How should I handle NaNs in rolling quantiles?

Use a custom rolling quantile function that filters out NaN values before calculation to match SQL's 'quantileExactExclusive' behavior and prevent NaN poisoning of future results.

What is the 5-Gate Oracle Validation framework?

It is a structured verification process to check Signal Count, Timestamps, Entry Prices, Exit Types, and Kelly Fractions to ensure Python and SQL logic are perfectly aligned.

Why does my Python backtest have fewer trades than my SQL query?

This is often due to missing 'hedging=True' and 'exclusive_orders=False' settings in the Backtest constructor, which prevents overlapping trades in the default backtesting.py configuration.

How do I fix entry price mismatches in stats._trades?

Sort the results by 'EntryTime' rather than the default 'ExitTime'. Because overlapping trades can exit out of order, the default sort often breaks the mapping between SQL signals and Python trades.

Backtesting Oracle Validation

Name: Backtesting Oracle Validation
Author: terrylica

byterrylica

•

数据科学与机器学习

Configures backtesting.py for SQL oracle validation and high-precision range bar pattern analysis.

This skill provides specialized configurations and anti-pattern prevention for synchronizing backtesting.py results with ClickHouse SQL data. It ensures bit-atomic replicability across overlapping trades, multi-position modes, and complex rolling quantile calculations. By implementing the 5-Gate Oracle Validation framework, quant researchers can eliminate discrepancies in signal counts, entry prices, and exit logic between database-driven research and Python-based execution simulations, specifically optimized for range bar methodologies.

主要功能

01Rolling quantile NaN handling to prevent feature poisoning

025-Gate Oracle Validation framework (Signal, Price, Kelly)

03Multi-position mode configuration for concurrent overlapping trades

04Correct trade sorting by EntryTime for SQL parity

0519 GitHub stars

06Time-barrier logic and fixed fractional sizing templates

使用场景

01Troubleshooting signal count mismatches between database and backtest

02Validating ClickHouse SQL sweep results against Python trade simulations

03Backtesting range bar strategies with concurrent positions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add terrylica/cc-skills backtesting-py-oracle

For use in Claude.ai and ChatGPT

Download Skill