The brain keeps no complete copy of an event. Recalling something means assembling a fresh account from a small number of cues, and the material used in that assembly includes everything known at the present moment. Where an outcome has already been revealed, the outcome participates in the reconstruction and becomes part of what the original expectation is remembered to have been. Subjectively this feels identical to simply remembering, and there is no internal marker distinguishing a reconstructed detail from a preserved one.
In markets the mechanism produces a very consistent pattern. After an event, most participants find they had a sense of it coming. That sense carries the full texture of genuine recollection, without the thinner quality an inference would have. They are neither lying nor exaggerating. What they are unable to do is restore the state of uncertainty they occupied before the outcome was known. Once a gap has been filled with an answer, it cannot be emptied again on request, and the answer now sits in the space where the doubt used to be.
The first consequence is that learning gets blocked. Learning from experience requires comparing a prior expectation against a subsequent outcome. Where the expectation has already been revised by the outcome, the comparison amounts to checking a document against itself. Whatever the underlying accuracy, the comparison returns broad agreement and therefore generates no correction. Many people spend years in markets without visible improvement despite paying close attention throughout, and this is one of the structural reasons why attention alone does not produce progress.
The second consequence is that self-assessment drifts upward. Each event leaves behind an impression of having been broadly right, and those impressions accumulate without anything arriving to offset them. After several years, somebody's estimate of their own judgement may have separated substantially from their actual record, with no internal mechanism available to raise an alarm. Every piece of internally available evidence points the same way. That is precisely what one would expect if the evidence were being generated by the thing it is meant to test.
The only effective remedy is to externalise expectations. Written material does not get reconstructed. It fixes at the moment of writing and can be retrieved later exactly as it was. That includes the parts which now look foolish and would otherwise have been quietly smoothed away. That is the essential difference between a written record and a memory, and it is the one function of writing that nothing else performs. Everything else writing does could in principle be done some other way.
What gets recorded matters more than whether anything is recorded. Noting direction alone has limited use, since there are only two directions and the base rate for guessing is already high enough to obscure any real skill. What helps considerably more is recording a level of confidence together with the specific reasoning behind it. Six months later, the question worth examining is not whether the direction was right. It is whether the high-confidence calls actually outperformed the low-confidence ones, which is a question about calibration.
Most people skip this, usually explaining that it seems like a lot of effort. The writing is not the effortful part, and a usable entry takes perhaps two minutes. The effortful part arrives later, when the old record has to be read without the benefit of any softening. An unedited record from twelve months ago carries far more information than any act of recall. Its capacity to be uncomfortable scales roughly in proportion to how much it carries. That relationship is worth expecting in advance, since being surprised by it is what stops most people continuing.