StandardError — Standard Error

How far price scatters around its own regression line. Low means the straight line explains the move; high means the trend is a bad description of what happened. The natural companion to LinearRegressionSlope — slope tells you the direction, this tells you whether to believe it.

Signature

StandardError(period, on=<source>)

Argument	Type	Required	Notes
period	int Param or literal	Yes	Regression window. Default 20, bounds 3–400 (three points minimum for a residual).
on=	symbol, OHLCV field, or indicator	No	Source. Defaults to the trade symbol’s close.

Formula

$$ SE = \sqrt{\frac{\sum_i (C_i - \hat{C}_i)^2}{n-2}} $$

with $\hat{C}$ the least-squares fit. The n − 2 denominator is the regression convention: two degrees of freedom go to the slope and intercept.

Warmup

period bars; 0 until then.

Reading the value

se.value
se.value < atr.value

Example

class CleanTrendOnly(Strategy):
    """Trade the slope only when the fit is tight."""
    trade = Symbol()
    sizing = FixedCapital()

    def execute(self):
        se = StandardError(20, on=self.trade)
        slope = LinearRegressionSlope(20, on=self.trade)
        atr = ATR(14, on=self.trade)
        if slope.value > 0 and se.value < atr.value and self.trade.is_flat:
            self.buy(self.trade, reason=f"Clean uptrend, SE {se.value:.2f}")
        if slope.value < 0 and self.trade.is_long:
            self.sell(self.trade, self.trade.position, reason="Slope down")

Notes

• Standard regression statistics.
• Proven by an assert-via-trade run on SPY daily (matched a strategy-side residual computation to 1e-3 on every bar; zero alarm trades).
• In price units — compare it against ATR or divide by price to make it portable.