---
title: '缠论到底有没有用？我回测了162套组合，最好的年化18.88%'
source: 'https://www.youtube.com/watch?v=bFtaYxhNPUI'
video_id: 'bFtaYxhNPUI'
date: 2026-09-12
duration_sec: 217
channel: '有色哥 Yosef'
---

# 缠论到底有没有用？我回测了162套组合，最好的年化18.88%

> Source: [缠论到底有没有用？我回测了162套组合，最好的年化18.88%](https://www.youtube.com/watch?v=bFtaYxhNPUI)

## Summary

The video presents a unique approach to studying Chan Theory (禅论), a technical analysis framework for trading. The creator backtested 162 different interpretations of the theory, turning each into a reproducible, testable trading system. The goal is to resolve long-standing ambiguities in the theory by brute-force testing all possible rule combinations.

### Key Points

- **Annualized return of 18.88%** [00:00] — The creator achieved an 18.88% annualized return using their Chan Theory-based system, but emphasizes that this number is less important than the process behind it.
- **162 versions of Chan Theory** [00:12] — The creator backtested 162 different versions of Chan Theory, each implemented as runnable, reproducible code that can be verified for correctness.
- **The core problem: ambiguity in Chan Theory** [00:49] — Different practitioners define key concepts (fractals, strokes, segments, divergence) differently, leading to fundamentally different trading systems despite claiming to follow the same theory.
- **Examples of ambiguity** [01:16] — Definitions vary for fractals, the number of K-lines between strokes, what constitutes a broken segment, and what counts as divergence. Even the original text sometimes fails to distinguish between 'greater than' and 'greater than or equal to'.
- **Cascading consequences** [01:54] — Because Chan Theory builds layer by layer, small differences in definitions cascade: different fractals lead to different strokes, which lead to different segments, divergences, and ultimately buy/sell signals.
- **The solution: keep all interpretations** [02:07] — Instead of arguing which interpretation is correct, the creator decided to keep all plausible definitions and combine them systematically.
- **Combinatorial explosion** [02:34] — Three fractal definitions times three stroke definitions yields nine combinations; further variations in segments, divergences, and buy/sell points multiply the total to 162 distinct systems.
- **Brute-force testing** [02:58] — The creator implemented and backtested all 162 systems, treating the project as a form of 'brute-force cracking' of Chan Theory.

## Transcript

18.88% 这是我目前跑出来的禅论年化收益 不过在说这个数字之前 我得先解释一件更奇怪的事情
我这次一共回测了162套禅论 你可能会觉得这句话哪里不太对 禅论不是一套理论吗
对 正常来说应该是一套 但我最后硬是做出了 162套 而且这162套
每一套都是能跑起来的代码 每一套都可复现 可以回测 可以被证明对错的谗论交易系统 线段有没有被破坏 背驰有没有发生
非黑即白 不存在感觉 也没有应该 这恰恰就是我为什么要做这件事 因为我研究禅论越久,就越发现一个很有意思的问题,大家都说自己学的是禅论,可真让大家把规则一条一条写出来,他们脑子里的禅论根本不是同一个东西。
分形有人这么定 有人那么定义 比应该隔几根K线 有不同答案 线段怎么才算破坏 中书怎么才算背驰
各有各的理解 甚至有时候 常论原文也没有区分 大于和大于等于 听起来只差一个等号 但常论的问题就在这里他是一层一层往后推的分形变了笔就变笔一变线段中书背驰
买卖点 都跟着变 所以才会出现一个特别荒诞的现象 所有人学的 明明都是一套禅论 最后却能做出完全不同的脚印
我就在想 既然这么多年都没争清楚 那我为什么还要继续争 如果分形有三种说法 那三种都留下
笔有三种 也全留下 三种分形成三种笔 先变成九种 后面线段有分歧 继续成
中书有分歧 继续成 背驰有分歧 再成 最后再接不同的买点和卖点 只要一种说法
能够被严谨地写成算法 我就先不管 它到底是不是正宗 全部留下 全部组合 全部实现 然后
全部跑一遍 所以这件事情 后来就变得有点魔幻 别人研究禅论 是在努力搞懂一套禅论
而我研究到最后 给自己整出了 162套禅论 说白了 我想干的事情有点像 暴力破解禅论
既然不知道哪个答案是真的 那我就尽可能把所有能找到的答案 一个一个试过去