Does ConnorsRSI Work? 10 and 90 Tested on S&P 500 Stocks
ConnorsRSI crossing 10 and 90 on 483 S&P 500 stocks, 2006 to 2026: 4 of 8 signal tests differed clearly from ordinary days over the next 5 days.
The Test Bench › Mean reversion › ConnorsRSI
The three parts of ConnorsRSI, each worked through with an example, what 10 and 90 mean, and how it differs from RSI(2), with what our tests found.
ConnorsRSI is a short-term oscillator built for mean reversion: buying a stock after a sharp, recent fall. It starts from a short RSI and adds two measures of how stretched the move is, so a reading below 10 requires all three parts to be weak at the same time.
This page covers:
ConnorsRSI was published in 2012 by Connors Research, Larry Connors’ firm; its guidebook was written by Larry Connors, Cesar Alvarez and Matt Radtke. It averages three measures, each on a 0 to 100 scale:
The standard setting is ConnorsRSI(3, 2, 100), and GU Analyser uses it. The result runs from 0 to 100 and is read with the same extremes as RSI(2), 10 and 90.
ConnorsRSI = (RSI(3) of the close + RSI(2) of the streak + percent rank) ÷ 3. Each part, step by step:
Eight invented closes, with the streak count after each:
| Close | $50.00 | $50.60 | $51.10 | $50.40 | $49.80 | $49.50 | $49.50 | $50.20 |
|---|---|---|---|---|---|---|---|---|
| Streak | 0 | +1 | +2 | −1 | −2 | −3 | 0 | +1 |
The first close has no earlier close to compare with, so it starts at 0. Two higher closes make +1 and +2; three lower closes make −1, −2 and −3; an unchanged close resets the count to 0; the next higher close starts again at +1. ConnorsRSI then takes a 2-day RSI of these counts, so a long run of down closes pulls that part towards 0.
Suppose today the stock fell 3.1%. Among the previous 20 one-day returns in this illustration, 2 were lower, so the percent rank is 2 ÷ 20 × 100 = 10.0. ConnorsRSI uses the previous 100 days, and an equal return does not count as lower.
Suppose the same day gives an RSI(3) of 11.8 and a streak RSI of 6.4. With the percent rank of 10.0, ConnorsRSI is (11.8 + 6.4 + 10.0) ÷ 3 = 9.4: below 10, an oversold reading. Every part is low here; any one part at 30 or above would have kept the average at 10 or above.
Below 10, the three parts add up to less than 30, so each of them is below 30: the 3-day RSI is weak, the streak component is weak after a run of lower closes or a recent change in direction, and today’s return ranks in the lowest 30% of the last 100 days. Above 90 is the mirror image on the way up. The ConnorsRSI guidebook tests buy levels from 5 to 15; a lower level is stricter and fires less often.
Both aim at short pullbacks and both use 10 and 90. RSI(2) measures only the size of the last few up and down moves; ConnorsRSI adds how many days the move has run and how unusual today’s move is against the last 100 days. A reading below 10 therefore also needs today’s return to rank low, and in our test ConnorsRSI crossed below 10 far less often than RSI(2).
| RSI(2) | ConnorsRSI | |
|---|---|---|
| Introduced by | Larry Connors and Cesar Alvarez, 2008 | Connors Research, 2012 |
| Reads | Gains and losses, smoothed over 2 days | A 3-day RSI, the streak and the 100-day rank |
| Usual zones | 10 and 90 | 10 and 90 (the guidebook tests 5 to 15) |
| Below 10 on the chart’s invented days | 11 | 2 |
| Trades in our backtest (typical stock, 20 years) | 132 | 35 |
In our event test on 483 S&P 500 stocks above their 200-day average, the 5 days after ConnorsRSI fell below 10 returned 0.29 percentage points more than ordinary days; after RSI(2) fell below 10, 0.19 percentage points more than ordinary days. Their intervals overlapped, so the test did not show that one was better. In a separate backtest using each as the entry of one dip-buying rule, RSI(2) finished ahead of ConnorsRSI on 365 of 483 stocks, with a median total return per stock over 2006 to 2026 of +45% for RSI(2) against +13% for ConnorsRSI.
The guidebook that introduced ConnorsRSI uses it to buy pullbacks on the long side only. The entry is a ConnorsRSI reading below a level between 5 and 15, and the guidebook adds filters of its own, among them ADX(10) above 30 and an entry on a further pullback below the signal day’s close. Our tests use the simpler form shared across this series: every crossing of 10 or 90, split by the 200-day moving average.
