AverageDeviation — Average Deviation

Mean absolute deviation from the window’s mean. Unlike standard deviation it does not square, so a single outlier moves it much less — the robust dispersion measure, and the one inside CCI.

Signature

AverageDeviation(period, on=<source>)

Argument	Type	Required	Notes
period	int Param or literal	Yes	Window. Default 20, bounds 1–400.
on=	symbol, OHLCV field, or indicator	No	Source. Defaults to the trade symbol’s close.

Formula

$$ \text{AD} = \frac{1}{n}\sum_i |x_i - \bar{x}| $$

Warmup

period bars; partial windows use the bars available.

Reading the value

ad.value
ad.value < sd.value

Example

class RobustBands(Strategy):
    """Bands built from absolute deviation rather than standard deviation."""
    trade = Symbol()
    sizing = FixedCapital()

    def execute(self):
        ad = AverageDeviation(20, on=self.trade)
        sma = SMA(20, on=self.trade)
        if self.trade.close < sma.value - 2 * ad.value and self.trade.is_flat:
            self.buy(self.trade, reason="Two mean-deviations low")
        if self.trade.close > sma.value and self.trade.is_long:
            self.sell(self.trade, self.trade.position, reason="Back to the mean")

Notes

• Proven by an assert-via-trade run on SPY daily (matched a strategy-side reference to 1e-3 on every bar; zero alarm trades).
• For a normal distribution AD ≈ 0.8 σ; on fat-tailed data the gap widens, and that gap is itself informative.