Post Gonzaga Aike Model Update

[Aike]

We actually slipped one spot. Something you see a lot when two good teams play. They both knock each other down some. We dropped by 0.17 and Gonzaga dipped by 0.26.

Latest Update;

1. Auburn 4.75
2. Gonzaga 4.04
3. Tennessee 3.95
4. Duke 3.38
5. Kansas 3.19
6. Mississippi St 2.742
7. Kentucky 2.736
8. Connecticut 2.60
9. Houston 2.52
10. Iowa St 2.51
11. Marquette 2.49
12. Cincinnati 2.37
13. Penn St 2.26
14. Alabama 2.24
15. Michigan 2.20
16. Florida 2.18
17. Texas Tech 2.13
18. Georgia 2.12
19. Ohio St 2.10
20. Mississippi 2.02
21. Maryland 1.94
22. Arkansas 1.86
23. Baylor 1.85
24. Michigan St 1.81
25. Oregon 1.70
29. Clemson 1.57
30. St John’s 1.56
33. BYU 1.45
35. Purdue 1.39
37. Oklahoma 1.35
38. Louisville 1.32
39. Vanderbilt 1.29
42. North Carolina 1.17
44. Texas 1.05
48. Indiana 0.95

I think it’s worth pointing out that 20 teams are currently projected to win 2 tournament games or better. Obviously only 16 can actually accomplish this.

45 teams are currently projected to win 1+ tournament games. Again, only 32 can pull this off.

What this basically means is that at least at this point in the season, there is a lot more strength at the top of college basketball than in recent seasons.

Will be interesting to see how this plays out once we get into conference play and quality meets quality. May have more effects like we saw last night, with 2 good teams pulling each other down.

For the record, no team has won the tournament in the years I’ve been modeling that wasn’t at least a 2 entering the tournament.

 

[Cowtown Cat]

WOW! How good is Mississippi State?! I’m about to go look at the SEC results per team.

 

[Aike]

WOW! How good is Mississippi State?! I’m about to go look at the SEC results per team.

Beat Pitt by 33.

 

[ukjenning44]

We beat two teams ahead of us, this model must be broken lol

 

[Aike]

We beat two teams ahead of us, this model must be broken lol

Right. Because those were the only games any of the teams played. Gonzaga didn’t kill Baylor. Duke didn’t beat Auburn.

 

[ukjenning44]

Right. Because those were the only games any of the teams played. Gonzaga didn’t kill Baylor. Duke didn’t beat Auburn.

Baylor is garbage. Best win is Arkansas

 

[catben]

How do you square U Conn at 8 with 3 straight losses to mediocre teams?

Just asking.

 

[Aike]

Baylor is garbage. Best win is Arkansas

Thanks bro

 

[Aike]

How do you square U Conn at 8 with 3 straight losses to mediocre teams?

Just asking.

Data is aggregated. Doesn’t take streaks or trends into account.

They also just beat Baylor, and every other win they have is by 35+.

I’d say the stats they’re accumulating in all the blowouts are propping them up. If they start piling up more Ls during conference play, they will drop. But I wouldn’t count them out just yet.

 

[bowfreak]

My favorite thread every time it is posted.

It is a model! It is a tool! Aike created the model, it is not necessarily his opinion.

 

[Aike]

My favorite thread every time it is posted.

It is a model! It is a tool! Aike created the model, it is not necessarily his opinion.

Haha yeah. It’s a data point among a sea of them. None of them ever have or ever will get it perfect.

 

[travisbickle]

@Aike does your model’s predicted tournament wins sum up to 63 (or 67 with the First Four games)? Based on what I’m seeing out of this list I’d say it doesn’t.

What do the wins sum up to, and does that total stay constant or fluctuate?

 

[JonathanW]

Seeing UConn in your top 10, do you consider/use data from prior year?

Also, how do you treat these not-really-Neutral court games (like UK vs Gonzaga in Seattle, KU playing a team in KC, UK playing in Louisville, etc…)?
I called those semi-home or semi-road games, placing a value on them halfway between Home/Road and Neutral.
I had calculated (from a large sample size) that home teams win 63% of games, only looking at conference games so that on average it is a game of equals.

 

[Aike]

@Aike does your model’s predicted tournament wins sum up to 63 (or 67 with the First Four games)? Based on what I’m seeing out of this list I’d say it doesn’t.

