Connors RSI: three components built for short-term mean reversion

What it measures

Connors RSI averages three things: a very short RSI of price, an RSI of the current up or down streak length, and the percentile rank of the latest one-bar return against recent returns.

Each component measures a different kind of extremity — price level, persistence, and magnitude — and the average only reaches its own extremes when all three agree.

It was built explicitly for short-term mean reversion and tested on holds of a few days. Using it as a trend tool is using it against its design.

How it is calculated

These are the steps Basis performs, verified against the published definition.

1. 1

   Compute a short-period RSI of the close, typically three periods.

2. 2

   Count the current streak of consecutive up or down closes and compute an RSI of that streak value.

3. 3

   Compute the percentile rank of the latest one-bar percentage change within a lookback of recent changes.

4. 4

   Average the three components equally.

Settings

Three for the RSI, two for the streak RSI and one hundred for the percent-rank window is the published set. The short periods are deliberate; lengthening them removes the short-term extremity the indicator exists to find, and turns a mean-reversion tool into a slow oscillator that is worse than RSI at the job RSI already does.

How to read it

• Readings below ten and above ninety are the conventional extremes, and they are genuinely rare because three conditions must align.
• It is a fade signal by construction: extreme readings suggest reversion, not continuation.
• In a strong trend it will produce repeated extreme readings that keep being wrong, which is the known failure mode of every mean-reversion tool.
• Because three components are averaged, a reading near fifty carries almost no information — it can mean everything is neutral or that the components disagree completely.

Where it misleads

Related indicators