BIG getting crazier

[Max_Power]

MSU loses to Ohio State at home. If I were on the committee, I’d pick Iowa and leave MSU out of the NCAA tournament.

But the computers love MSU. They are - despite losing back to back home games - No. 24 this morning. IDK how MSU gamed the NET rankings this year but it pretty wild:

24

23

Michigan St.

Big Ten

17-11

3-5

1-2

13-4

3-7

5-4

5-0

4-0

53

52

Northwestern

Big Ten

19-8

3-6

2-1

14-1

4-5

5-2

3-0

7-1

This all being said, NU and Collins need to upgrade the schedule. I would like to see NU add a home and away against an SEC and XII team - I'd love to see a Florida and Iowa State situation over the next couple of years. As MSU's schedule shows, it does not hurt and it would be fun.

 

[PurpleWhiteBoy]

But the computers love MSU. They are - despite losing back to back home games - No. 24 this morning. IDK how MSU gamed the NET rankings this year but it pretty wild:

24

23

Michigan St.

Big Ten

17-11

3-5

1-2

13-4

3-7

5-4

5-0

4-0

53

52

Northwestern

Big Ten

19-8

3-6

2-1

14-1

4-5

5-2

3-0

7-1

This all being said, NU and Collins need to upgrade the schedule. I would like to see NU add a home and away against an SEC and XII team - I'd love to see a Florida and Iowa State situation over the next couple of years. As MSU's schedule shows, it does not hurt and it would be fun.

The trick is not to upgrade the schedule. The trick is to blow teams out.

Michigan State is so overrated because they Beat Baylor by 24, beat Stony Brook by 44 and beat Penn State by 31.
If they win each of those games by 10 instead, they lose 69 points / 26 games or roughly 2.6 points in the ratings. That would be about 23 spots lower in the ratings.

 

[phatcat]

NU and Collins need to upgrade the schedule

In addition to PWBs response, I want to reiterate that, when creating the schedule, it would have been impossible to know that Det Mercy and DePaul would be THAT BAD.

 

[DaCat]

The trick is not to upgrade the schedule. The trick is to blow teams out.

Michigan State is so overrated because they Beat Baylor by 24, beat Stony Brook by 44 and beat Penn State by 31.
If they win each of those games by 10 instead, they lose 69 points / 26 games or roughly 2.6 points in the ratings. That would be about 23 spots lower in the ratings.

Wow, that's why these ratings suck. I didn't know the margin of victory would factor in so significantly.

 

[hdhntr1]

Maryland is a big big game for many reasons, imo most of all because it will be the most rested this team has been in a long time. Need to capitalize on the bye weekend.

Mostly the dings get a chance to heal. That said, after a layoff, a lot of times teams come out a little flat. Hopefully not at MD as at this point that seems like the most winnable. But if they play the way they did yesterday they will be tough. Of course the officials helped them out a lot rather than Rutgers

 

[hdhntr1]

The team we played last year, Boise State, was 29 in NET prior to the tournament and we were 41. We were a 7 seed, Boise was a 10. Rutgers was 40 in NET last year and didn’t make the tournament. NET is important but it’s just one of many things considered. Strength of record doesn’t get talked about enough but it’s important for the committee too and we rank quite favorably there (currently 26th). We did last year, too, which is why we were seeded above our predictive metrics.

But right now NE has us at around 55 and SOS in the 60s. That is danger zone

 

[phatcat]

DI Men's Basketball Rankings - NCAA Men's Basketball NET Rankings | NCAA.com

Get NCAA college basketball rankings from the Associated Press, USA Today Coaches poll and the NCAA NET Rankings.

www.ncaa.com

Someone explain to me why these teams have better Net rankings: McNeese, Cincy, Fla Atlantic, St John's, Villanova, Nebraska. Is it only margin of victory (in the victories they have)? Because their résumés are not good. We do have that single CSU stain, but McNeese and Fla Atl have Q4 losses too

 

[hdhntr1]

But the computers love MSU. They are - despite losing back to back home games - No. 24 this morning. IDK how MSU gamed the NET rankings this year but it pretty wild:

24

23

Michigan St.

