GarmanKlassVolatility — Garman-Klass Volatility

An OHLC volatility estimator: the log range and the log open-to-close move combined, which is more efficient than either the close-to-close or the pure range estimator.

Signature

GarmanKlassVolatility(period, annualization, on=<source>)

Argument	Type	Required	Notes
period	int Param or literal	Yes	Window. Default 20, bounds 1–400.
annualization	int Param or literal	Yes	Bars per year. 252 daily. Bounds 1–100000.
on=	symbol	No	Source symbol. Defaults to the trade symbol. Reads open, high, low, close.

Formula

$$ \sigma_{GK} = \sqrt{A \cdot \frac{1}{n}\sum_i \left[\tfrac{1}{2}\ln^2!\frac{H_i}{L_i} - (2\ln 2 - 1)\ln^2!\frac{C_i}{O_i}\right]} \times 100 $$

The per-bar term can go slightly negative on a doji; the average is floored at zero before the square root.

Warmup

period bars.

Reading the value

gk.value

Example

class GkRegime(Strategy):
    """Trade only in the calm half of the volatility distribution."""
    trade = Symbol()
    sizing = FixedCapital()

    def execute(self):
        gk = GarmanKlassVolatility(20, 252, on=self.trade)
        sma = SMA(50, on=self.trade)
        if gk.value < 18 and self.trade.close > sma.value and self.trade.is_flat:
            self.buy(self.trade, reason=f"GK vol {gk.value:.1f}")
        if gk.value > 30 and self.trade.is_long:
            self.sell(self.trade, self.trade.position, reason="Vol regime shift")

Notes

• Mark Garman and Michael Klass, Journal of Business (1980).
• Like Parkinson’s, it assumes no drift and no overnight gap.