- Apr 12 4:55 PMI finally did what we talked about a long time ago here. I looked at
how well we can predict the pace of a single game based upon the two
teams' average paces. I used a formula like Stratomatic uses:
GamePace = TmA_avgpacediff + TmB_avgpacediff
So if TmA is 4 possessions per game faster than normal and TmB is 1
possession per game slower than league average, then they
are "expected" to play at a pace 3 possessions faster than average.
Over the last few years, this has an average error of -0.07 poss per
game. It actually slightly underpredicts pace, which surprised me,
but not by any large amount. The mean absolute error was 3.1, as
opposed to 3.8 using just the league average as a predictor. It
predicted about 71% of games correctly with regard to whether they
would be faster or slower than average.
This implies to me that a team's average pace has a pretty small
impact on the actual game pace. Reducing 3.8 to 3.1 is only about
20%. There is a fair amount of "other" factors that may account for
pace changes. I'm not sure how much we can identify in those other
Time to brainstorm: What independent factors are there that can be
used to a priori predict the pace of a game? Playoffs used to be a
significant one -- the pace used to get a lot slower in the playoffs.
I haven't looked in a while. If that's the case, competitiveness
(how close the teams are and how good they are) would be another
factor (correlated to playoffs). Not sure what else.
Note: I did correct for overtime games.
- Next post in topic >>