The Honest Number诚实的数字
Most books on Chinese metaphysics only show cases that worked. The failures are quietly omitted. We think that's the wrong approach — and we're not the first to say so. As one senior practitioner put it: "Presenting only successful examples isn't scientific, and it isn't honest. It makes readers think prediction is infallible. That helps no one."
大多数术数著作只展示测准的案例，测错的悄悄略去。我们认为这是不对的。而且我们不是第一个这么说的人。一位资深实践者曾直言："著书只讲测准的例子，不讲测不准的例子，这不是科学的实事求是的态度。让读者以为预测能百发百中，谁也帮不了。"
So here is the honest number: experienced Qimen practitioners, working with proficiency, report an accuracy rate of roughly 70–80 %. That figure carries two meanings. First, out of a hundred readings, seventy to eighty produce a result that matches what actually happens. Second, within a single complex reading, seventy to eighty percent of the specific claims hold up.
所以这里是诚实的数字：经验丰富、运用纯熟的奇门实践者，报告的准确率大约在 70–80%。这个数字包含两层含义。第一，测一百人或一百件事，有七八十件与实际结果吻合。第二，测一个人或一件事的多个细节，其中七八成以上的判断经得起验证。
That is not a weakness; it is a feature of any honest predictive system. Even modern medicine does not cure 100 % of patients. Even weather forecasting, backed by satellites and supercomputers, misses. The Qing-dynasty editors of the Siku Quanshu, the largest scholarly compilation in Chinese history, called Qimen Dunjia "the most principled of all the arts" (于方技之中，最有理致). Principled does not mean infallible. It means the system has structure, logic, and a trackable chain of reasoning, which is precisely what makes it possible to measure, improve, and eventually teach to a machine.
这不是缺陷，这是任何诚实的预测体系的特征。现代医学也做不到治愈率 100%。天气预报有卫星和超级计算机支撑，也会失误。清代《四库全书》编纂者（中国历史上最大的学术丛书）称奇门遁甲"于方技之中，最有理致"。有理致不等于百发百中。它意味着这个体系有结构、有逻辑、有可追溯的推理链条，而正是这一点，使得衡量、改进、并最终教给机器成为可能。
"Accuracy is always relative;
inaccuracy is absolute.
Just as absolute truth is the sum
of countless relative truths.""测准只是相对的，
测不准才是绝对的。
正如绝对真理
只是无数相对真理的总和。"
Why Not 100 %? The Four Limits为什么不是 100%：四大局限
Understanding why readings miss is more useful than pretending they don't. Four factors account for nearly all inaccuracy in traditional Qimen prediction:
厘清为什么会测不准，比假装从不失手有用得多。传统奇门预测中，几乎所有的不准确都可以归结为四个因素：
What Raises Accuracy: The Signal-Strength Rule什么提高准确率：信号强度法则
Not all readings are created equal. Practitioners have long observed that accuracy varies predictably with how strongly the question connects to the moment of asking. We call this the signal-strength rule:
并非所有解读的准确率都一样。实践者早已观察到，准确率随"问题与问事时刻的关联强度"呈可预测的变化。我们称之为信号强度法则：
| Scenario | Signal | Typical accuracy |
|---|---|---|
| Person comes specifically to ask about this matter | Strong | Highest: the moment and the question are tightly coupled |
| Close family member asks on someone's behalf | Good | High: they carry significant information about the person |
| Casual acquaintance asks out of curiosity | Weak | Lower: thin connection between the moment and the subject |
| Random, vague question with no real stakes | Noise | Unreliable: the chart has no strong signal to lock onto |
| 场景 | 信号 | 典型准确率 |
|---|---|---|
| 当事人专程为此事前来问测 | 强 | 最高：时刻与问题紧密耦合 |
| 直系亲属替当事人来问 | 良好 | 较高：携带当事人大量信息 |
| 泛泛之交出于好奇随口一问 | 弱 | 较低：时刻与当事人关联稀薄 |
| 随机、含糊、无关痛痒的问题 | 噪音 | 不可靠：盘面无强信号可锁定 |
This is why Qimen has always valued the quality of the question. A focused, specific, genuinely urgent question produces better readings than a vague "tell me about my future." The chart is a mirror: if you face it with a clear question, you get a clear reflection.
这就是为什么奇门一直重视提问的质量。一个聚焦、具体、真正紧迫的问题，比一句含糊的"帮我看看未来"能产生好得多的解读。盘面是一面镜子：你带着清晰的问题面对它，就得到清晰的映像。
What AI Changes, and What It Doesn'tAI 改变了什么，又没改变什么
Look at the four limits again. AI directly addresses the two most common ones:
再看一遍那四大局限。AI 直接针对最常见的两个：
| Limit | Human reader | AI reader |
|---|---|---|
| Human error | Fatigue, distraction, skill gaps; varies reader to reader, day to day | Never tired, never distracted. Applies the same rules consistently every time |
| Theoretical disagreements | Each reader follows their own school; you may not know which | Uses a declared, fixed methodology; you know exactly which rules are applied |
| Time-information looseness | Cannot control | Cannot control; same constraint applies |
| System's own limits | Cannot exceed | Cannot exceed; AI reads the chart, it doesn't rewrite the system |
| 局限 | 人类解盘师 | AI 解盘 |
|---|---|---|
| 人为失误 | 疲劳、分心、功力差异，因人而异，因日而异 | 不疲劳、不分心。每次一致地应用同一套规则 |
| 理论分歧 | 各人各师各法门；你可能不知道他用的哪一派 | 使用公开、固定的方法论，你清楚知道应用了哪些规则 |
| 时间信息松散性 | 无法控制 | 无法控制，同样的约束 |
| 体系本身的局限 | 无法超越 | 无法超越，AI 解读盘面，不改写体系 |
In short: AI makes readings more consistent, not more magical. It eliminates the variance that comes from human fatigue and skill gaps. It makes the methodology transparent. But it operates within the same theoretical framework; it cannot exceed the system's inherent ceiling. Anyone who tells you AI makes Qimen 100 % accurate is selling something.
