CSNbbs

Final score: UAB 63 - UNC 59

KenPom.com provides graphs, of all completed games, showing the probability of winning at each possession throughout the games. This service is for subscribers only. So, I'm not going to show the graph of the UNC@UAB game. However, let me try pulling some of the data out of the graph and presenting it here. I'm not going to make any attempt at explaining how Ken Pomeroy comes up with his probabilities. As a reminder, Pomeroy used to be a computer programmer for the National Weather Service, and decided a few years ago that he could make a living on his college basketball ratings.

I'll list the time remaining in each half - who is favored to win and their probability of winning - the score - and the leverage. Leverage gives you an idea of how important each possession is to the probability of winning. Leverage runs from low to medium, to high. Low leverage means possessions at that point are not that important to the probability of winning. High leverage means each possession could swing the probability of winning 10% or more. In other words, low leverage implies a relatively boring game (my words), while high leverage implies an intense and exciting game.

FIRST HALF
20:00 - UNC 76.6%
15:00 - UNC 65% (UAB 12 - UNC 6) - medium
10:00 - UAB 53% (UAB 24 - UNC 14) - high
05:00 - Toss up (UAB 26 - UNC 18) - high
00:00 - UAB 75% (UAB 37 - UNC 25) - medium

SECOND HALF
15:00 - UAB 65% (UAB 39 - UNC 32) - high
10:00 - UAB 57% (UAB 49 - UNC 44) - high
05:00 - UAB 75% (UAB 52 - UNC 46) - high
04:00 - UAB 74% (UAB 52 - UNC 46) - high
03:00 - UAB 80% (UAB 54 - UNC 50) - high
02:00 - UAB 77% (UAB 57 - UNC 54) - high
01:00 - UAB 92% (UAB 58 - UNC 54) - high

Once the clock got down to 45 seconds or so, the leverage finally dropped to low-medium. For the majority of the second half, the game was very intense as indicated by the high leverage. Each possession was critical to the probability of winning.

UNC's largest lead was 4-2, with 18:45 left in the first half. That reduced UAB's probability of winning to 19.5%. UAB's largest lead was 33-18, with 2:25 left in the first half.

Notice how UAB's probability of winning actually dropped from the 5 minute mark, in the second half, to the 4 minute mark even though UAB's lead remained 6 points. On the surface that may not make sense, but remember the probability of winning is dependent on which team possesses the ball. The leverage was high, so there were some large swings in the probability during those last five minutes.

Why put these probability graphs online?

From the horse's mouth.

Quote:My goal is to have an easy reference for the evolution of a game in a way that goes beyond the final score to truly characterize the competitiveness of the contest.

Pomeroy started these probability graphs during the 2010 season, and yet I'm just now paying attention to them.

I could share the probabilities from the New Mexico game, but it's painful.

(12-05-2013 01:57 PM)KevMo4UAB Wrote: [ -> ]Final score: UAB 63 - UNC 59

KenPom.com provides graphs, of all completed games, showing the probability of winning at each possession throughout the games. This service is for subscribers only. So, I'm not going to show the graph of the UNC@UAB game. However, let me try pulling some of the data out of the graph and presenting it here. I'm not going to make any attempt at explaining how Ken Pomeroy comes up with his probabilities. As a reminder, Pomeroy used to be a computer programmer for the National Weather Service, and decided a few years ago that he could make a living on his college basketball ratings.

I'll list the time remaining in each half - who is favored to win and their probability of winning - the score - and the leverage. Leverage gives you an idea of how important each possession is to the probability of winning. Leverage runs from low to medium, to high. Low leverage means possessions at that point are not that important to the probability of winning. High leverage means each possession could swing the probability of winning 10% or more. In other words, low leverage implies a relatively boring game (my words), while high leverage implies an intense and exciting game.

FIRST HALF
20:00 - UNC 76.6%
15:00 - UNC 65% (UAB 12 - UNC 6) - medium
10:00 - UAB 53% (UAB 24 - UNC 14) - high
05:00 - Toss up (UAB 26 - UNC 18) - high
00:00 - UAB 75% (UAB 37 - UNC 25) - medium

SECOND HALF
15:00 - UAB 65% (UAB 39 - UNC 32) - high
10:00 - UAB 57% (UAB 49 - UNC 44) - high
05:00 - UAB 75% (UAB 52 - UNC 46) - high
04:00 - UAB 74% (UAB 52 - UNC 46) - high
03:00 - UAB 80% (UAB 54 - UNC 50) - high
02:00 - UAB 77% (UAB 57 - UNC 54) - high
01:00 - UAB 92% (UAB 58 - UNC 54) - high

Once the clock got down to 45 seconds or so, the leverage finally dropped to low-medium. For the majority of the second half, the game was very intense as indicated by the high leverage. Each possession was critical to the probability of winning.

UNC's largest lead was 4-2, with 18:45 left in the first half. That reduced UAB's probability of winning to 19.5%. UAB's largest lead was 33-18, with 2:25 left in the first half.

A statistical model that gave us a 92% chance of winning up 4 with a minute to go apparently didn't input all the profanity I used during the New Mexico game. Twice. About fifteen minutes apart. Let's just say my internal confidence statistics when we were up 4 with a minute left to UNC did not approach 92%.