What is Walk-Forward Efficiency (WFE)?

WFE is a robustness metric defined as the ratio of out-of-sample performance to in-sample performance, helping identify strategies that are likely to fail in live environments.

Does this require specific financial data?

The skill is optimized for financial time-series (using Sharpe ratios and annualization factors), but the underlying logic can be applied to any temporal walk-forward validation task.

Why use log-spaced epoch candidates?

Log spacing is used because early changes in epoch counts usually have a greater impact on convergence than changes at high epoch counts, ensuring better coverage of the search space.

How does this skill differ from early stopping?

While early stopping monitors validation loss during training, this skill evaluates discrete epoch candidates post-hoc to find the optimal 'efficient frontier' across multiple folds.

Adaptive Walk-Forward Epoch Selection

Name: Adaptive Walk-Forward Epoch Selection
Author: terrylica

byterrylica

•

Data Science & ML

Optimizes training epochs for machine learning models using Walk-Forward Efficiency and efficient frontier analysis.

Adaptive Walk-Forward Epoch Selection (AWFES) provides a principled framework for selecting optimal training epochs in time-series and quantitative models. By evaluating discrete epoch candidates across the efficient frontier, it helps developers maximize performance transfer between in-sample and out-of-sample data. This skill automates the calculation of Walk-Forward Efficiency (WFE) and implements data-driven signal-to-noise thresholds to prevent overfitting in financial strategies and predictive workflows.

Key Features

01Walk-Forward Efficiency (WFE) metric calculation

02Efficient frontier analysis for training cost vs. performance

038 GitHub stars

04Signal-to-noise ratio derived Sharpe thresholds

05Bayesian carry-forward priors for cross-fold optimization

06Log-spaced epoch candidate generation for uniform coverage

Use Cases

01Automating walk-forward optimization (WFO) in quant research

02Minimizing overfitting in algorithmic trading strategy development

03Optimizing hyperparameter sweeps for financial time-series models

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add terrylica/cc-skills adaptive-wfo-epoch

For use in Claude.ai and ChatGPT

Download Skill