Big Ten

17-11

3-5

1-2

13-4

3-7

5-4

5-0

4-0

53

52

Northwestern

Big Ten

19-8

3-6

2-1

14-1

4-5

5-2

3-0

7-1

This all being said, NU and Collins need to upgrade the schedule. I would like to see NU add a home and away against an SEC and XII team - I'd love to see a Florida and Iowa State situation over the next couple of years. As MSU's schedule shows, it does not hurt and it would be fun.

Those kind of things were in the BIG ACC challenge and Gavin games which we were not involved in this year

 

[PurpleWhiteBoy]

DI Men's Basketball Rankings - NCAA Men's Basketball NET Rankings | NCAA.com

Get NCAA college basketball rankings from the Associated Press, USA Today Coaches poll and the NCAA NET Rankings.

www.ncaa.com

Someone explain to me why these teams have better Net rankings: McNeese, Cincy, Fla Atlantic, St John's, Villanova, Nebraska. Is it only margin of victory (in the victories they have)? Because their résumés are not good. We do have that single CSU stain, but McNeese and Fla Atl have Q4 losses too

I wrote up 2 algorithms - one based on margin of victory (like KenPom and NET) and one based on probability of winning each game. PWB-Record is the method that rewards teams for winning and doesn't care much about margin. The PWB-Points algo reduces margin of victory for overtime games (duh) and weights the games much more appropriately, with games against comparable teams counting significantly more than mismatches. I'm very confident that my "Points" approach is better than the NET method. (shocker?)

Team	NET	PWB-Points	PWB-Record	Biggest Win	NET Impact
McNeese	52	63	50	beat Houston Christian by 51	+12 spots
Northwestern	53	38	29	beat OSU by 25	+8 spots
Cincy	45	47	61	beat GA Tech by 35	+10 spots
FAU	37	44	44	beat VA Tech by 34	+12 spots
Nebraska	41	36	22	beat Purdue by 16	+8 spots
St. Johns	44	42	48	beat Villanova by 20	+5 spots
Villanova	38	31	49	beat Seton Hall by 26	+10 spots

So McNeese was supposed to beat dreadful Houston Christian by about 23. Instead they won 105-54. I'd estimate that moves McNeese up by about 12 spots in the NET.

Of course, teams have bad games where they play poorly against bad teams and get penalized heavily. NU's 2 point loss to Chicago State is just barely worse than its 30 point beatdown at Illinois. Our 3rd worst underperformance (beating Western Michigan by 4) basically cancels our best win (beating Ohio State by 25). So the NET thinks we suck because of those 2 games.

 

[Gatabowl]

Wow, that's why these ratings suck. I didn't know the margin of victory would factor in so significantly.

Barely lose to or beat good teams, destroy okay to good and bad teams.

This is why it’s not unsportsmanlike to play until the final buzzer.

 

[PurpleWhiteBoy]

Barely lose to or beat good teams, destroy okay to good and bad teams.

This is why it’s not unsportsmanlike to play until the final buzzer.

That's like saying its okay to rob a bank because your old college roommate robbed a bank.

It is okay to be a jerk because you might gain something by being a jerk.

Got it.

 

[Gatabowl]

That's like saying its okay to rob a bank because your old college roommate robbed a bank.

It is okay to be a jerk because you might gain something by being a jerk.

Got it.

If everyone is a jerk no one is

 

[SmellyCat]

If everyone is a jerk no one is

 

[Gatabowl]

The more I think about it the more I’m convincing myself that the Maryland game is going to be the key game. It’s going to be a quad 1 game, we are neck and neck in kenpom, it’s more proof to ourselves and the committee that we can win on the road, and it’s another opportunity to play one of the best scoring players in the country which is great experience for the tournament and a test for our team in a post-Ty era.

