Big Ten rankings

[PurpleWhiteBoy]

I ran a Sagarin-style evaluation of the Big Ten teams using only conference games to determine the rankings below...
The difference in the ratings of two teams tells you expected margin of victory on a neutral court.
Home court is usually assigned an advantage of about 3 points.

Team	Won - Loss	Rating
Purdue	13-4	106.89
Michigan	9-7	103.54
Northwestern	11-5	103.46
Indiana	10-7	103.00
Maryland	10-7	102.80
Michigan State	9-7	102.58
Illinois	9-7	101.77
Iowa	9-8	101.19
Rutgers	9-7	101.13
Penn State	7-9	97.85
Wisconsin	8-9	97.77
Ohio State	3-13	94.52
Nebraska	7-10	94.06
Minnesota	1-15	88.83

 

[timbfischer]

Cool. How did you do it?

 

[Hungry Jack]

Passes the sniff test

 

[PURPLECAT88]

This matches the eye test pretty well. As for home court adjustments, it seems as if for Iowa and Maryland you should add ten for home and subtract ten for road. Iowa in particular just can't shoot at all on the road. It's as if Brian Ferentz takes over the offense every time they leave Carver-Hawkeye.

 

[CappyNU]

Per Torvik, home/road splits in conference are as follows, from most extreme in favor of home to most extreme in favor of road:

1) Maryland (9-0, 1-7) - 2nd at home, 12th on road, dead last in road offensive efficiency, can we be the first to knock them off at home?
2) Nebraska (5-3, 2-7) - 11th at home, 14th on road, dead last in road defensive efficiency

3) PSU (5-3, 2-6) - 6th at home, 10th on road, glad we get them at home
4) Iowa (7-1, 2-7) - 7th at home, 11th on road, best offense by far, worst defense by far at home, mediocre offense, bad D on road
5) Indiana (7-1, 3-6) - 3rd at home, 8th on road, offense much worse on road
6) Michigan (7-2, 2-5) - 4th at home, 6th on road, should have better record but couldn't close out some road games. Disappointing we had to deal with Covid/dead legs for the home game

7) MSU (6-2, 3-5) - 5th at home, 4th on road, defense elite at home, good on road
8) Purdue (7-1, 6-3) - 1st at home, 2nd on road, top defense on road but better offense at home
9) Rutgers (6-2, 3-5) - 8th at home, 5th on road, injuries are messing with the numbers a bit, but they have best defense at home
10) Wisconsin (5-4, 3-5) - 12th at home, 7th on road

11) OSU (2-5, 1-8) - 13th at home, 9th on road, they're just bad. Zed Key injury in close loss to Purdue just derailed their season immediately after beating us
12) Illinois (6-2, 3-5) - 10th at home, 3rd on road, for whatever reason their efficiency numbers are better on the road than at home, but they win more at home
13) NU (6-3, 5-2) - 9th at home, 1st on road, defense is about the same either way, but we have the...top offense in the conference on the road(?)
14) Minny (0-8, 1-7) - 14th at home, 13th on road, woof.

 

[PurpleWhiteBoy]

Cool. How did you do it?

I found the scores for all the Big Ten games in a nice format.
I downloaded those into an Excel spreadsheet.
I used a macro that I had sitting around from a few years back.

So you start with all the teams ranked equally at 100. (or 80 or whatever). Then you compare the first game "expected outcome" with the actual outcome. Lets say it was Minnesota at Purdue. Both teams are rated 100. 3 point home court advantage. So Purdue should win by 3. Purdue actually wins by 20. Thats one data point that says Purdue is 17 points better than Minnesota. That game's "error" is +17 points. So you look at each game, totalling the "errors" for each team. After looping thru all the games, you use the total error (per team) to adjust each team's actual ranking. So if Purdue's total error over 15 games was +45 points, you'd raise their rating by 3 points.

Then you run it again with the new ratings for each team. And again. And so on.

Eventually the ratings stop changing because the total error has been minimized.