What do the wins sum up to, and does that total stay constant or fluctuate?

It’s based on 63 wins. I don’t count the First Four as wins in the model. It could sum up to a little more or less than 63 depending on the strength of the field in any given year.

I think this is the right way to do it because you have odd matchups in the First Four.

An 11 beating an 11 is not the same as a 16 beating a 16. Also some 11s are playing 6s and some 16s are playing 1s on the actual first round, It’s apples to oranges.

I could probably play around with it, but it’s simple enough to not count a win until the round of 64.

Once the bracket is set, it basically will self-adjust according to the odds on individual matchups.

No matter what the prior prediction looks like, you are locked in based on competition. If you’re a 2 playing a 3 in the second round, you only have a 40% chance of advancing. So you’re “effective score” may be adjusted based on competition, and your predicted win total may end up different than your raw score would suggest.

 

[Aike]

Seeing UConn in your top 10, do you consider/use data from prior year?

Also, how do you treat these not-really-Neutral court games (like UK vs Gonzaga in Seattle, KU playing a team in KC, UK playing in Louisville, etc…)?
I called those semi-home or semi-road games, placing a value on them halfway between Home/Road and Neutral.
I had calculated (from a large sample size) that home teams win 63% of games, only looking at conference games so that on average it is a game of equals.

Prior year is not included. I am not personally adjusting based on location. I use a publicly available SOS calculation that I’ve found to be useful in my models. This calculation “should” be accounting for where a game was played.

I’ve thought about calculating my own separate SOS, but haven’t really had the need. May do it just for fun sometime, and would probably use a Ken Massey linear algebra approach, with adjustments similar to what you suggest for odd locations.

As far as using my model for predictions, it is always predicting based on a neutral setting, as all NCAA Tournament games are on a neutral court. One should adjust road or “semi-neutral” point spreads based on this information.

Your 63% number sounds spot on, as 63% odds of victory would equate to a spread of around 3.5. This is inline with commonly accepted average homecourt advantage.

 

[Aike]

I left out 43. Missouri.

They’re about to get a bump.

 

[KYFOSSIL]

Right. Because those were the only games any of the teams played. Gonzaga didn’t kill Baylor. Duke didn’t beat Auburn.

But Goncaga has 2 losses and ewe beat Duke

 

[Aike]

But Goncaga has 2 losses and ewe beat Duke

Ok

 

[KYFOSSIL]

Ok

Good answer

 

[KyFanInTenn]

Always appreciate you posting this.

Feel like Auburn is pretty clearly the best team to me as your model shows. Know they lost to Duke but I was so impressed with them throughout that game. Just not a lot of weaknesses there. Tennessee has been awfully impressive each time I've watched them as well - feel like this is the most balance they've had since the Grant/Admiral team. Duke is obviously talented and a really good team, but I just don't like their guard play and feel like it might let them down in March.

I truly don't know how good we actually are, but we've beaten two clear top 5 teams. I never felt we were actually "better" but this team absolutely knows how to win. I still think we have a ton of room to grow too. This coaching staff and the toughness/maturity of the players is just going to allow us to punch above our weight class in every single game.

 

[Aike]

Good answer

There wasn’t a question

 

[KYFOSSIL]

There wasn’t a question

Ok I mention two factors that I believed should factor in lowering those teams projections but Im not familiar how the model works.

 

[Friedas_Boss]

We actually slipped one spot. Something you see a lot when two good teams play. They both knock each other down some. We dropped by 0.17 and Gonzaga dipped by 0.26.

Latest Update;

1. Auburn 4.75
2. Gonzaga 4.04
3. Tennessee 3.95
4. Duke 3.38
5. Kansas 3.19
6. Mississippi St 2.742
7. Kentucky 2.736
8. Connecticut 2.60
9. Houston 2.52
10. Iowa St 2.51
11. Marquette 2.49
12. Cincinnati 2.37
13. Penn St 2.26
14. Alabama 2.24
15. Michigan 2.20
16. Florida 2.18
17. Texas Tech 2.13
18. Georgia 2.12
19. Ohio St 2.10
20. Mississippi 2.02
21. Maryland 1.94
22. Arkansas 1.86
23. Baylor 1.85
24. Michigan St 1.81
25. Oregon 1.70
29. Clemson 1.57
30. St John’s 1.56
33. BYU 1.45
35. Purdue 1.39
37. Oklahoma 1.35
38. Louisville 1.32
39. Vanderbilt 1.29
42. North Carolina 1.17
44. Texas 1.05
48. Indiana 0.95

I think it’s worth pointing out that 20 teams are currently projected to win 2 tournament games or better. Obviously only 16 can actually accomplish this.

