roll_cvar¶

Defined in fynance.metrics

roll_cvar(X, alpha=0.05, w=252, method='historical')[source]

Rolling Conditional Value-at-Risk of a price/equity curve.

Causal trailing estimate of cvar: each output point uses only the w returns observed up to and including that point, so the output has a w-point NaN head.

Parameters:

Xarray_like: Time-series of price, performance or index (a single curve), shape (T,).
alphafloat, optional: Tail probability, in (0, 1). Default is 0.05.
wint, optional: Size of the trailing window, in number of returns. Default is 252.
method{‘historical’, ‘gaussian’, ‘cornish_fisher’}, optional: Estimation method, see cvar. Default is ‘historical’.

Returns:

np.ndarray[np.float64, ndim=1] of shape (T,): Rolling Conditional Value-at-Risk; the first w values are NaN.

See also

cvar, roll_var, roll_mdd

Examples

>>> import numpy as np
>>> X = np.array([100., 99., 103., 95., 101., 98., 104., 90., 108., 97.,
...                102.])
>>> out = roll_cvar(X, alpha=0.4, w=5, method='historical')
>>> np.isnan(out[:5]).all()
True
>>> round(out[5], 4)
0.0537