Connors RSI (CRSI)

Haiyue

April 1, 2026

11min

I. What is Connors RSI

Connors RSI (CRSI) is a composite momentum indicator that synthesizes three independent momentum components into a single oscillator. It combines the RSI of short-term price changes, the RSI of consecutive up/down day streaks, and the percentile rank of recent price changes, aiming to capture short-term price reversal opportunities more precisely than traditional RSI.

Historical Background

CRSI was developed by Larry Connors and Cesar Alvarez, published across several of Connors’ books including Short Term Trading Strategies That Work. Larry Connors is the founder of TradingMarkets.com and is renowned for researching and promoting short-term mean reversion strategies. He found that traditional RSI(14) is too slow for short-term trading, while very short-period RSI (such as RSI(2)) is sensitive but one-dimensional, so he designed CRSI to fuse multiple dimensions of short-term momentum information.

Indicator Classification

Type: Oscillator, displayed in a separate panel
Category: Momentum — composite indicator
Default Parameters: RSI period $= 3$ , Streak RSI period $= 2$ , Rank period $= 100$
Value Range: 0 — 100

Three Components of CRSI

RSI(Close, 3): Short-term price momentum
RSI(Streak, 2): Momentum of consecutive up/down streak
PercentRank(ROC(1), 100): Percentile rank of today’s return over the past 100 days

II. Mathematical Principles and Calculation

Core Formula

$CRSI = \frac{RSI(C, 3) + RSI(Streak, 2) + PercentRank(ROC_1, 100)}{3}$

Component 1: RSI(Close, 3)

Standard RSI formula with period shortened to 3:

$RSI_3 = 100 - \frac{100}{1 + RS_3}$

Component 2: RSI(Streak, 2)

Streak (consecutive up/down count) definition:

If $C_t > C_{t-1}$ , Streak increments by 1 (from 0 or positive)
If $C_t < C_{t-1}$ , Streak decrements by 1 (from 0 or negative)
If $C_t = C_{t-1}$ , Streak resets to zero

Example: 3 consecutive up days — Streak = 3; 2 consecutive down days — Streak = -2.

Then calculate RSI(2) on the Streak series:

$RSI_{Streak} = RSI(Streak, 2)$

Component 3: PercentRank(ROC(1), 100)

ROC(1): Single-day return

$ROC_1 = \frac{C_t - C_{t-1}}{C_{t-1}} \times 100$

PercentRank: Percentile rank of current ROC(1) among the past 100 ROC(1) values

$PercentRank = \frac{\text{count of past 100 values less than current ROC(1)}}{100} \times 100$

Step-by-Step Calculation Logic

Calculate RSI(Close, 3): 3-period RSI using Wilder Smoothing
Build Streak series: Calculate the consecutive up/down count for each day
Calculate RSI(Streak, 2): 2-period RSI on the Streak series
Calculate ROC(1): Daily percentage return
Calculate PercentRank: Rank of current ROC(1) within the past 100 periods
Composite: Simple average of all three

Why Three Components?

RSI(3) captures the absolute level of short-term price momentum
RSI(Streak, 2) captures the persistence/continuity of price movement
PercentRank captures the historical rarity of today’s change

Each has a different focus; combined, they provide a more comprehensive short-term momentum profile.

III. Python Implementation

import numpy as np
import pandas as pd


def rsi_wilder(series: pd.Series, period: int) -> pd.Series:
    """Calculate RSI using Wilder Smoothing"""
    delta = series.diff()
    gain = delta.where(delta > 0, 0.0)
    loss = (-delta).where(delta < 0, 0.0)

    avg_gain = gain.copy()
    avg_loss = loss.copy()
    avg_gain.iloc[:period] = np.nan
    avg_loss.iloc[:period] = np.nan

    first_idx = period
    avg_gain.iloc[first_idx] = gain.iloc[1:first_idx + 1].mean()
    avg_loss.iloc[first_idx] = loss.iloc[1:first_idx + 1].mean()

    for i in range(first_idx + 1, len(series)):
        avg_gain.iloc[i] = (avg_gain.iloc[i - 1] * (period - 1) + gain.iloc[i]) / period
        avg_loss.iloc[i] = (avg_loss.iloc[i - 1] * (period - 1) + loss.iloc[i]) / period

