The RSI indicator, explained properly — most traders use it wrong

The Relative Strength Index is the most widely used — and most widely misused — indicator in retail trading. The folklore version is simple: RSI above 70 means overbought, sell; below 30 means oversold, buy. Backtest that rule on crypto and it loses money with impressive consistency. The indicator isn't broken. The folklore is.

What RSI actually measures

RSI, developed by J. Welles Wilder, compares the size of recent gains to the size of recent losses over a lookback window — 14 bars by default. It outputs a number from 0 to 100. High RSI means recent up-moves have dominated; low RSI means down-moves have. That is the whole of it. RSI is a momentum oscillator: it describes the recent past. It does not predict a reversal, and it never claimed to.

The crucial misunderstanding is treating RSI as a level when it behaves as a regime gauge. In a strong trend, RSI can sit above 70 for weeks while price keeps climbing. Anyone shorting each touch of 70 is fighting the trend and paying for it.

Why 70/30 is a trap

The 70/30 thresholds were chosen for 1970s commodity markets and a 14-day period. Crypto is more volatile and trends harder. Two consequences follow. First, in a bull leg, RSI overbought is a sign of strength, not exhaustion — "overbought" markets routinely get more overbought. Second, the same RSI value means different things in different volatility regimes. A static 70 line cannot adapt; your rules have to.

Three uses that actually backtest well

RSI earns its keep when it is one condition among several, not a standalone trigger:

• Mean reversion inside a range — only when a separate filter confirms the market is ranging (e.g. ADX below 20), a deep RSI reading is a usable pullback signal. The filter does the heavy lifting; RSI just times the entry.
• Trend pullback entries — in an uptrend, a dip to RSI 40-50 (not 30) is a 'buy the dip' timing tool. You are using RSI to enter a trend, not to fade it.
• Divergence as a warning, not a trigger — when price makes a higher high but RSI makes a lower high, momentum is fading. Treat it as a reason to tighten stops or stop adding, not as a standalone short.

Tuning the period — and not over-tuning it

A shorter RSI period (say 7) is jumpier and fires more signals; a longer one (21) is smoother and slower. It is tempting to sweep the period to find the value that maximised past returns — and that is exactly how you overfit. If RSI-13 backtests beautifully and RSI-12 and RSI-14 fall apart, you have not found a parameter, you have found noise. Validate any RSI rule with walk-forward so the period is chosen on data the optimizer hasn't seen.

strategy:
  name: rsi_pullback
  indicators:
    - { id: rsi,  kind: RSI, period: 14 }
    - { id: ema,  kind: EMA, period: 200 }
    - { id: adx,  kind: ADX, period: 14 }
  rules:
    entry:
      all:
        - { type: above, left: close, right: ema }   # uptrend only
        - { type: below, left: rsi,   right: 45 }     # shallow pullback
    exit:
      any:
        - { type: above, left: rsi, right: 65 }
  risk:
    size_pct: 0.5
    stop_loss_atr: 2.0

In Noon Barbari, RSI is one of the built-in indicators in the indicator library, and the visual designer lets you gate it behind trend and volatility filters without code. Build the rule above, backtest it, then walk-forward the period. The free tier covers all of that for one strategy.

The short version: RSI is a momentum gauge, not a reversal alarm. Use it to time entries within a regime you have already identified by other means — and the indicator that loses money as a standalone signal starts pulling its weight.