A strong SOS Often Translates to Losing More Games

[Just Sayin]

Per KenPom Rankings of SOS- All Division 1 Teams

Record: Top 10 SOS teams
3-6
1-7
0-8
1-7
2-8
2-5
3-2
1-7
1-9
2-6


Record: Bottom 10 SOS teams
5-4
6-1
7-3
5-4
3-4
7-1
7-1
6-1
5.2
7-2

This actually makes sense. The stiffer the competition, the tougher it is to win and vice-versa. The value of SOS when running a multiple regression to ascertain how much weight to give different variables which contribute to winning games for prediction purposes is negligible. Better to leave it out. It is however intuitive that top-ranked teams with good records and who have tougher schedules are probably much better than other teams with similar W-L records but with average SOS. But using SOS as a dependant variable and including it in a regressiona formula will not be a better predictor of wins than by leaving it out. There rarely is a positive correlation between wins and SOS as the sample above shows.

[vok22]

Interesting you bring this up. A few years ago I made a very simple regression model in attempt to predict Valpo basketball conference (so I could avoid having to create baseline rankings for every team) games. While the model did go 16-2, I did realize that the impact of SOS was pretty much negligible in forecasting games. Also of note, the model went 16-2 in the regular season, but something like 1-3 in the postseason for Valpo. This was the year we went to the MVC championship game, so not surprising. However, I thought it was interesting because it really highlighted march MADNESS to me. The regular season was somewhat predictable, but all bets are off once March starts.

[wh]

I don't think you can draw any logical conclusions about SOS at this early stage of the season. Right now, every team from 1-10 is a mid major. Once conference play begins Major programs will dominate the top 50.

[Just Sayin]

Interesting you bring this up. A few years ago I made a very simple regression model in attempt to predict Valpo basketball conference (so I could avoid having to create baseline rankings for every team) games. While the model did go 16-2, I did realize that the impact of SOS was pretty much negligible in forecasting games. Also of note, the model went 16-2 in the regular season, but something like 1-3 in the postseason for Valpo. This was the year we went to the MVC championship game, so not surprising. However, I thought it was interesting because it really highlighted march MADNESS to me. The regular season was somewhat predictable, but all bets are off once March starts.

Same experience. It's tough to predict the big dance. I'm gathering data for all teams and running previous years best models with some tweaking. It's fun to work the whole year trying different things that you can't wait to try the following year. 

[Just Sayin]

Interesting you bring this up. A few years ago I made a very simple regression model in attempt to predict Valpo basketball conference (so I could avoid having to create baseline rankings for every team) games. While the model did go 16-2, I did realize that the impact of SOS was pretty much negligible in forecasting games. Also of note, the model went 16-2 in the regular season, but something like 1-3 in the postseason for Valpo. This was the year we went to the MVC championship game, so not surprising. However, I thought it was interesting because it really highlighted march MADNESS to me. The regular season was somewhat predictable, but all bets are off once March starts.

Do you mind sharing what model you created?  We can take this discussion off- topic.

[Just Sayin]

I don't think you can draw any logical conclusions about SOS at this early stage of the season. Right now, every team from 1-10 is a mid major. Once conference play begins Major programs will dominate the top 50.

A few years ago, I ran a straight correlation for every team post-tournament Numbers to test the relation between win% and sos.  The CORR was negative.

[wh]

I don't think you can draw any logical conclusions about SOS at this early stage of the season. Right now, every team from 1-10 is a mid major. Once conference play begins Major programs will dominate the top 50.

A few years ago, I ran a straight correlation for every team post-tournament Numbers to test the relation between win% and sos.  The CORR was negative.

Ah, that is revealing.