Flip a fair coin ten times and seven heads is unremarkable. Watch a player for ten games and seven overs feels like destiny. The mathematics is identical; only the storytelling changes. This article covers how much evidence a sample really carries, why some stats settle down quickly while others take a season, and what to do when the data you have is all the data there is.
Football hands out the fewest opportunities of any major sport, then asks you to project them anyway. The research that survives an NFL season starts from role and volume, respects game script, and treats seventeen games as the tiny sample it is.
Every prop tool shows you a trailing hit rate, and almost every reader treats it as a forecast. It is not one. This article separates the two numbers that dominate prop research, explains why eight of the last ten overs is weaker evidence than it feels, and shows where a hit rate genuinely earns its keep.
Two researchers can look at the same prop, run sound processes, and land on 54 percent and 57 percent. Neither is wrong yet. Learning to think in that language, small percentages, long runs, honest error bars, is the difference between research and a highlight reel with numbers on it.
Open any prop research screen and you may see three numbers describing the same event. One came from a model. One is a product a platform wants to sell you. One is a distillation of what the sharpest markets believe. Confusing them is the most common structural error in prop research.
Ask a projection what a player will do tonight and the honest answer is not a number. It is a shape: a range of outcomes with probabilities attached. This article walks through how that shape gets built, layer by layer, and why the shape matters more than the average sitting in the middle of it.