Win at Maryland and guarantee at least 11-9, with the path ahead including two home games as comfortable favorites and a chance for another quad 1 road game against a team we match up well against.

 

[Secho99]

But right now NE has us at around 55 and SOS in the 60s. That is danger zone

I’m not worried at this moment for a few reasons:

1) Our descriptive metrics (the ones that are based on our results so far) that the committee considers are very good. SOR is 26, KPI is 30.

2) Our predictive metrics are fine. KenPom is 43, Torvik is 37. These are a bit skewed by our relatively poor performances in wins over Binghamton, WMU and Jackson State, plus the Chicago State debacle. Our defensive efficiency took a bigger hit allowing 75 to them than it did allowing 96 to Illinois.

3) NET itself isn’t great (53), but the committee knows it’s disproportionately influenced by the CSU loss and we look good in the most important part of NET: Quad 1 and 2 record (we are 4-5 in Quad 1, 5-2 in Quad 2, which is exactly the same record in both quads as Illinois).

4) I don’t believe raw SOS is significantly considered by the committee. Other metrics build in the quality of your games.

 

[SimpsonElmwood]

Cats have been blown out just 1 time all season. They’ve played a few bad games but won all but 1. It been an impressive year.

 

[PurpleWhiteBoy]

I’m not worried at this moment for a few reasons:

1) Our descriptive metrics (the ones that are based on our results so far) that the committee considers are very good. SOR is 26, KPI is 30.

2) Our predictive metrics are fine. KenPom is 43, Torvik is 37. These are a bit skewed by our relatively poor performances in wins over Binghamton, WMU and Jackson State, plus the Chicago State debacle. Our defensive efficiency took a bigger hit allowing 75 to them than it did allowing 96 to Illinois.

3) NET itself isn’t great (53), but the committee knows it’s disproportionately influenced by the CSU loss and we look good in the most important part of NET: Quad 1 and 2 record (we are 4-5 in Quad 1, 5-2 in Quad 2, which is exactly the same record in both quads as Illinois).

4) I don’t believe raw SOS is significantly considered by the committee. Other metrics build in the quality of your games.

Since you seem to have looked into how KenPom actually does his calculations, its worth reminding people that "Efficiency" is something of a smokescreen.

Its is simply Points/Possessions. The only way it differs from Points/Minute is that it counts possessions and uses that as the denominator. The value of that approach is not obvious, if it exists. It does enable him to talk about "Pace" (possessions per minute) but my guess is that the standard error (actual score - expected score) barely improves when he backfits the results to his ratings.

All games are weighted equally, regardless of when they were played, regardless of the competitiveness of the two opponents. Overtime is treated as extra possessions. All of these are serious flaws.

If you were to estimate NU scores 1.75 points per minute and KenPom were to say NU has an offensive efficiency of 1.05, I know which one sounds more complicated, but I'm not sure which one is more informative.

 

[CappyNU]

If you were to estimate NU scores 1.75 points per minute and KenPom were to say NU has an offensive efficiency of 1.05, I know which one sounds more complicated, but I'm not sure which one is more informative.

Well, if you were using this information to predict performance in future games, then clearly KenPom's number would be more informative, since games can have different numbers of possessions in them. If NU scores 1.75 points per minute it's a lot different if there's 2 possessions in that minute or 3, or 4.

 

[hdhntr1]

The more I think about it the more I’m convincing myself that the Maryland game is going to be the key game. It’s going to be a quad 1 game, we are neck and neck in kenpom, it’s more proof to ourselves and the committee that we can win on the road, and it’s another opportunity to play one of the best scoring players in the country which is great experience for the tournament and a test for our team in a post-Ty era.

Win at Maryland and guarantee at least 11-9, with the path ahead including two home games as comfortable favorites and a chance for another quad 1 road game against a team we match up well against.