A ConnorsRSI rule states each of these parts:
Mean reversion trading explains the entries, exits and risks of these rules, with ConnorsRSI beside RSI(14), RSI(2) and Stochastic RSI.
How to read these results. An event study compares the days after a signal with ordinary days of the same stocks; a backtest runs one complete rule with costs and reports a median total return per stock over the whole period. “Clear” means the 95% interval excluded zero under every test setting we ran. All tests use today’s S&P 500 members projected back, which favours rules that buy dips, and price returns without dividends.
The event studies compared the days after a signal with ordinary days for the same S&P 500 stocks, 2006 to 2026; the backtest ran the signals as one trading rule.
Each study states the signal, the stocks, the period and what it compares the signal with. The results show what happened in the tested sample, and each study states what the test cannot establish.
ConnorsRSI crossing 10 and 90 on 483 S&P 500 stocks, 2006 to 2026: 4 of 8 signal tests differed clearly from ordinary days over the next 5 days.
Four RSI entries under one dip-buying rule (above the 200-day average, out on a close above the 5-day average) on 483 S&P 500 stocks, 2006 to 2026, against buying and holding.
An indicator Connors Research published in 2012. It averages three parts: a 3-period RSI of the close, a 2-period RSI of the up or down streak, and the rank of today’s change among the last 100 days’ changes. Readings run from 0 to 100, and 10 and 90 are the usual extremes.
ConnorsRSI = (RSI(3) of the close + RSI(2) of the streak + the 100-day percent rank of today’s change) ÷ 3. Both RSIs use Wilder’s smoothing, as the standard RSI does.
The number of consecutive closes in one direction: +1, +2, +3 for higher closes, −1, −2, −3 for lower closes, and 0 after an unchanged close. ConnorsRSI takes a 2-day RSI of that count, so a long run of down closes pushes it towards 0.
The share of the previous 100 one-day returns that were lower than today’s, from 0 to 100. A fall larger than any in the last 100 days gives 0; an ordinary day gives a value near the middle.
They are the usual extremes. Below 10, all three parts are below 30: the 3-day RSI is weak, the streak component is weak after a run of lower closes or a recent change in direction, and today’s return ranks in the lowest 30% of the last 100 days. Above 90 is the mirror image. The ConnorsRSI guidebook tests buy levels from 5 to 15.
Larry Connors’ firm, Connors Research, published it in 2012, with a guidebook by Larry Connors, Cesar Alvarez and Matt Radtke. Connors and Alvarez had earlier made the 2-day RSI widely known.
Yes. Each part counts bars, so on hourly bars the percent rank looks back 100 hourly bars. The 10 and 90 levels come from tests on daily bars, and our tests use daily bars too.
The standard setting is ConnorsRSI(3, 2, 100): a 3-day RSI of the close, a 2-day RSI of the streak and a 100-day percent rank. GU Analyser uses that setting.
In our event test on 483 S&P 500 stocks above their 200-day average, the 5 days after ConnorsRSI fell below 10 returned 0.29 percentage points more than ordinary days; after RSI(2) fell below 10, 0.19 percentage points more than ordinary days. Their intervals overlapped, so the test did not show that one was better. In a separate backtest using each as the entry of one dip-buying rule, RSI(2) finished ahead of ConnorsRSI on 365 of 483 stocks, with a median total return per stock over 2006 to 2026 of +45% for RSI(2) against +13% for ConnorsRSI.
In our test on 483 S&P 500 stocks, 4 of 8 ConnorsRSI comparisons showed a clear 5-day difference from ordinary days. The study page lists each one. In a separate backtest using the same signal as the entry of one dip-buying rule, ConnorsRSI ranked third of the four entries by median total return per stock over 2006 to 2026 (+13%), and trailed buying and holding on most stocks.
Yes. In GU Analyser ConnorsRSI is in the Momentum group: build the entry and exit rules on a visual canvas and run a backtest or an event study on daily bars, with no code.
Want to test your own ConnorsRSI rule?
Add ConnorsRSI, set the level and the trend filter, and measure what followed the signal across your stocks, or backtest it with an entry, an exit and a cost on each trade.
Create a free account and build your own strategies → or read how ConnorsRSI is calculated in the app →The methods: how a backtest is built · how an event study works