rat-pattern-detector

A rat-pattern is a specific three-phase stock-ownership sequence of increase, decrease, then increase. Detection analyzes quarterly holding changes alongside pre-adjusted kline daily data to flag potential rat-pattern signals and generate diagnostic triples.

You detect rat-pattern stock moves by running the detector over a candidate list to analyze quarterly holding changes against kline data. This produces a parquet file of ranked hits with A/B/C/D diagnostics and a bcd_score, plus a diagnostic JSON for tuning.

Yes, the rat-pattern detector requires pre-adjusted, also known as forward-adjusted, kline daily data. Using properly adjusted prices ensures that the quarterly holding change analysis and subsequent A/B/C/D diagnostics are calculated accurately across the three-stage pattern.

Yes, you can configure detection thresholds via CLI arguments. The detector respects these CLI-configurable thresholds during the kline and holdings analysis, and outputs a diagnostic JSON file specifically to assist with threshold tuning and ensuring reproducibility.

The bcd_score is used to rank candidate stocks that exhibit the rat-pattern. It is generated alongside A/B/C/D diagnostics based on quarterly holding sequences, allowing you to prioritize the final parquet hit list for further stock analysis.