    rs = avg_gain / avg_loss
    return 100 - 100 / (1 + rs)


def calc_streak(close: pd.Series) -> pd.Series:
    """Calculate consecutive up/down day streak"""
    streak = pd.Series(0, index=close.index, dtype=float)
    for i in range(1, len(close)):
        if close.iloc[i] > close.iloc[i - 1]:
            streak.iloc[i] = max(streak.iloc[i - 1], 0) + 1
        elif close.iloc[i] < close.iloc[i - 1]:
            streak.iloc[i] = min(streak.iloc[i - 1], 0) - 1
        else:
            streak.iloc[i] = 0
    return streak


def percent_rank(series: pd.Series, period: int) -> pd.Series:
    """Calculate percentile rank"""
    def rank_func(x):
        current = x[-1]
        past = x[:-1]
        return np.sum(past < current) / len(past) * 100
    return series.rolling(window=period + 1, min_periods=period + 1).apply(
        rank_func, raw=True
    )


def connors_rsi(close: pd.Series,
                rsi_period: int = 3,
                streak_period: int = 2,
                rank_period: int = 100) -> pd.DataFrame:
    """
    Calculate Connors RSI

    Parameters
    ----------
    close : pd.Series
        Close price series
    rsi_period : int
        Price RSI period, default 3
    streak_period : int
        Streak RSI period, default 2
    rank_period : int
        PercentRank lookback period, default 100

    Returns
    -------
    pd.DataFrame
        DataFrame with RSI, RSI_Streak, PercentRank, and CRSI columns
    """
    # Component 1: RSI(Close, 3)
    rsi_close = rsi_wilder(close, rsi_period)

    # Component 2: RSI(Streak, 2)
    streak = calc_streak(close)
    rsi_streak = rsi_wilder(streak, streak_period)

    # Component 3: PercentRank(ROC(1), 100)
    roc1 = close.pct_change() * 100
    pct_rank = percent_rank(roc1, rank_period)

    # Composite
    crsi = (rsi_close + rsi_streak + pct_rank) / 3

    return pd.DataFrame({
        "RSI_3": rsi_close,
        "RSI_Streak": rsi_streak,
        "PctRank": pct_rank,
        "CRSI": crsi
    }, index=close.index)


# ========== Usage Example ==========
if __name__ == "__main__":
    np.random.seed(42)
    dates = pd.date_range("2024-01-01", periods=200, freq="D")
    price = 100 + np.cumsum(np.random.randn(200) * 1.0)

    df = pd.DataFrame({
        "date": dates,
        "open":  price + np.random.randn(200) * 0.3,
        "high":  price + np.abs(np.random.randn(200) * 0.6),
        "low":   price - np.abs(np.random.randn(200) * 0.6),
        "close": price,
        "volume": np.random.randint(1000, 10000, size=200),
    })
    df.set_index("date", inplace=True)

    # Calculate Connors RSI
    result = connors_rsi(df["close"])
    df = pd.concat([df, result], axis=1)

    print("=== Connors RSI Results (last 15 rows) ===")
    print(df[["close", "RSI_3", "RSI_Streak", "PctRank", "CRSI"]].tail(15).to_string())

    # Mean reversion signals
    df["signal"] = np.where(
        df["CRSI"] < 10, 1,      # Extremely oversold -> Buy
        np.where(df["CRSI"] > 90, -1, 0)  # Extremely overbought -> Sell
    )

    print("\n=== Extremely Oversold Signals (CRSI < 10) ===")
    print(df[df["signal"] == 1][["close", "CRSI"]].head(10).to_string())
    print("\n=== Extremely Overbought Signals (CRSI > 90) ===")
    print(df[df["signal"] == -1][["close", "CRSI"]].head(10).to_string())

IV. Problems the Indicator Solves

1. Short-Term Mean Reversion Timing

CRSI is specifically designed for short-term mean reversion strategies:

CRSI < 10: All three dimensions show extreme oversold — strong buy signal
CRSI > 90: All three dimensions show extreme overbought — strong sell signal

2. Multi-Dimensional Momentum Assessment

Component	Dimension	Information
RSI(3)	Price momentum	Speed of short-term price changes
RSI(Streak, 2)	Persistence	Strength of consecutive up/down sequences
PercentRank	Historical position	Historical rarity of today’s change

3. Quantifying Consecutive Streaks

Traditional indicators ignore the “continuity” information of price movement. Through the Streak component, CRSI can identify overbought/oversold conditions after consecutive days of rise or decline.