45 teams are currently projected to win 1+ tournament games. Again, only 32 can pull this off.

What this basically means is that at least at this point in the season, there is a lot more strength at the top of college basketball than in recent seasons.

Will be interesting to see how this plays out once we get into conference play and quality meets quality. May have more effects like we saw last night, with 2 good teams pulling each other down.

For the record, no team has won the tournament in the years I’ve been modeling that wasn’t at least a 2 entering the tournament.

Don't look now, the Fighting Mark Byington's are making a move.

Dores had a nice road win today in Dallas against TCU (neutral court game in the same city where the school is located) by 7. Y'all are going to love watching Tyler Tanner if you haven't seen him yet.

I can't believe how good and deep the SEC actually is, insane.

Thanks for sharing the rankings.

 

[Aike]

Don't look now, the Fighting Mark Byington's are making a move.

Dores had a nice road win today in Dallas against TCU (neutral court game in the same city where the school is located) by 7. Y'all are going to love watching Tyler Tanner if you haven't seen him yet.

I can't believe how good and deep the SEC actually is, insane.

Thanks for sharing the rankings.

Good luck to you guys, except when you’re playing the Cats. Feel free to knock off anyone else you want.

 

[Friedas_Boss]

Good luck to you guys, except when you’re playing the Cats. Feel free to knock off anyone else you want.

Same to ya!

I think the game in Memorial is going to be absolutely electric this year. Can't wait.

 

[blue cincy]

Wonder what happens to KU after losing at Mizzou and not really being in that game what so ever.

 

[BBUK]

We actually slipped one spot. Something you see a lot when two good teams play. They both knock each other down some. We dropped by 0.17 and Gonzaga dipped by 0.26.

Latest Update;

1. Auburn 4.75
2. Gonzaga 4.04
3. Tennessee 3.95
4. Duke 3.38
5. Kansas 3.19
6. Mississippi St 2.742
7. Kentucky 2.736
8. Connecticut 2.60
9. Houston 2.52
10. Iowa St 2.51
11. Marquette 2.49
12. Cincinnati 2.37
13. Penn St 2.26
14. Alabama 2.24
15. Michigan 2.20
16. Florida 2.18
17. Texas Tech 2.13
18. Georgia 2.12
19. Ohio St 2.10
20. Mississippi 2.02
21. Maryland 1.94
22. Arkansas 1.86
23. Baylor 1.85
24. Michigan St 1.81
25. Oregon 1.70
29. Clemson 1.57
30. St John’s 1.56
33. BYU 1.45
35. Purdue 1.39
37. Oklahoma 1.35
38. Louisville 1.32
39. Vanderbilt 1.29
42. North Carolina 1.17
44. Texas 1.05
48. Indiana 0.95

I think it’s worth pointing out that 20 teams are currently projected to win 2 tournament games or better. Obviously only 16 can actually accomplish this.

45 teams are currently projected to win 1+ tournament games. Again, only 32 can pull this off.

What this basically means is that at least at this point in the season, there is a lot more strength at the top of college basketball than in recent seasons.

Will be interesting to see how this plays out once we get into conference play and quality meets quality. May have more effects like we saw last night, with 2 good teams pulling each other down.

For the record, no team has won the tournament in the years I’ve been modeling that wasn’t at least a 2 entering the tournament.

P_SS on Auburn and car-kicker U got spanked...

 

[runt#69]

Wonder what happens to KU after losing at Mizzou and not really being in that game what so ever.

Chip off Bill Self's toupee' on the Mount Rushmore of college basketball coaches? Make him go toupee - less for life, all chiseled in stone ?

I think the penalty fits the crime of losing that game.

 

[ukjenning44]

Wonder what happens to KU after losing at Mizzou and not really being in that game what so ever.

I think they move up two spots in this broken model.

 

[BBUK]

The model is not broken.

