risk lab

Risk Lab — past Sharpe is the easy part.

Pairs-trading P&L is non-Gaussian. This page reports the metrics desks actually run after Sharpe: VaR, CVaR/Expected Shortfall, Ulcer Index, Pain & Sterling ratios, information ratio, return-distribution moments, and a stationary-bootstrap CI on the Sharpe.

Pair

Distributional risk

historical, daily-bar

Skew-0.37left-skewed

Excess kurt18.300 = Gaussian

VaR 95%-0.79%

VaR 99%-2.44%

CVaR 95%-1.81%

CVaR 99%-3.17%

Ulcer idx0.1856

Max losing streak6consecutive losing bars

Quality ratios

Sharpe (point)-0.41

Sortino-0.16

Calmar-0.16

Pain ratio-0.23

Sterling-0.66ann. ret / avg ann. DD

Pain ratio = annual return / Ulcer Index. Sterling = annual return divided by the average yearly drawdown. Both are smoother than Calmar (which depends on a single worst-DD point) and tend to correlate better with subjective sleeplessness.

Sharpe — stationary bootstrap CI

400 resamples · block ≈ 21 bars

Median Sharpe-0.43

90% CI-0.86, 0.055th–95th percentile

95% CI-0.90, 0.142.5th–97.5th

Median total-19.0%

Total 90% CI-37%, +1%

Information ratio-0.44vs market

Politis & Romano (1994) stationary bootstrap: random block lengths preserve the strong-mixing property of the original return series so the implied Sharpe distribution is honest under serial correlation. A wide 95% CI on Sharpe is the simplest way to spot data-mined results.

Risk Lab — past Sharpe is the easy part.

Distributional risk

Quality ratios

Sharpe — stationary bootstrap CI

Daily-return distribution

Bootstrap Sharpe distribution