sigmaquant.performance.metrics.excess_kurtosis#
- sigmaquant.performance.metrics.excess_kurtosis(returns)#
Compute the excess kurtosis of a return series.
Excess kurtosis measures the degree of tail heaviness of a distribution relative to a normal distribution. A normal distribution has zero excess kurtosis.
- Parameters:
returns – Sequence of periodic returns or PnL values.
- Returns:
Excess kurtosis of the input series.
- Return type:
float
Notes
A normal distribution has zero excess kurtosis.
NaN values are ignored. No finite-sample bias correction is applied.
The excess kurtosis estimator is:
\[\widehat{\kappa} = \frac{1}{T} \sum_{t=1}^{T} \left( \frac{r_t - \bar{r}}{s} \right)^4 - 3\]