Every scouting department has had this conversation: a name comes up, someone pulls up his numbers, and they’re eye-catching: 19 points, great efficiency, strong on-off numbers. Then someone asks the question that matters more than any of those figures: “how many games is that over?”
Small samples lie. Not because the player is cheating the numbers, but because basketball is noisy at the individual-game level. A player can catch fire for six games against a soft stretch of schedule, or a role change from an injury elsewhere on the roster can temporarily inflate his usage. Neither tells you what he’ll do over a full season, and neither tells you what he’ll do for you.
There’s no single magic number of games that makes a stat line trustworthy; it depends on minutes, on how spiky the stretch was, and on whether the sample includes real variation in opponent quality. But a useful habit is to ask two questions before acting on any stat line: how many minutes has he actually played, and does the number hold up if you split the sample in half? If a player’s per-36 production in his first half of games looks nothing like his second half, you’re not looking at a skill level yet; you’re looking at variance.
This is exactly why MelonIQ applies minimum-minutes and minimum-games floors before showing a projection or including a player in cross-league comparisons; a hot two-week stretch shouldn’t be enough to move a player onto (or off) a shortlist. The floor doesn’t tell you a player is good or bad. It just tells you whether you’re looking at a signal yet, or still looking at noise.
The players worth chasing are the ones whose numbers hold up under that scrutiny, not the ones who happened to have a good month at the right time.