简言之：AI 让解读更一致，而非更玄妙。它消除了人类疲劳和功力差异带来的波动，让方法论透明可查。但它在同一个理论框架内运作，无法超越体系本身的天花板。谁告诉你 AI 能让奇门 100% 准确，那他在卖东西。
The real value of AI in Qimen is not supernatural accuracy; it is accessibility and consistency. A skilled practitioner still sees nuances that AI may miss. But AI makes competent, rule-consistent readings available to anyone, at any hour, without needing to find a master, and it shows its reasoning every time.AI 在奇门中的真正价值不是超自然的准确，而是可及性与一致性。一个功力深厚的师傅仍然能看到 AI 可能遗漏的细微之处。但 AI 让任何人、在任何时间、无需拜师，就能获得合格且规则一致的解读，而且每次都展示推理过程。
How We Validate: QimenIt's Approach我们怎么验证：QimenIt 的方法
Saying "we use AI" is easy. Proving it works is harder. Here is how we hold ourselves accountable:
说"我们用了 AI"很容易。证明它有效才难。以下是我们对自己的问责方式：
1. Chart generation is deterministic. Given the same date, time, and method, QimenIt always produces the same chart. There is no randomness, no hidden variation. You can verify this yourself: cast the same moment twice and compare.
1. 排盘是确定性的。给定相同的日期、时间和方法，QimenIt 永远生成相同的盘。没有随机性，没有隐藏的变量。你可以自己验证：同一时刻起两次盘，对比一下。
2. Yong Shen selection follows published rules. The AI picks the Yong Shen based on the question category using the classical systems we document openly. It doesn't invent its own logic.
2. 用神选取遵循公开规则。AI 根据问题类别、按照我们公开记录的经典体系选取用神。它不会自己发明逻辑。
3. Reasoning is shown, not hidden. Every reading explains which palaces were examined, which symbols were read, and why. You see the chain of logic, not just a conclusion. If you know Qimen, you can audit every step.
3. 推理过程展示而非隐藏。每次解读都说明查看了哪些宫位、读取了哪些符号、为什么。你看到的是逻辑链条，不仅仅是结论。如果你懂奇门，可以逐步审查。
4. We benchmark against practitioner consensus. We regularly test the AI's readings against interpretations from experienced human practitioners. Where the AI diverges from practitioner consensus, we investigate: sometimes the AI missed a nuance; sometimes it caught something the practitioners overlooked. Both outcomes improve the system.
4. 我们以实践者共识为基准。我们定期将 AI 的解读与经验丰富的人类实践者的解读进行比对。当 AI 偏离实践者共识时，我们会排查原因：有时是 AI 遗漏了细微之处；有时是它捕捉到了实践者忽略的信息。两种结果都能改进系统。
5. We track user feedback. When users tell us a reading resonated, or didn't, that signal feeds back into our evaluation. We don't claim every reading is right. We claim every reading is improvable.
5. 我们追踪用户反馈。当用户告诉我们某次解读击中了要害（或者完全不对），这个信号会回到我们的评估体系中。我们不宣称每次都对，我们宣称每次都可以更好。
Our Promise: Honest Over Impressive我们的承诺：诚实大于精彩
We would rather give you an honest 75 % than a dishonest 100 %. That means:
我们宁可给你诚实的 75%，也不给你虚假的 100%。这意味着：
- We will never claim Qimen prediction is infallible, because it isn't, and saying so would be dishonest.
- We will always show our reasoning so you can judge for yourself.
- We will tell you when a question is too vague to read well, rather than pretending to answer it.
- We will keep improving. The system learns from every reading, every correction, every piece of feedback.
- 我们永远不会宣称奇门预测百发百中，因为它做不到，这么说是不诚实的。
- 我们会始终展示推理过程，让你自己判断。
- 当问题太模糊无法好好解读时，我们会直接告诉你，而不是硬编一个答案。
- 我们会持续改进。系统从每次解读、每次纠正、每条反馈中学习。
Cast a free chart and see for yourself: not just the answer, but the reasoning behind it. Then decide whether the reading makes sense for your situation. That's all we ask.免费起一张盘，亲自看看：不只是答案，还有背后的推理。然后判断这次解读对你的情况是否有意义。我们只要求这一点。
Start your free reading →免费起一张盘 →