And reality is that it is possibly the easiest for us to win (and even that will be tough on the road) with recent upgrades in performance by IA and MN and having to play MSU at Breslin being the other games

 

[PurpleWhiteBoy]

Well, if you were using this information to predict performance in future games, then clearly KenPom's number would be more informative, since games can have different numbers of possessions in them. If NU scores 1.75 points per minute it's a lot different if there's 2 possessions in that minute or 3, or 4.

Here's where I agree...

If I wanted to know how many possessions to expect in the next game, I'd want an estimate of how many possessions each team normally has. So Ken Pom would help me.

But if I wanted to know how many points to expect my team to score...

I might say we score 1.7 per minute and they allow 2.0 per minute, so I estimate we score 1.85 per minute or 74 points.
And KenPom would say "We play 70 possessions per game and they play 75 possessions per game, but we score 0.97 points per possession and they allow 1.07 points per possession, so 72.5 possessions X 1.02 points per possession = 74 points per game.

He has to estimate 4 inputs and I have to estimate 2. Normally you'd want to see the results improving enough to justify the additional complexity. And you have to be careful that the inputs are actually independent of each other. My instinct is that as possessions per game increase, points per possession decrease somewhat, which would mean they aren't independent.

 

[Secho99]

Since you seem to have looked into how KenPom actually does his calculations, its worth reminding people that "Efficiency" is something of a smokescreen.

Its is simply Points/Possessions. The only way it differs from Points/Minute is that it counts possessions and uses that as the denominator. The value of that approach is not obvious, if it exists. It does enable him to talk about "Pace" (possessions per minute) but my guess is that the standard error (actual score - expected score) barely improves when he backfits the results to his ratings.

All games are weighted equally, regardless of when they were played, regardless of the competitiveness of the two opponents. Overtime is treated as extra possessions. All of these are serious flaws.

If you were to estimate NU scores 1.75 points per minute and KenPom were to say NU has an offensive efficiency of 1.05, I know which one sounds more complicated, but I'm not sure which one is more informative.

The O and D efficiencies are just points scored/allowed per 100 possessions, adjusted to account for quality of opponent and location of the game, expressed as a prediction of how a team would fare against an average team. Because it's expressed on a per 100 possession basis, it wouldn't be affected by tempo or whether a game goes to OT. I think he includes tempo (possessions per game) only as a way to take the per 100 numbers and apply them to predicting the result of a 40-minute game. You're right that each possession is weighted equally, but maybe the argument there is that over entire seasons teams should have a relatively similar number of "important" possessions.

NU's KenPom efficiency margin right now is +15.33, which means we should beat an average team by about 15 points per 100 possessions, but since our adjusted tempo is 64.5 possessions per game, that would suggest that we would be favored over an average team by about 10 points in a 40-minute game on a neutral court. The median team (#181) in KenPom right now is Cal Baptist. The most average team (smallest difference between O and D efficiency) is Missouri State (#163).

As I understand it, raw tempo is just the number of possessions per game. For example, our game against Michigan is a 59-possession game for us: 64 FGA + 4 trips to the line (Langborg's and-1 FT doesn't count as a possession) + 5 TOs - 14 offensive rebounds = 59 possessions. The adjusted tempo is then possessions for each game adjusted for opponent (a 65-possession game is "fast" for a game against Virginia but "slow" for a game against Arizona, so adjusted tempo accounts for that).

KenPom certainly has its flaws but it's good enough that sportsbooks were forced to start using it as the basis for setting lines because bettors were exploiting discrepancies between books and KenPom to make money.

 

[drewjin]

IU downs Wisconsin! NU alone in third place in B1G, for the time being!

 

[NJCat]

IU downs Wisconsin! NU alone in third place in B1G, for the time being!

