Why does a run of wins feel different from a run of losses of the same length?
Why this matters. Both runs distort, and they distort in opposite directions. Reading either as information about your skill is the error, and it is much easier to spot in someone else.
Four in a row your way, and you feel you have understood something. Four against, and you feel you have lost whatever you had. The lengths are identical.
Why the two feel different
A run of gains is experienced as confirmation — evidence that a method works. A run of losses is experienced as threat — evidence that something must change immediately. So the same sequence length produces expansion in one direction and contraction in the other, and neither response was decided.
The question streaks cannot answer
Whether you are skilled. And this is the point where this domain earns its existence, because the reason is structural rather than psychological.
Conceptual frame, and it is the spine of this domain: markets are a wicked learning environment in Hogarth's sense — feedback is noisy, delayed and confounded, so experience does not convert into skill on its own. Kahneman and Klein (2009) set the condition explicitly: intuition can be trusted where the environment is regular enough to hold patterns and feedback is prompt enough to learn from. Where those fail, subjective confidence is not evidence of skill.
A four-length run is not a sample. It is not a small sample either — it is not a sample of anything, because you would need to know how many similar runs occur by chance in a series like yours, and that number, unlike the feeling, can be worked out in advance.
The number exists. It comes from arithmetic, and it belongs in the file before the run. A process with a 50% hit rate produces four losses in a row from one starting point in sixteen; across a hundred decisions the longest losing run to expect is about six. Technical Analysis does the sums in How can a system that wins 40% of the time make money?. Prices are no streakier. On the S&P 500 fund SPY, 5,033 sessions from 2006-09-01 to 2026-09-04 (measured 2026-09-07), the longest run of down days was eight; a coin with the same 45% down-rate gives ten. Write the expected longest run beside the process. A streak inside it is not information.
The artefact
A decision-quality tally kept independently of outcomes: for each entry, did it meet its criteria at the time — yes or no. Kept separately from whether it made money.
When you next feel a streak means something, the tally is what you consult. If decision quality was constant while outcomes clustered, the streak is telling you about variance, not about you.
Try it now
Take your last eight closed positions. Mark each "criteria met at entry: yes/no", ignoring outcomes entirely. Then compare that column with the outcomes. Any clustering that survives is worth examining; the rest is noise wearing a pattern.