What is a 'binding metric' in this context?

A binding metric is a constraint that, if relaxed, would immediately increase the number of surviving configurations. Identifying these helps researchers focus on the most restrictive quality dimensions.

How does the intersection logic improve selection?

Unlike weighted sums which can hide a failure in one metric with high scores in others, intersection logic requires a configuration to pass every individual cutoff, ensuring multi-dimensional excellence.

Why does this skill use percentiles instead of raw values?

Raw values often have incompatible scales (e.g., a ratio vs. a raw count). Percentiles normalize all metrics to a 0-100 range, making independent cutoffs and cross-metric comparison mathematically sound.

Which optimization algorithms are supported?

The methodology is designed to work with Optuna, supporting TPE (Tree-structured Parzen Estimator) for single-objective search and NSGA-II for multi-objective Pareto frontier discovery.

Evolutionary Metric Ranking

Name: Evolutionary Metric Ranking
Author: terrylica

byterrylica

•

Data Science & ML

Optimizes multi-metric configuration selection using evolutionary search and percentile-based cutoff filtering for quantitative research.

Evolutionary Metric Ranking provides a systematic framework for filtering and ranking configurations across multiple, often conflicting, quality metrics. By normalizing raw values into percentile ranks and employing multi-objective evolutionary optimization via Optuna, it identifies the most robust 'survivor' sets. This skill is essential for quant researchers and data scientists who need to move beyond simple weighted scoring to find configurations that meet stringent requirements across all dimensions of performance.

Key Features

01Automated identification of inert and binding performance dimensions

02Independent per-metric cutoff tuning to identify binding constraints

03Forensic post-analysis for cross-objective consistency and feature frequency

04Multi-objective evolutionary optimization using TPE and NSGA-II samplers

0520 GitHub stars

06Percentile-based normalization for scale-agnostic metric comparison

Use Cases

01Selecting high-quality quantitative trading strategies from backtest results

02Filtering machine learning model hyperparameters across multiple validation metrics

03Performing forensic analysis on complex optimization frontiers to find universal champions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add terrylica/cc-skills evolutionary-metric-ranking

For use in Claude.ai and ChatGPT

Download Skill