Chan Theory Buy/Sell Signals

Problem

When judging A-share moves, price action or single moving-average cues can be misleading, especially in ranges that look like reversals but end as noise. This skill applies the Chan trend framework to turn raw K-line data into a repeatable analysis chain: K-line inclusion, fractal detection, stroke division, pivot identification, and MACD divergence detection. It then classifies signals into first, second, and third buy/sell points, with confidence-aware trading references.

Workflow and Limits

The main entry point is scripts/chan_signal.py, which fetches public TDX K-line data through pytdx and supports daily, weekly, 60-minute, and 30-minute periods. The script computes Chan structure first and returns JSON. A higher-level response can combine latest price, change, current trend, price position relative to the pivot, signal rank, and confidence to form a recommendation. In multi-period use, larger periods are better for direction, while shorter periods help locate entries and exits.

Limits matter: market data comes from public quotes and may reflect the latest close outside trading hours. Default parameters such as minimum stroke length and pivot tolerance may need review in extreme conditions. The simplified segment algorithm and MACD-based divergence can produce false signals in choppy markets. Outputs should retain risk disclaimers and should not be treated as direct buy/sell orders.