 

[Aike]

Wonder what happens to KU after losing at Mizzou and not really being in that game what so ever.

Bound to take a bite out of them, but Missouri isn’t bad. They look like a tournament team, but it’s really early and obviously the SEC is brutal.

 

[BBUK]

Don't look now, the Fighting Mark Byington's are making a move.

Dores had a nice road win today in Dallas against TCU (neutral court game in the same city where the school is located) by 7. Y'all are going to love watching Tyler Tanner if you haven't seen him yet.

I can't believe how good and deep the SEC actually is, insane.

Thanks for sharing the rankings.

Hi Fred, how's it going? Well I hope ...

(They call him Fred ..) 😉

 

[Friedas_Boss]

Hi Fred, how's it going? Well I hope ...

(They call him Fred ..) 😉

All good, just happy to have a decent basketball team again. Hope you’re well, my friend.

 

[UK67KU59]

Beat Pitt by 33.

Would like to see where they land after the stinker of an egg they laid against Prarie View A&M 😳

I watched them beat SMU in a true road environment and they looked good.

 

[Aike]

Would like to see where they land after the stinker of an egg they laid against Prarie View A&M 😳

I watched them beat SMU in a true road environment and they looked good.

Wild that game was so close. Had to have hurt.

 

[JonathanW]

Prior year is not included. I am not personally adjusting based on location. I use a publicly available SOS calculation that I’ve found to be useful in my models. This calculation “should” be accounting for where a game was played.

I’ve thought about calculating my own separate SOS, but haven’t really had the need. May do it just for fun sometime, and would probably use a Ken Massey linear algebra approach, with adjustments similar to what you suggest for odd locations.

As far as using my model for predictions, it is always predicting based on a neutral setting, as all NCAA Tournament games are on a neutral court. One should adjust road or “semi-neutral” point spreads based on this information.

Your 63% number sounds spot on, as 63% odds of victory would equate to a spread of around 3.5. This is inline with commonly accepted average homecourt advantage.

I really do not like/agree-with the standard SOS calculations, which I think is a simple average ranking across their opponents.

Let me give you an example, for 2 top 20 teams who play 10 games:

Team A plays vs 1 top 20, 2 in 20-50, 2 in 50-100, 5 in 100-200, 0 in 200-350 for an average SOS of 98.
Team B plays 3 top 20, 3 in 20-50, 1 in 50-100, 0 in 100-200, 0 in 200-300, and 3 in 300-350, for an average SOS of 118.
So Team A gets the better (lower) SOS, even though relative to their own ranking Team B has had a tougher schedule playing 6-7 teams with a legit chance of beating them, while Team A played only 3-5 teams with a legit change of beating them.

So if you are a top 20 team, whether you play a team ranked 250 or 350 is irrelevant (you should always win that game). But if you are a team at 200 or even 150, then that 250 game may be tougher than the 350 game.



For my ranking system, I didn't use SOS, actually SOS was just a by-product of my rankings so I didn't put much/any effort into calculating SOS. I calculated the probability (based on game data) for Team X to beat Team Y (on a neutral court), and did that for every possible matchup (so 350x 350 matrix of probabilities). Then for each team their score was the sum of all of their (350) probabilities.
Factors used to calculate the probabilities included: margin of victory (MoV), was game played Home or Away or Neutral or Semi-Home or Semi-Away, did the game go into OT (which affects MoV), and how long ago was the game played because I weighted more recent games (say in Feb-Mar) slightly more than distant games (Nov). Using those factors, if 2 teams played, I would have a probability of who would win if they played again (on a neutral court). Of course that left a very incomplete (350x350) matrix. But each time I multiplied the matrix by its inverse, it added more data, the first time it added opponents of your opponents, the 2nd time it added opponents of your opponent's opponents, and so on. So the 1 additional factor in that calculation was how much to weight how Team X did playing vs Team Y, compared to how well Team X's opponents also did vs Team Y. Let's just say it was a complex SAS program I wrote for this.

 

[Wildcats1st]

WOW! How good is Mississippi State?! I’m about to go look at the SEC results per team.

Every team in the SEC is like the 1990 unlv team lol

 

[Aike]

I re-ran just now. We stayed at 7. Kansas and Mississippi St fell back. Iowa St and UConn jumped us.