Tag:rankings
Posted on: January 11, 2010 12:58 am
  •  
 

My college basketball rating system, if you care!

My scoring system is set-up as follows:

The first component is margin of victory/defeat and this is how I have it set-up:

For Wins there are 5 groupings of margin (1-5pts), (6-10), (11-20), (21-30), and (over 30):

for the 5 groupings, these points are awarded:
home:    10, 12, 14, 16, 17
neutral:  11, 14, 17, 19, 20
away:     13, 16, 19, 22, 24

So examples to be sure the above is understood...If a team wins at home by 1-5 points (grouping #1), they get a margin point total of 10. If a team wins on the road by more than 30 points, they would get a margin score of 24 points.

For Losses, there are only 4 groupings of margins (1-5 pts), (6-10), (11-20). and (over 20 pts)

home:    0, -2, -4, -6
neutral:  1, 0, -2, -4
away:    3, 1, 0, -2

So, if you lose a game by 11-20 pts, on a neutral court, you would get a margin score of -2. If you lose a true away game, by 5 or less, then you would get a margin score of 3 points.

Thats the summary of the margin component of my system:

The second part of my scoring system is the opponent's rating component:

Again, it is seperated into a Win and a Loss subcomponents, and each of those has different ratings levels:

For both wins and losses the buckets, or rating ranges are:

group 1:(347-201), group 2: (200-151), group 3 (150-101), group 4 (100-51), group 5: (50-26), group 6: (25-11), group 7 (10-6), and group 8 is (5-1)

For winning, the ratings points you receive are:

gr#1 (1 pt), gr#2 (2 pts), gr#3 (4), gr#4 (6), gr#5 (12), gr#6 (18), gr#7(24), gr#8 (30 pts).

For losing, the ratings points by group are:

gr#1 (-8), gr#2 (-4), gr#3 (0), gr#4 (2), gr#5 (6), gr#6 (11), gr#7 (12), gr#8 (14)

So, if you win against the  #29 rated team (group #5), you would get 12 points for the rating score component.

If you lost to a team in the top 5 (gr#8), you would get 14 points for the rating score component.

The two scores would be combined to give you a total score for each game played.

So, if you won on a neutral court by 16 points, over a team in the top 10 (gr#7: 6-10), then you would get 14 points for the margin score and 12 points for the rating component, for a total of 28 points for that game...

I worked with the components - group scores, until I got comparable values for a loss to a top team being equal to or a higher score than a minimum win at home. So that when conference play starts and good teams get beat by great teams, they do not become disadvantaged to strong teams in weak conferences beating up on cupcakes...

Best possible score: Winning on the other team's home court by over 30 (24 pts) plus a win over a top 5 team rating score of (30 pts): For a maximum total single game score of 54 (Highly unlikely)...

The lowest score for winning a game would be: Winning at home, by 5 or fewer points, against a team rated worse than 200...that would be worth 10+1=11 points.

The Lowest possible score would be: For losing to a team that is worse than 200, on your home court, by more than 20 points. The score for that would be: (-6) margin score plus (-8) rating score, total of -14 points

Then to take into account the fact that teams play a different number of games in any given week, or at any point in the season, I divide the total score by the # of games played to put the score on a per game basis.

IMPORTANT:
My understanding is that the computer models that Sagarin and KenPom use are very sophisticated math models, but they encompass one or both of two components, those components are often genericly referred to as, Elo Chess, and Pure Score...
Elo Chess is only interested in calculating the probability of winning/losing...
Pure Score is interested in the margin of winning...

So, I have very loosely incorporated both of these calculating methods into my scoring system, but at a very simplified level, and with constraints...I cap the margin of victory, and treat all wins over 30 points, and all losses over 20 points as equivalent. So that running a score up to infinity does a team no good, but if a team can not consistently win games by over 30, they are not maxxing their scoring potential in my system. Some teams just aren't talented enough to beat even average teams by 30 points. This to me is a sign of a team that might be good, but not great.

Can you follow the scoring? Any questions?

 

Category: NCAAB
Tags: rankings
 
 
 
 
 
The views expressed in this blog are solely those of the author and do not reflect the views of CBS Sports or CBSSports.com