Connors RSI in Practice: Three Calculations and Reversal Misjudgment

Trading Strategies

22 October 2026

Connors RSI's reversal signals often "come too early." This is not a parameter tuning problem—it is a built-in characteristic of the indicator itself. TradingView's official documentation openly admits that in an uptrend, Connors RSI may issue a sell signal while the market continues to rise, creating a false signal. The most effective way to reduce misjudgment is not to change the threshold from 90/10 to 95/5, but to use a trend filter to block all counter-trend signals. StockCharts' ChartSchool reaches the same conclusion: Connors RSI is designed for short-term overbought and oversold conditions, and false signals when used alone are expected behavior, not a malfunction.

OKX Exchange
A leading global cryptocurrency platform,suitable for both beginners and experienced traders.
New user benefit: 20% off trading fees upon registration!！

Three Calculations: Each One Measures a Different Kind of "Extreme"

Connors RSI takes the average of three values ranging from 0 to 100. The default parameters are (3, 2, 100). Understanding what each component measures is more important than memorizing the formula.

First component: RSI(3). This is the standard Wilder RSI, but with the period shortened from the common 14 to 3. It captures price momentum over the last three candles. The shorter the period, the more sensitive it is to price changes. This is one reason Connors RSI reacts faster than traditional RSI. In highly volatile markets, RSI(3) may hit extreme values on its own, but the three-part average dilutes it.

Second component: Streak RSI(2). First, count the number of consecutive up or down days: consecutive gains are recorded as +N, consecutive losses as −N, and flat days reset to zero. Then a 2-period RSI is calculated on this "streak series." This component measures how long the trend has lasted, not how far it has moved. For example, if the market falls for five consecutive days but each day only drops slightly, RSI(3) may not be very low, but the streak value will be large in absolute terms, and Streak RSI will be close to 0. It captures the fact that the market has "been falling for too long."

Third component: Percent Rank(100). Calculate today's single-day percentage change and determine where it ranks among the last 100 daily percentage changes. If today's gain is larger than 95 of the past 100 days' gains, this component reads 95. If today's loss is larger than 90 of the past 100 days' changes (meaning it is only better than 10 days), this component reads 10. It measures how extreme the single-day move is, from a different angle than RSI or Streak.

After averaging the three components, Connors RSI ranges from 0 to 100. Connors' original overbought and oversold lines are 90 and 10, not the traditional RSI levels of 70/30. The reason is that the three-part average smooths out individual extreme values. If the thresholds were not pushed further out, signals would be so frequent that they would lose their screening value.

The Root of Reversal Misjudgment: The Indicator Is "Catching a Falling Knife"

Connors RSI is designed for mean reversion, but mean reversion assumes that "price will return to the mean." If the mean itself is moving quickly, the reversion direction is no longer a reversal—it becomes a continuation of the trend.

A typical misjudgment scenario: an asset is in a clear downtrend, and Connors RSI drops below 10, triggering an "oversold buy" signal. But in a downtrend, oversold conditions can persist for a long time, and prices continue to fall. Buying immediately means getting trapped. TradingView's documentation states it plainly: when the market is in a downtrend, Connors RSI may issue a short-term buy signal, but the trend has not changed. StockCharts also reminds traders that when the indicator hits extreme values, they should check the trend, volume, and other indicators instead of placing an order directly.

Quantitative backtesting data supports this from another angle. Backtrader's documentation records a test of a Connors RSI composite strategy: on 17 years of gold daily data, using a rule of "enter with a limit order when the closing price is 0.3% below yesterday's close," only 38 trades were executed, with a 78.95% win rate and a maximum drawdown of 6.42%. The combination of few trades, a high win rate, and shallow drawdowns itself shows that strict entry conditions—not frequently triggered threshold signals—are what make this indicator usable. If you traded mechanically on the 90/10 thresholds alone, the number of trades would be far higher, and the win rate and drawdown would likely not look nearly as good.

Three Practical Ways to Reduce Misjudgment

First, adding a trend filter is more effective than adjusting thresholds. If you use Connors RSI for long trades, only accept oversold signals when price is above a long-term moving average (such as the 200-day SMA or 100-day SMA). INDODAX's tutorial gives similar advice: set the buy condition as "price above the 200-day moving average AND CRSI < 10." The goal is to wait for pullbacks in an uptrend, not to bottom-fish in a downtrend. Backtrader's classic RSI2 strategy also uses an SMA as a "safety belt": buy oversold conditions only when the trend is up, and stay in cash when the trend is down. The cost of this filter is that you will miss some bottom reversal opportunities—when price forms a bottom below the moving average, you will not be triggered. But the number of false signals it filters out is usually far greater than the number of true signals it misses.

Second, do not use Connors RSI to catch "tops." Overbought signals (>90) also appear frequently in uptrends, and the market can remain overbought for a long time. TradingView explicitly points out that sell signals in an uptrend may be false signals. If you are shorting, using Connors RSI's overbought signals alone carries high risk. A safer approach is to use it as a hint that "the rally may be slowing down," combined with price patterns or weakening trend strength for confirmation—not to enter as soon as you see 90.

Third, accept that "early signals" are a feature, not a flaw. All three components of Connors RSI are short-period calculations. It is inherently a leading indicator. Being leading means it will issue signals earlier than the actual trend reversal, and some of those signals will be invalidated by subsequent price action. If you want signals to be more "accurate," you need to accept later entries, which means switching to longer-period indicators or waiting for price confirmation. Using Connors RSI while expecting it not to be early is like asking an indicator designed for sensitivity to become sluggish.

My Judgment for You

If you are trading manually, first determine your trend direction using Connors RSI (moving averages or higher timeframe structure), and only accept signals in the direction of the trend. Ignore all counter-trend signals. Keep the thresholds at 90/10 or more extreme. Do not lower them to 80/20 just to "catch more opportunities."

If you are writing a strategy backtest, add trend filter conditions to your entry rules, then observe whether the number of trades drops significantly. If the number of trades is still high after adding the filter, your thresholds are too loose, or the volatility structure of the instrument you are testing is not suitable for this indicator. The 38-trades-in-17-years case from Backtrader's documentation can serve as a reference for "reasonable use."

If you are already experiencing frequent false signals, check two things first: Are you trading directly on the 90/10 thresholds without any trend filter? Is the platform you are using calculating Percent Rank in a way that matches the standard? NexusFi's documentation mentions that different platforms may handle flat days in the streak calculation and inclusion/exclusion rules in Percent Rank differently. A 2–3 point discrepancy can cause signals to appear one candle earlier or later. Manually verifying with 5–10 known data points is more useful than adjusting parameters.

OKX Exchange
A leading global cryptocurrency platform,suitable for both beginners and experienced traders.
New user benefit: 20% off trading fees upon registration!！