What do your own records say about you that you would never guess?

Why this matters. This is the payoff of the entire domain. Everything before it was collecting the inputs.

Once a year of records exists, questions become answerable that no amount of reflection could touch. Not "am I disciplined?" — questions with numbers.

What becomes visible

How often your written invalidation condition was actually met before you exited. Whether decisions marked rushed differ from the rest. Whether your high conviction entries outperformed your low conviction ones. How often you moved a target and whether it helped. Whether your late-session decisions differ from your morning ones — the measurement Course 1 left you with when it removed the two discredited studies behind decision fatigue.

Every one is a question about you, answerable from your own file, and unanswerable any other way.

One more proportion, the only one that grades a number you wrote. Group entries by the probability stated at entry. For each group, count the fraction in which the case held: the position ended without the written condition being met. If the entries you called 70% held about 70% of the time, the number carries information; if they held 45% of the time, it does not, and conviction does not change that. This is calibration, and it is the one feedback a market cannot corrupt, because you set the question and the answer is yes or no.

Why the input has to be contemporaneous

Because a record written after the outcome is a memory, and memory is reconstructed around what happened. This is why the attribution asymmetry Course 1 described survives ordinary honest review: the outcome supplies the material for the explanation. Only notes written before the outcome was known escape that.

What this is not

Not performance measurement. Your returns are a different question, and they are dominated by variance at any sample size you will personally reach. These are base rates about your decisions, not about your results — and that is what makes them stable enough to learn from.

The artefact

A base-rate sheet, computed once or twice a year from the records: five or six proportions, each with the count it was computed from.

The count matters as much as the proportion. Three out of four is not a finding.

Nor is fifteen out of twenty, quite. At thirty decisions, the honest band around a proportion is still about twenty points either side, so "I exit early 60% of the time" and "40% of the time" are barely distinguishable. The sheet becomes useful when a proportion sits far from where you assumed it did — that is a finding at any count — and when the same proportion holds up two years running.

Try it now

Pick one question you would like answered about yourself and write down what you would need to have recorded to answer it. If you are not recording it, that is today's change.