Collecting 300 rows is not the same as observing 300 independent market events.
One shock creates multiple rows
If decisions occur at the same moment across BTC, ETH, SOL, and several altcoins, it is easy to count each row as an independent sample. In reality, one market-wide shock may be driving many of them at once.
The horizons also overlap
When 1-second, 3-second, 5-second, and 10-second outcomes overlap, and decisions are repeated close together, adjacent samples can be strongly dependent. Treating the row count as n can therefore understate uncertainty.
Three hundred fully matched decisions is the maturity gate for beginning effect review, not a number that automatically guarantees statistical sufficiency.
What should be measured separately?
The nominal sampling unit may be the decision, but simultaneous multi-symbol decisions, repeated signals, overlapping horizons, and shared BTC shocks should be tracked separately. Time-block inference rules should be defined in advance rather than chosen after seeing the effect size.