Trading Concept
Sample Size
The number of independent observations needed to distinguish a real pattern from random noise — small samples are unreliable, and most traders draw conclusions too early.
A coin flipped 10 times that comes up heads 7 times does not prove the coin is biased. The probability of 7 or more heads in 10 fair flips is about 17% — it happens all the time. A strategy that wins 7 of its first 10 trades has not proven it has edge. And a strategy that loses 7 of its first 10 trades has not proven it lacks edge.
The mathematics of sample size are unforgiving. To detect a 55% win rate with 95% confidence (distinguishing it from a fair coin), approximately 400 trades are required. To detect a 52% win rate — which can be highly profitable with proper sizing — requires over 2,500 trades. Most traders evaluate their strategies after 20–50 trades. At that sample size, noise dominates signal.
This is the most practical concept in the entire library. When the bot has been running for a week and has 15 trades, the user does not have enough data to evaluate performance. They have an anecdote. The ledger records every trade, building toward a sample that eventually becomes meaningful. The discipline is in waiting for the sample to grow before drawing conclusions — and understanding that even 100 trades may not be enough to separate skill from luck.