Kewei Hou, Chen Xue and Lu Zhang assembled 452 published asset pricing anomalies and attempted to replicate each one under a single consistent method.
Under that treatment 65 percent could not clear the conventional hurdle of an absolute t-statistic of 1.96. In the trading frictions category the failure rate was 96 percent. Applying the higher multiple-testing hurdle of 2.78 raised the overall failure rate to 82.1 percent, leaving roughly 81 of the original 452.
Now the finding, and it is about you rather than about the field.
These were peer-reviewed findings by professional researchers with better data and stronger incentives for care than any individual has.
About four in five did not survive an independent replication.
Whatever you already believe about this field arrived through a chain that includes several such findings, repeated by people who did not check them.
The diagnostic is not testing whether you have been paying attention. It is testing which of the things you absorbed are load bearing.
This is the first thing you do here, and it takes about thirty minutes. It exists because per the course autopsy your own estimate of what you know is not a reliable input to a learning plan.
Six domains, measured separately
Because the normal shape is uneven. Somebody who has held crypto for five years may know a great deal about custody and nothing about how a market clears. Somebody from a trading background may have the reverse.
A single score hides that. Per part one below, the placement that follows is per domain, so the path skips what you already have and does not skip what you do not.
And the questions adapt. The diagnostic stops in each domain when the estimate is precise enough to place you, which is why it is shorter for some people than others.
Part one: the survival rate.
Of 452 published anomalies, 82.1 percent failed at the stricter hurdle:
452 x (1 - 0.821) = about 81 survived
Part two: what that implies about received knowledge.
Those 452 were the findings that reached publication. Per part one about 81 held up.
Everything you have heard about this field passed through people who were not checking, so the proportion that is load bearing is unlikely to be better than the proportion that was load bearing at the source.
Part three: which is not a reason to distrust everything.
Per part one, 81 findings did survive. The exercise is separating them, and per the course autopsy the sensation of confidence does not do that.
So the diagnostic measures instead. Per part two it is faster than arguing about it and per O100-04 the calibration curve tells you how much weight your own confidence deserves.
The one instruction for this lesson.
If the result places you lower than you expected in a domain, take the lower path. Per the course autopsy the accuracy of a strong feeling of knowing is often below 80 percent, and per O100-04 you will get a personal number for this within the hour.
If it places you higher, take the higher one. Per part three the diagnostic is not designed to flatter or to humble, and skipping material you already hold wastes the only resource this course is trying to protect, which is your attention.
And redo it later. Per O100-04's curve, the diagnostic is re-measured at every gate, so the placement is a starting point rather than a verdict.
I have been in this space for years, so I can skip the basics.
Time in the space and coverage of the material are different quantities, and per part two the second is what the placement measures.
What is true. Experience is real, and per part three the diagnostic will place you higher wherever you have it. Nobody is asked to repeat material they hold.
What years do not guarantee. Per the autopsy, exposure to a field means exposure to what circulates in it, and about four in five published findings in an adjacent field did not survive replication. Per part two what circulates is not filtered.
And the gaps are usually specific. Per the six domains section, long experience in one domain frequently coexists with none in another, which is the shape a single self-assessment cannot report.
So per P6 the accurate framing: years give you domains, not coverage, and the diagnostic finds which. Per the callout, accept the result in both directions.
The diagnostic measures competency across six domains separately, because the normal shape is uneven and a single score hides it, and it adapts until the estimate is precise enough to place you. It exists because your own estimate is not a reliable input: per the course autopsy, accuracy stops tracking confidence above about three to one, and at a stated 98 or 99 percent it is often below 80. And the material you have absorbed came through a filter that does not work: of 452 published asset pricing anomalies replicated under one consistent method, 82.1 percent failed at the multiple-testing hurdle, leaving about 81. That is not a reason to distrust everything, because 81 did survive. It is the reason placement is measured rather than asked. Accept the result in both directions, and expect to be re-measured at every gate.