This is a brute force approach, but it gets the job done.

 

[eastbaycat99]

I found the scores for all the Big Ten games in a nice format.
I downloaded those into an Excel spreadsheet.
I used a macro that I had sitting around from a few years back.

So you start with all the teams ranked equally at 100. (or 80 or whatever). Then you compare the first game "expected outcome" with the actual outcome. Lets say it was Minnesota at Purdue. Both teams are rated 100. 3 point home court advantage. So Purdue should win by 3. Purdue actually wins by 20. Thats one data point that says Purdue is 17 points better than Minnesota. That game's "error" is +17 points. So you look at each game, totalling the "errors" for each team. After looping thru all the games, you use the total error (per team) to adjust each team's actual ranking. So if Purdue's total error over 15 games was +45 points, you'd raise their rating by 3 points.

Then you run it again with the new ratings for each team. And again. And so on.

Eventually the ratings stop changing because the total error has been minimized.

This is a brute force approach, but it gets the job done.

Had someone tried to run your macro when I was in school, the Vogelback CDC6400 would have melted, I think!

 

[timbfischer]

I found the scores for all the Big Ten games in a nice format.
I downloaded those into an Excel spreadsheet.
I used a macro that I had sitting around from a few years back.

So you start with all the teams ranked equally at 100. (or 80 or whatever). Then you compare the first game "expected outcome" with the actual outcome. Lets say it was Minnesota at Purdue. Both teams are rated 100. 3 point home court advantage. So Purdue should win by 3. Purdue actually wins by 20. Thats one data point that says Purdue is 17 points better than Minnesota. That game's "error" is +17 points. So you look at each game, totalling the "errors" for each team. After looping thru all the games, you use the total error (per team) to adjust each team's actual ranking. So if Purdue's total error over 15 games was +45 points, you'd raise their rating by 3 points.

Then you run it again with the new ratings for each team. And again. And so on.

Eventually the ratings stop changing because the total error has been minimized.

This is a brute force approach, but it gets the job done.

Would you be willing to share the macro with me?

 

[PurpleWhiteBoy]

Would you be willing to share the macro with me?

Its an excel macro written in vba or whatever, using cells in a spreadsheet.
So its pretty specific to how the data is laid out.
I'd rather just send you the logic and you can implement it in whatever way you see fit.
Its not complicated.

 

[timbfischer]

Its an excel macro written in vba or whatever, using cells in a spreadsheet.
So its pretty specific to how the data is laid out.
I'd rather just send you the logic and you can implement it in whatever way you see fit.
Its not complicated.

Sure. tim.b.fischer@gmail.com

 

[timbfischer]

Thanks, dude!

 

[Hungry Jack]

I found the scores for all the Big Ten games in a nice format.
I downloaded those into an Excel spreadsheet.
I used a macro that I had sitting around from a few years back.

So you start with all the teams ranked equally at 100. (or 80 or whatever). Then you compare the first game "expected outcome" with the actual outcome. Lets say it was Minnesota at Purdue. Both teams are rated 100. 3 point home court advantage. So Purdue should win by 3. Purdue actually wins by 20. Thats one data point that says Purdue is 17 points better than Minnesota. That game's "error" is +17 points. So you look at each game, totalling the "errors" for each team. After looping thru all the games, you use the total error (per team) to adjust each team's actual ranking. So if Purdue's total error over 15 games was +45 points, you'd raise their rating by 3 points.

Then you run it again with the new ratings for each team. And again. And so on.

Eventually the ratings stop changing because the total error has been minimized.

This is a brute force approach, but it gets the job done.

This sounds like a form of regression.

I have a lot of personal experience in regression.

 

[DaCat]

Would you be willing to share the macro with me?

ChatGPT baby!

 

[DaCat]

I ran an old-fashioned scoreboard algorithm from ESPN standings to determine the rankings.

Big Ten Conference

5PUR
21NU
MD
17IU
ILL
MSU
RUTG
MICH
IOWA
WISC
PSU
NEB
OSU
MINN