Connors RSI: 3-Component Mean Reversion for Equity Indices

Connors RSI is a proprietary three-component oscillator designed specifically for mean reversion in equity indices and liquid ETFs — combining price RSI, momentum streak RSI, and rate-of-change percentile into a single composite signal with documented historical edge.

Section 1: Core Mechanics

Larry Connors designed ConnorsRSI to address a key weakness in standard RSI: a single RSI reading combines multiple different market phenomena — price distance from mean, momentum streak duration, and relative rate of change — into one number. ConnorsRSI separates and measures each component independently, then averages them.

Formula

Component 1 — Short RSI of price:

Standard RSI calculation over 3 periods. Captures whether price is overbought or oversold on a very short-term basis. RSI(3) is intentionally aggressive — it reaches extremes faster than the standard RSI(14).

Component 2 — Streak RSI:

First, compute the streak: the number of consecutive days price closed higher (positive integer) or lower (negative integer). Example: 3 consecutive up days = streak of +3. 2 consecutive down days = streak of -2. A reversal day resets the streak to ±1.

RSI applied to the streak series over 2 periods. This measures whether the current run of consecutive up/down days is extreme relative to recent streak history.

Component 3 — PercentRank of 1-day ROC:

Where is the 1-day percentage change, and PercentRank measures where today's ROC falls in the distribution of the past 100 daily ROC values. Output: 0 (lowest ever) to 100 (highest ever).

Final composite:

Inputs

• Price source: Close price (standard)
• RSI(3) period: Fixed at 3 — not typically varied
• Streak RSI period: Fixed at 2 — not typically varied
• PercentRank lookback: 100 days — the primary adjustable parameter

Parameters

Output Range and Meaning

ConnorsRSI outputs 0 to 100. Unlike standard RSI, the operative thresholds are much more extreme:

• CRSI below 10 = extreme oversold — primary long signal for index ETFs
• CRSI above 90 = extreme overbought — primary exit / short signal
• CRSI at 50 = neutral — no signal

Visual Behavior

CRSI plots as a single smooth line between 0 and 100. It drops to extreme lows more frequently than RSI(14) but less frequently than StochRSI — representing a middle ground in sensitivity. In trending markets it will drift in one direction without reverting.

Section 2: Interpretation & Signals

Signal Zones

Entry and Exit Rules

Long entry: CRSI drops below 10 on a confirmed daily bar close. Enter the next day's open. This is the primary signal for S&P 500 ETFs (SPY, QQQ, IWM), sector ETFs, and large-cap liquid stocks.

Hold period: Connors research showed optimal holds of 1 to 5 trading days. The mean reversion typically completes within this window. Do not hold longer — if CRSI does not rise above 30 within 5 days, the thesis has failed.

Exit: CRSI rises above 70 (conservative exit) or 90 (aggressive exit to capture full reversion). Most implementations use 70 as the standard exit threshold.

Short entry (more cautious): CRSI rises above 90 on a confirmed close. Enter short at next day's open. The short side is less reliable than the long side — indices have an upward drift bias.

Streak Component Deep Dive

The streak is the most unique element of ConnorsRSI. A streak of +5 means price closed higher for 5 consecutive days. RSI(2) of this streak series captures whether that streak length is extreme relative to recent streak history. A streak of +5 occurring when the 2-period average streak is only +1 produces a high RSI(2) of streak — adding to the overbought CRSI reading.

Divergence

Standard divergence analysis is less applicable to CRSI than to price-based oscillators. Instead, watch for CRSI re-entry from above: CRSI drops below 10, price bounces back to neutral, then CRSI drops below 10 again. The second re-entry with a shallower price low (price not making a new low) is the stronger signal — momentum of selling is exhausted.

False Signals

Best Market Conditions

CRSI is purpose-built for:

• SPY, QQQ, IWM, XLF, and similar large-cap liquid ETFs
• S&P 500 constituent large-cap stocks in correction phases
• Range-bound or mild bear phases, not crash conditions (use market-regime filter)

Chart Setup — ConnorsRSI Below 10 Entry Signal

Section 3: Pass vs. Live — Real-Time Reliability

All three components of ConnorsRSI are calculated on confirmed closed bars. The streak count updates only when a new bar closes — there is no intra-bar ambiguity. PercentRank compares today's confirmed ROC against the prior 100 confirmed daily closes. This makes ConnorsRSI one of the cleaner indicators for backtesting — no lookahead bias, no repaint risk.

Section 4: Practical Use Cases

Real example and documented performance: Larry Connors' research tested ConnorsRSI below 10 signals on the S&P 500 from 1995 through 2010. Over 1,100 tested signals with 1-to-5-day holds, the strategy showed approximately 74-76% winning trades with average gains of 0.8% per trade (annualized to approximately 15% net return). The edge decays significantly when applied to individual stocks — the index-level diversification effect is part of what makes CRSI work.

Section 5: Pseudo Code

INPUT: close_prices[], rsi_period=3, streak_rsi_period=2, pct_rank_period=100

PROCESS:
  Step 1: Calculate Component 1 — RSI(3) of close
            avg_gain, avg_loss via Wilder smoothing (RMA)
            c1[i] = RSI value at bar i using 3-period lookback

  Step 2: Calculate streak for each bar i:
            if close[i] > close[i-1]:
                if streak[i-1] > 0:
                    streak[i] = streak[i-1] + 1
                else:
                    streak[i] = 1
            elif close[i] < close[i-1]:
                if streak[i-1] < 0:
                    streak[i] = streak[i-1] - 1
                else:
                    streak[i] = -1
            else:
                streak[i] = 0  # flat day resets streak

  Step 3: Calculate Component 2 — RSI(2) of streak
            c2[i] = RSI value at bar i using streak[] series and 2-period lookback

  Step 4: Calculate Component 3 — PercentRank of 1-day ROC over 100 bars
            roc[i] = (close[i] - close[i-1]) / close[i-1] * 100
            window = roc[i - pct_rank_period + 1 : i + 1]
            rank = sum(1 for r in window if r <= roc[i])
            c3[i] = (rank / pct_rank_period) * 100

  Step 5: ConnorsRSI = average of three components
            crsi[i] = (c1[i] + c2[i] + c3[i]) / 3

OUTPUT:
  crsi[] — values 0 to 100
  buy_signal: crsi[i] < 10 and close[i-1] > sma_200[i-1] (regime filter)
  exit_signal: crsi[i] > 70

EDGE CASES:
  - Minimum bars: rsi_period + pct_rank_period + streak_rsi_period + buffer ≈ 110 bars
  - Flat close (close[i] == close[i-1]): streak = 0, treat as streak reset
  - Crash environment: disable when SPY below 200-day SMA — return DISABLED flag
  - PercentRank ties: count all bars with ROC equal to or less than current

Section 6: Parameters & Optimization

Standard Parameter Conventions

Parameter Impact

Market-Specific Adjustments

• SPY/QQQ: Default parameters — designed specifically for these instruments
• Sector ETFs (XLF, XLE): Raise buy threshold to 12 — sectors move more than the broad index
• International ETFs (EEM, EFA): Extend PercentRank to 150 — international cycles are longer
• Individual large-cap stocks: Use with 200-day SMA filter AND earnings calendar exclusion

Section 7: Synergies & Conflicts

Section 8: Common Mistakes

Section 9: Cheat Sheet