RSI — Relative Strength Index

A momentum oscillator bounded between 0 and 100. Originally Welles Wilder’s design.

Signature

RSI(period, on=<source>)

Argument	Type	Required	Notes
period	int Param or literal	Yes	Lookback. Wilder’s default is 14.
on=	symbol, OHLCV field, or indicator	No	Source. Defaults to the trade symbol’s close.

Formula

For each bar, compute change Δ = close − previous_close. Split into:

$$ \text{gain}_t = \max(\Delta, 0), \qquad \text{loss}_t = \max(-\Delta, 0) $$

Smooth with Wilder’s exponential moving average over period bars:

$$ \bar{G}t = \frac{(N-1) \bar{G}{t-1} + \text{gain}_t}{N}, \qquad \bar{L}t = \frac{(N-1) \bar{L}{t-1} + \text{loss}_t}{N} $$

Then:

$$ \text{RS} = \frac{\bar{G}_t}{\bar{L}_t}, \qquad \text{RSI} = 100 - \frac{100}{1 + \text{RS}} $$

When loss = 0, RSI is 100 (max). When gain = 0, RSI is 0 (min).

Warmup

period + 1 bars (you need one bar to compute the first change, then period to smooth).

Reading the value

rsi.value
rsi > 70
rsi < 30
rsi.crossed_above(30)       # leaving oversold
rsi.crossed_below(70)       # leaving overbought

Example

class RSIMeanReversion(Strategy):
    """Buy oversold, sell overbought."""
    period = Param(14, min=5, max=30)
    oversold = Param(30, min=15, max=40)
    overbought = Param(70, min=60, max=85)
    trade = Symbol()
    sizing = FixedCapital()

    def execute(self):
        rsi = RSI(self.period, on=self.trade)
        self.stop_loss(3.0)

        if rsi.crossed_above(self.oversold) and self.trade.is_flat:
            self.buy(self.trade, reason=f"RSI {rsi.value:.1f}")

        if rsi.crossed_below(self.overbought) and self.trade.is_long:
            self.sell(self.trade, self.trade.position)

Notes

• Lower period (5–10) reacts to short swings; higher (14–30) tracks longer trends.
• The classic 30/70 thresholds are starting points — different instruments and timeframes calibrate differently. Make them Param()s and let the optimizer find the right pair.
• Smooth RSI to reduce false signals: smooth = EMA(3, on=rsi).