# What Is the Median Principle Heuristic in Zhangxuefeng-skill?

> Understand the Median Principle heuristic in Zhangxuefeng-skill. Learn how to make better career and education choices by focusing on the middle 50% of outcomes, ignoring extremes for practical decisions.

- Repository: [花叔/zhangxuefeng-skill](https://github.com/alchaincyf/zhangxuefeng-skill)
- Tags: deep-dive
- Published: 2026-06-12

---

**The Median Principle is a data-driven decision heuristic that advises evaluating career and education choices based on the outcomes of the middle 50% of people who made the same decision, deliberately ignoring both top performers and worst-case scenarios.**

The Median Principle (中文 "中位数原则") serves as one of eight core decision-making heuristics in the `zhangxuefeng-skill` open-source project. This pragmatic rule-of-thumb prioritizes statistical median outcomes over outliers when generating advice about majors, careers, or educational institutions. As implemented in the alchaincyf/zhangxuefeng-skill repository, the principle ensures recommendations reflect sustainable, typical trajectories rather than aspirational exceptions.

## Understanding the Median Principle Heuristic

### Definition and Core Logic

According to the source documentation in [`README.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/README.md), the Median Principle is explicitly defined as: "不看顶尖不看最差，看中间50%的人过得怎么样" (Don't look at the top, don't look at the worst, look at how the middle 50% are doing).

This heuristic operates on a three-step methodology:

1. **Collect real-world outcome data** for graduates of specific majors or universities
2. **Rank these outcomes** by metrics like salary, employment rate, and job stability
3. **Select the median segment** as the primary reference point for recommendations

### Why Ignore the Extremes?

The principle filters out outliers because extreme success stories and failure cases are rarely reproducible for the typical individual. By focusing on the median 50%, the skill generates advice aligned with the most probable and sustainable career trajectory, avoiding the distortion caused by exceptional cases.

## Source Files and Implementation

The heuristic is documented across multiple files in the repository:

- **[`README.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/README.md)**: Lists the Median Principle among eight decision heuristics at the repository root
- **[`references/research/05-decisions.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/references/research/05-decisions.md)**: Reiterates the definition as the second heuristic in the decision research documentation
- **[`SKILL.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/SKILL.md)**: Contains the markdown that agents load to provide Zhang Xuefeng-style answers
- **[`examples/demo-conversation.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/examples/demo-conversation.md)**: Provides sample dialogue demonstrating the principle in action

## Practical Usage Examples

The Median Principle is not a separate API endpoint but a logical sub-routine embedded in the skill's answer generation pipeline.

### Direct Prompt Invocation

```text
> 用张雪峰的视角帮我分析：选读计算机专业的中位数就业前景怎么样？

```

When processing this query, the skill retrieves average salary, employment rate, and typical job titles for the median 50% of computer science graduates, basing its recommendation on reproducible outcomes (e.g., "约10-12万年薪，主要进入互联网公司研发或运维岗位") rather than rare high-profile cases.

### Implicit Invocation via Employment-Backward Method

```text
> 张雪峰，给我做一次就业倒推：如果我想学金融，应该先查哪些中位数数据？

```

The response generates a step-by-step data collection plan, explicitly stating it will "查看金融相关专业毕业生的中位数去向" (check the median destinations of finance graduates) and compare median salaries before forming an opinion.

## Integration with the Employment-Backward Model

The Median Principle aligns with Zhang Xuefeng's broader "employment-backward" (就业倒推法) methodology. Rather than relying on aspirational claims or anecdotal evidence, the skill derives advice from actual post-graduation data sets, applying the median filter to ensure statistical relevance.

## Summary

- The Median Principle filters out career and education outliers to focus on the middle 50% of outcomes
- It is explicitly defined in [`README.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/README.md) and [`references/research/05-decisions.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/references/research/05-decisions.md) as one of eight core heuristics
- The principle operates automatically during answer generation, not as a separate function call
- It supports the "employment-backward" model by prioritizing reproducible, data-driven trajectories over exceptional cases

## Frequently Asked Questions

### How does the Median Principle differ from using average statistics?

The Median Principle specifically targets the middle 50% of outcomes, whereas averages can be skewed by extreme high or low values. By explicitly excluding both top performers and worst-offenders as documented in the source files, the heuristic provides a more realistic baseline for typical individuals that cannot be distorted by outlier salaries or unemployment rates.

### Is the Median Principle a standalone function in the codebase?

No. According to the source analysis, the Median Principle functions as a logical sub-routine embedded within the skill's answer generation pipeline. When users query career or education data, the skill automatically applies this heuristic when relevant data is available in [`SKILL.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/SKILL.md) or reference files.

### Where is the Median Principle documented in the repository?

The principle is documented in two primary locations: [`README.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/README.md) lists it among eight decision heuristics, and [`references/research/05-decisions.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/references/research/05-decisions.md) reiterates the definition as the second heuristic in the decision research file. Both files contain the exact Chinese phrasing: "不看顶尖不看最差，看中间50%的人过得怎么样".

### Can users trigger the Median Principle manually?

Yes. Users can invoke the principle explicitly by asking about "中位数" (median) outcomes in their prompts, or implicitly through the "employment-backward" method (就业倒推法), which automatically incorporates median data collection steps as shown in [`examples/demo-conversation.md`](https://github.com/alchaincyf/zhangxuefeng-skill/blob/main/examples/demo-conversation.md).