4. Relative Position Assessment

The PercentRank component answers a key question: “Where does today’s change rank historically?” If the rank is extremely low (today’s decline is worse than most of the past 100 days), it suggests oversold conditions.

Caution

V. Advantages, Disadvantages, and Use Cases

Advantages

Advantage	Description
Multi-dimensional fusion	Combines three independent dimensions: price, persistence, and historical rank
Ultra-short-term sensitivity	RSI(3) and RSI(Streak, 2) with very short periods react extremely fast
Mean reversion specialized	Specifically optimized for short-term reversal strategies
Clear signals	Extreme values (< 10 or > 90) appear at moderate frequency with high win rates

Disadvantages

Disadvantage	Description
Fails in trending markets	Counter-trend signals during sustained trends can cause large losses
Short-term only	Designed for holding periods of 1-5 days
Requires trend filter	Cannot be used independently; must be combined with trend analysis
Complex calculation	Three components each with their own steps; relatively complex implementation

Use Cases

Short-term mean reversion strategies: Buying pullbacks in uptrends, selling bounces in downtrends
ETF trading: CRSI performs excellently on liquid ETFs
1-5 day holding period: Connors’ backtests show CRSI works best with short holding periods

Comparison	CRSI	RSI(14)	RSI(2)	StochRSI
Period	Ultra-short	Medium	Ultra-short	Medium-derived
Dimensions	Three	One	One	Second-order
Strategy Type	Mean Reversion	General	Mean Reversion	General
Thresholds	10/90	30/70	5/95	20/80

tip Practical Tips

Must add trend filter: Only consider CRSI oversold buy signals when price is above the 200-day moving average.
Extreme thresholds: Using CRSI < 5 or > 95 yields higher win rates, but trading frequency drops significantly.
Profit-taking rules: Connors suggests taking profit on longs when CRSI > 70, and covering shorts when CRSI < 30.
Suited for systematic execution: CRSI strategy rules are clear and well-suited for automated trading programs.

Simple Moving Average (SMA)

Exponential Moving Average (EMA)

Weighted Moving Average (WMA)

Double Exponential Moving Average (DEMA)

Triple Exponential Moving Average (TEMA)

Simple Moving Average (SMA)

Exponential Moving Average (EMA)

Weighted Moving Average (WMA)

Double Exponential Moving Average (DEMA)

Triple Exponential Moving Average (TEMA)

Connors RSI (CRSI)

I. What is Connors RSI

Historical Background

Indicator Classification

II. Mathematical Principles and Calculation

Core Formula

Component 1: RSI(Close, 3)

Component 2: RSI(Streak, 2)

Component 3: PercentRank(ROC(1), 100)

Step-by-Step Calculation Logic

III. Python Implementation

IV. Problems the Indicator Solves

1. Short-Term Mean Reversion Timing

2. Multi-Dimensional Momentum Assessment

3. Quantifying Consecutive Streaks

4. Relative Position Assessment

V. Advantages, Disadvantages, and Use Cases

Advantages

Disadvantages

Use Cases

I. What is Connors RSI

Historical Background

Indicator Classification

II. Mathematical Principles and Calculation

Core Formula

Component 1: RSI(Close, 3)

Component 2: RSI(Streak, 2)

Component 3: PercentRank(ROC(1), 100)

Step-by-Step Calculation Logic

III. Python Implementation

IV. Problems the Indicator Solves

1. Short-Term Mean Reversion Timing

2. Multi-Dimensional Momentum Assessment

3. Quantifying Consecutive Streaks

4. Relative Position Assessment

V. Advantages, Disadvantages, and Use Cases

Advantages

Disadvantages

Use Cases

Comparison with Related Indicators