Connors Relative Strength Index (CRSI)

Summary

Connors RSI: a 0 to 100 oscillator, the plain average of three short-term momentum readings on that same scale.

The first is Wilder's RSI of the closes over optInTimePeriod bars. The second is an RSI over optInStreakPeriod bars of the up/down streak, the signed length of the current run of higher or lower closes (+3 after three higher closes in a row, -2 after two lower ones); an unchanged close resets the streak to 0. The third is the percent rank of the bar's one-bar return, (close - previous close) / previous close, among the optInRankPeriod returns before it: the percentage of them strictly below it.

High readings mark a short-term overbought market and low readings an oversold one.

Formula

\begin{aligned} \mathrm{CRSI}_t &= \frac{\mathrm{RSI}(\mathrm{Close}, m_1)_t + \mathrm{RSI}(\mathrm{Streak}, m_2)_t + \mathrm{PercentRank}(\mathrm{ROC1}, P)_t}{3} \\[6pt] \mathrm{Streak}_t &= \begin{cases} \max(\mathrm{Streak}_{t-1},\ 0) + 1 & \text{if } \mathrm{Close}_t > \mathrm{Close}_{t-1} \\ \min(\mathrm{Streak}_{t-1},\ 0) - 1 & \text{if } \mathrm{Close}_t < \mathrm{Close}_{t-1} \\ 0 & \text{if } \mathrm{Close}_t = \mathrm{Close}_{t-1} \end{cases} \\[6pt] \mathrm{ROC1}_t &= \frac{\mathrm{Close}_t - \mathrm{Close}_{t-1}}{\mathrm{Close}_{t-1}} \\[6pt] \mathrm{PercentRank}(\mathrm{ROC1}, P)_t &= \frac{\left|\lbrace\, j : t-P \le j \le t-1,\ \mathrm{ROC1}_j < \mathrm{ROC1}_t \,\rbrace\right|}{P} \times 100 \end{aligned}

where $\mathrm{Close}$ is the input series, $m_1$ is optInTimePeriod, $m_2$ is optInStreakPeriod and $P$ is optInRankPeriod. Both RSIs are Wilder's RSI. The streak is defined from the second bar, taking the streak before it as 0, and $\mathrm{ROC1}_t$ is 0 when $\mathrm{Close}_{t-1} = 0$ .

Notes

CRSI has no unstable period of its own: both RSI legs take TA_FUNC_UNST_RSI. They start warming up on the first bar the call reads (the streak's RSI one bar later, the streak being a change), so a leg that needs fewer bars than the lookback warms up over the spare ones.
Finite input is a precondition. A NaN close does not propagate: the output stays finite and no error is reported, but the values that follow it are wrong.

Inputs

inReal — Close price series

Outputs

outReal — The averaged reading, 0 to 100

Parameters

Parameter	Type	Default	Accepted values	Description
`optInTimePeriod`	integer	3	2–100000	Period of the RSI of the closes
`optInStreakPeriod`	integer	2	2–100000	Period of the RSI of the up/down streak
`optInRankPeriod`	integer	100	2–10000	Number of earlier one-bar returns each return is ranked against

Properties

Numerical Stability: Initial Unstable Period — Inherited from RSI, which CRSI computes internally; tunable via RSI's unstable period.


☐ Overlap Input
✅ Independent Y-Axis i
☐ Candlestick
☐ Can Output NaN or ±Inf
☐ Identity at Period 1

Implementation

TA-Lib Definition: crsi.c · crsi.yaml

Native	File
C	`ta_CRSI.c`
Rust	`crsi.rs`
Java	`Core_CRSI.java`
C#	`Core_CRSI.cs`

TA-Lib is also available for Python, R and more using a wrapper.

Aliases

ConnorsRSI, Connors RSI, Connors Relative Strength Index