One Night Only: Offensive Packers-Redskins Game Review, Mythbusters Edition

[GTripp0012]

Originally Posted by warriorzpath

My question is what is a good enough "sample size"? The point in that is- since it's a statistical term, there must be statistical way to figure that out, right? And if not, why not? That's partly why I don't get that term being used in football.

It all depends on what you're looking at of course, but if you want a blanket number where you have a sample that's not TOO small, I'd say 30 is pretty good.

30 games? Sure, you can start to use W/L to evaluate a team at that point. 30 passes? You can start to make conclusions off how a guy looks throwing a football.

There is a statistical way to figure it out. It's called a significance test. Basically, a significance tests only goal is to determine how likely that a result is due to random chance. If the sample is adequately large, something can be deemed significant. If not, it will tell you that something is not significant to 'X' degree of certainty (typically 90% or 95%).

There is no sample size in which something can be 100% certain, but most people are okay with something that is 99.9% certain.

It's a really math-y answer, but I don't have a non math-y way to describe a significance test.