Chance of predicting exact basketball scores?

[kbfc]

[ QUOTE ]
6481 NBA games since 1990 with non-zero integer spread and total.

There were 254 pushes against spread [3.919%].

There were 100 pushes against total [1.543%].

Assuming these are independent (questionable), the likelihood of both occuring is 0.060%, slightly higher than the actual historical probability of 0.046%.

[/ QUOTE ]

Excuse me, as this is rushed, and I haven't really thought about it, but I think your original methodology is slightly flawed.

You should be looking for games that fit one of the following categories:
1) Odd total (ie: 189) AND odd spread (ie: -3).
2) Even total (ie: 190) AND even spread (ie: -2).

Otherwise, just take all of the non-zero spread games and round to the nearest integer result. As you did it, there are a whole class of games (even total/odd spread + odd total/even spread) that have no chance of push/push. This is probably why your estimation came out higher. The true historical rate is actually higher than you calculated it.

I have the last 2 seasons in a database. I don't have all the spread data for that time period in there yet, but I can just go off of my predicted scores, which will be close enough for these purposes. I'll check how often it hit exactly.

[ImBen]

Good point. I'll take another look...

[ImBen]

3224 games in which push/push was possible, so 0.093%, not 0.046%.