Distribution Metrics

Distribution metrics describe the statistical shape of your trade P&L data. They reveal whether returns follow a "normal" bell curve or have unusual characteristics like fat tails or asymmetry.

Quick Reference

Key Metrics Explained

Skewness

Measures whether returns lean left (negative) or right (positive):

Positive skewness is preferred — it means your outlier trades tend to be winners, not losers.

Kurtosis

Measures how extreme the tails of your distribution are:

High kurtosis means your strategy occasionally produces very large wins or losses. This makes results less predictable.

The 16 Distribution Profiles

AlgoChef combines skewness and kurtosis to classify your strategy into one of 16 profiles.

The following 16 distribution profiles are proprietary to AlgoChef and protected by copyright.

Understanding your distribution profile helps you select the most appropriate Monte Carlo simulation method for stress-testing.

Z-Score

The Z-Score from a "runs test" measures whether wins and losses are randomly distributed or clustered:

• Z-Score near 0 — Random sequence (no pattern)
• Negative Z-Score — Streaks (wins follow wins, losses follow losses)
• Positive Z-Score — Alternating (win-loss-win-loss patterns)

The Z-Probability tells you how statistically significant this pattern is. Above 95% means the pattern is likely real and not due to chance.

Tail Analysis

These metrics go beyond standard skewness/kurtosis to examine the specific behavior of extreme trades.

Tail Ratio is particularly useful. AlgoChef first identifies your statistical outliers — the trades that sit beyond the usual spread of your results, found with the interquartile method rather than a fixed percentile — then reports what share of them are wins. It runs from 0 to 1, and 0.5 is the neutral point: above 0.5 your extremes lean towards wins, below 0.5 your extreme losses outnumber your extreme wins. That is a warning sign for strategies that look profitable on average but carry hidden tail risk.

PnL Tail Ratio asks the same question by size rather than by count: of all the money moved by those outlier trades, what share came from the winning ones. A strategy can have more winning outliers than losing ones and still score badly here if the few losing outliers were far bigger.

Robust Statistics (Winsorized Metrics)

Standard metrics can be distorted by a few extreme outliers. Winsorized metrics trim the most extreme values before calculating, giving a more stable view of "typical" strategy behavior.

Core Edge Robust is the key metric here. If your strategy is profitable overall but not profitable after removing outlier trades, your edge depends on rare events rather than a repeatable process. This is critical for assessing long-term viability.