RSI Mean Reversion

Buy when RSI rebounds out of oversold territory; exit when it rebounds out of overbought. The textbook mean-reversion play.

class RSIMeanReversion(Strategy):
    """Buy oversold, sell overbought."""
    rsi_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.rsi_period, on=self.trade)

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

        if rsi.crossed_below(self.overbought) and self.trade.is_long:
            self.sell(self.trade, self.trade.position, reason=f"RSI overbought ({rsi.value:.1f})")

Walkthrough

Why crossover, not threshold

if rsi.value < 30: would fire on every bar RSI is below 30. That would either trigger a no-op (already long) or thrash if you cleared the position elsewhere.

crossed_above(30) fires for exactly one bar — the bar where RSI transitions from ≤ 30 to > 30. That’s the “leaving oversold” moment, which is the classic mean-reversion entry.

Threshold Params

Both 30 and 70 are Param() — let the optimizer find the best pair. Defaults match the Wilder convention; lower thresholds (40/60) trade more often, higher thresholds (20/80) trade more selectively.

F-string reasons

reason=f"RSI oversold ({rsi.value:.1f})" captures the RSI value at signal time. Reasons are saved with every fill and show up in the trade log — distinct from print(...) output, which appears on the Console tab.

No stop loss

Mean-reversion strategies often skip stops — the thesis is “price will revert to the mean,” and a stop loss can lock in the trough. But a runaway loss is a runaway loss. A reasonable middle ground:

self.stop_loss(5.0)   # wider than typical to give the mean-reversion thesis room

What to tune

Param	Typical range	Effect
rsi_period	5..30	Lower = more sensitive (more signals); higher = smoother.
oversold	15..40	Lower = entries only on extreme dips.
overbought	60..85	Higher = ride the trend longer.

Common modifications

Symmetric short side

if rsi.crossed_below(self.overbought) and self.trade.is_flat:
    self.sell_short(self.trade, reason=f"RSI overbought ({rsi.value:.1f})")

if rsi.crossed_above(self.oversold) and self.trade.is_short:
    self.buy_to_cover(self.trade, self.trade.position, reason="Cover short")

Smoothed RSI

If signals fire too often, smooth the RSI with an EMA:

rsi = RSI(self.rsi_period, on=self.trade)
smooth = EMA(3, on=rsi)

if smooth.crossed_above(self.oversold) and self.trade.is_flat:
    self.buy(self.trade)

Regime filter

Mean reversion fails in strong trends. Filter for trendless markets with Aroon — when neither .up nor .down is elevated, no recent extreme has printed and the market is ranging:

class RangingRSI(Strategy):
    """RSI mean reversion, gated on a trendless Aroon regime."""
    rsi_period = Param(14, min=2, max=50)
    oversold = Param(30, min=10, max=45)
    trade = Symbol()
    sizing = FixedCapital()

    def execute(self):
        rsi = RSI(self.rsi_period, on=self.trade)
        aroon = Aroon(14, on=self.trade)

        # Only enter when no recent high/low extreme has printed (ranging market)
        if aroon.up < 50 and aroon.down < 50 and rsi.crossed_above(self.oversold) and self.trade.is_flat:
            self.buy(self.trade)