbought readings in downtrends, which avoids its worst failure mode. Support and resistance refine entry timing at the extreme. Volume can confirm that an oversold reading is a genuine washout rather than the start of a breakdown. Because it is itself a composite, adding a second, slower momentum tool guards against acting on a fleeting spike. The consistent principle is that CRSI supplies a high-quality short-term extreme, and a trend filter supplies the context that decides whether that extreme is a buying opportunity or a warning to stay away.
Where it fails
CRSI's defining weakness is trending markets: in a strong downtrend it becomes oversold and stays oversold, so buying its low readings without a trend filter means repeatedly catching falling knives. Its speed makes it noisy, and it reaches extremes so often that undisciplined traders overtrade it. Its edge is short-lived, so holding a CRSI trade too long usually gives back the reversion. Being a composite, it is more complex and less transparent than a plain RSI, and its three components can occasionally send mixed signals masked by the average. The remedies are a strict trend filter, fixed short holding periods, and treating it as one component of a rules-based system rather than a discretionary all-purpose oscillator.
A worked example
Suppose a stock trading well above its 200-day moving average sells off hard for three straight days. On the third day, the 3-period RSI of price reads 8 (deeply oversold), the 2-period RSI of the down-streak reads 4 (a persistent, extreme down run), and today's one-day decline ranks in the 6th percentile of the last 100 daily returns, giving a Percent Rank of 6. Connors RSI = (8 + 4 + 6) / 3 = 6, an extremely oversold composite reading. Because the stock is still above its long-term average, a mean-reversion trader treats this sub-10 reading as a buy signal, entering near the close and planning to exit when CRSI rebounds above 50 or after a few days. The three components all agreeing on extreme oversold conditions is exactly the confluence CRSI is designed to capture.