Connors RSI

TL;DR. Three readings from 0 to 100 averaged into one: a 3-bar RSI of price, a 2-bar RSI of how many bars in a row price has risen or fallen, and where today's one-bar return ranks among the last 100. Above 90 is stretched up, below 10 stretched down.

#What it measures

• Short-term strength. The 3-bar RSI reacts within a bar or two.
• The streak. Three down closes in a row is a streak of -3. A lengthening up streak pushes its RSI toward 100 and a lengthening down streak toward 0, so long runs push the reading to the extreme.
• How unusual today's move is. The rank is the percent of the last 100 one-bar returns that were smaller than today's.

#Formula

#Inputs in AlgoLift

#Outputs in AlgoLift

#How to read it

• Below 10: a sharp, unusual drop, often after a run of down bars. The classic long setup for mean reversion.
• Above 90: the mirror.
• Between 30 and 70: nothing unusual. Most bars sit here.
• One component dominating: a long streak can hold the reading at an extreme even after price steadies. Check the streak before fading.

#Best in, worst in

Best in liquid instruments with a mean-reverting short-term habit, such as index futures and large stocks, traded on pullbacks inside a longer trend.

Worst in news-driven runs, where a stretched reading keeps stretching.

#Three setups

#1. Pullback in an uptrend

Enter long when CRSI is below 10 (a Comparison) while price is above an SMA of 200 bars. Exit when CRSI rises above 70, through Set Exit Condition.

#2. Short the spike in a downtrend

The mirror: CRSI above 90 below the 200-bar average.

#3. Tighter entries

Require CRSI below 5 instead of 10 for fewer, more stretched entries, and compare the trade count and the average trade before keeping it.

#Common mistakes

• No trend filter. Fading every extreme in both directions trades against strong moves. Pick the side with a longer average.
• A stop that is too tight. Mean-reversion entries buy weakness; a stop inside the normal noise exits before the snap back.
• Over-tuning the thresholds. 10 and 90 are the published levels; a rule that only works at 7 is fit to history (overfitting in trading).