RollingOmega

Description

RollingOmega is the Omega ratio (Keating and Shadwick, 2002) over a trailing window of returns about a threshold,

omega = sum( (return - threshold)+ ) / sum( (threshold - return)+ ),

the total gain above the threshold divided by the total shortfall below it. Unlike the Sharpe ratio, Omega uses the whole return distribution, so it reflects skew and fat tails rather than just the mean and variance. A value above 1 means gains outweigh losses at that threshold. A window with no downside (a zero denominator) has an undefined ratio and returns NaN, as does a NaN return (nan_policy: ignore).

Examples

Usage plot

import numpy as np
import plotly.graph_objects as go
from plotly.subplots import make_subplots
from screamer import RollingOmega

rng = np.random.default_rng(1)
n = 500
r = rng.standard_normal(n) * 0.01 + 0.0015          # a small positive drift
omega = RollingOmega(60)(r)

fig = go.Figure()
fig.add_trace(go.Scatter(y=omega, name='Omega', line=dict(color='steelblue')))
fig.add_hline(y=1.0, line=dict(color='gray', dash='dot'),
              annotation_text='gains = losses')
fig.update_layout(title='RollingOmega: gains vs losses about the threshold (>1 favours gains)',
                  xaxis_title='bar', yaxis_title='Omega',
                  margin=dict(l=20, r=20, t=60, b=20), legend=dict(orientation='h', yanchor='bottom', y=1.02, xanchor='right', x=1))
fig.show()