According to the predictor model, if NU just wins its last 2 home games, and Illinois bets Wisky at Kohl, NU is a 4 seed, playing the winner of Wisky-IU/RU......I like that path to the semi's......

http://bball.notnothing.net/bracket.php?conf=big10m

 

[PurpleWhiteBoy]

The O and D efficiencies are just points scored/allowed per 100 possessions, adjusted to account for quality of opponent and location of the game, expressed as a prediction of how a team would fare against an average team. Because it's expressed on a per 100 possession basis, it wouldn't be affected by tempo or whether a game goes to OT. I think he includes tempo (possessions per game) only as a way to take the per 100 numbers and apply them to predicting the result of a 40-minute game. You're right that each possession is weighted equally, but maybe the argument there is that over entire seasons teams should have a relatively similar number of "important" possessions.

NU's KenPom efficiency margin right now is +15.33, which means we should beat an average team by about 15 points per 100 possessions, but since our adjusted tempo is 64.5 possessions per game, that would suggest that we would be favored over an average team by about 10 points in a 40-minute game on a neutral court. The median team (#181) in KenPom right now is Cal Baptist. The most average team (smallest difference between O and D efficiency) is Missouri State (#163).

As I understand it, raw tempo is just the number of possessions per game. For example, our game against Michigan is a 59-possession game for us: 64 FGA + 4 trips to the line (Langborg's and-1 FT doesn't count as a possession) + 5 TOs - 14 offensive rebounds = 59 possessions. The adjusted tempo is then possessions for each game adjusted for opponent (a 65-possession game is "fast" for a game against Virginia but "slow" for a game against Arizona, so adjusted tempo accounts for that).

KenPom certainly has its flaws but it's good enough that sportsbooks were forced to start using it as the basis for setting lines because bettors were exploiting discrepancies between books and KenPom to make money.

Thanks for the response.

I'm not sure Ken Pom counts the number of possessions. I heard him recently say that he doesn't keep track of substitutions and lineups (which suggests he's using box score only). My understanding is that he estimates possessions from the box score, with free throws being the wildcard. That may have changed. I do know that Evan Miyakawa does all sorts of things with lineups to evaluate the individual players. Not sure about Torvik.

The main thing that matters, though, is how does KenPom (or the NET) aggregate the game by game "Points Per Possession" into a single number for each team. My understanding is that he uses a simple average, after adjusting for the opponent. This is the number that matters when you go to "predict" the next game.

But when you switch your frame of reference to "possessions" instead of minutes, you add an extra layer of complexity that may not improve the accuracy of the method.

In my approach, I use Points Sored Per Minute and Points Allowed per minute. This has what KenPom calls "tempo" baked into it and I can easily forecast a final score without having to bother estimating how many possessions each team will have. Simplicity is always preferred unless the added element actually does something significant to reduce the error in the dataset.

 

[ricko654321]

The trick is not to upgrade the schedule. The trick is to blow teams out.

Michigan State is so overrated because they Beat Baylor by 24, beat Stony Brook by 44 and beat Penn State by 31.
If they win each of those games by 10 instead, they lose 69 points / 26 games or roughly 2.6 points in the ratings. That would be about 23 spots lower in the ratings.

For KenPom purposes, if the average # of possessions per team in those games is ~67 it’s actually about 1.5x that number so a difference of 3.9…

Yes failing to rack up larger margin of victory (particularly against the bad teams) has been our biggest problem for the predictive metrics this year (and probably last year too). Either play a tougher non-con schedule, or else blow out the bad teams you play.

In reality it seems like something like strength of record matters more to the NCAA committee than NET or KenPom rating. But given how widely quoted those are, it is annoying.

 

[phatcat]

ailing to rack up larger margin of victory (particularly against the bad teams) has been our biggest problem

I've said it before, but I'm glad this isn't used in football. NU won 8 games and had a negative point differential. In 2018, NU was +1 per game on a 9 win team. 2015 was 10 wins, same +1 point differential.