Live win probability for the NCAA basketball championship games is available now at wp.advancednflstats.com/bball. (Final games are here. Games from the previous day are here.)
The model's approach is very similar to my football model. Basketball is a much simpler sport, though. There's no field position, down, or distance in basketball. Score and time remaining are really the only significant factors.
Live NCAA Basketball Win Probability
published on 4/05/2010
Subscribe to:
Post Comments (Atom)
Is that really true? Most teams have a standard configuration of one center, two forwards, and two guards in at almost all times, but there are variations on that, especially in the tricky world of NCAA. And not all players are equal, in terms of field goal percentage, 3-point percentage, and so on. Seems to me you could break down basketball to be nearly as complex as football if you looked deeply into it. Or are you saying that you've done that and found that indeed, for purposes of who is going to win a game in play right now it doesn't much matter who's handling the ball?
"Score and time remaining are really the only significant factors."
Possession is super important late in the game, but I guess that's pretty obvious.
I'd think that late in the game, the state of the shot clock is pretty important as well. The state of the ballhandler's dribble could possibly be important as well.
Possession is obviously critical. In my NBA version, there is a "PV" (possession value) stat that indicates the value of having the ball in terms of WP at each point of the game. It's a good indication of "leverage."
I would wager that total team fouls and number of fouls committed by starting players could contribute a higher level of predictability for a win probability model...
This is the first time I've seen your basketball graphs, but at first glance I am surprised to see so much movement early in the games. You often hear people complain about how the first 3/4 of a basketball game don't matter, but if that were the case you would expect WP to hover closer to 50% throughout the 1st half. Is this complaint unfounded, or are the graphs from this weekend atypical?
@brett
http://www.advancednflstats.com/2009/03/ncaa-basketball-win-probability.html
New Mexico were playing carrying its injured star, Darington Hobson, who suffered the injury during their last victorious game against Montana.
Is there any way to incorporate rankings instead of assuming two teams are equal? Say, for instance, Ken Pomeroy's ratings.
@eric:
http://kenpom.com/blog/index.php/weblog/in-game_win_probabilities/
Ken Pomeroy has been working on just that.