Manipulating the NET

[djcon57]

If you went by point differential in the NFL this year, the SAINTS would have made the playoffs over the LIONS.

 

[Bayside44]

How so? They don't use point differential?

This is my last post on this. Coaches manage the game to win the game, not to win the game by a certain spread. We had a guy here for 47 years that was the master at that.

Hence my original post that the margin of victory part of this is stupid.

 

[djcon57]

I think that also goes back to what I was saying before (and other people have as well) that once you get past a certain spot (top 20?) there just isnt much of a difference.

I'll use Ken Pom because they have projected scores for games. Virginia Tech is 30 spots ahead of us at Ken Pom. And at KP, we're projected to beat them tomorrow by, 51% win probability. So they're 30 spots ahead of us, and when they come to our place, we're rated slightly better.

Edit: Also, look, if you're just counting the losses, fine, but we've run pretty hot in close games and I know people probably won't like this comment but the fact of the matter is we probably should have lost a few more games at which point the NET/KP ratings would probably look a little more in line.

I get what you are saying, but the FACT is we won the games, that should count more.

FWIW, I don't mind that we are low in the NET, I don't mind if thats the reason we don't get in, etc (we should win more if that's the case), I DO mind that people are suggesting Red should coach accordingly. It's stupid.

 

[Knicks411]

I'm aware, I have kept track of this sort of thing over the years. We are 5-2 in games decided by five points or less. The previous two years we were 7-13 combined.

All I am asking is there should be somewhat less of an emphasis(or cap like they did originally) on point differential. At some point you're encouraging teams to schedule weak teams to boat race them and to play to just keep the game close instead of to win(end of game fouling etc).

My gut says if they did some kind of a discount on point differential (like I mentioned above) it A) wouldnt really matter a whole lot, and B) would serve to just bring the best teams down and the worse teams higher.
I took a look at our schedule, our net rating this year is +1.8. We've outscored teams by 1.77 points per 100 possessions this season. That works out to 36 total points all season.

I went through our schedule and used a 15 point cap on MOV. So if we won a game by more than 15 points, then it would count for 15, if we lost by more than 15, it would count for 15. Our total point differential for all season increased from 36 points to 43 points, a difference of 0.3 points per 100 possessions. Our average game is around 71 possessions, so about 0.24 points per game.

 

[Fly Rodder]

90% posts about "metrics" on here are like what dogs would post about vacuum cleaners. They're not sure what they are, or what they're used for, but they're pretty damned sure that they're evil and out to get them.

 

[Knicks411]

The main point that I am probably struggling to illustrate is the takes of "we should reward actually winning games" and "we should cap MOV" or whatever kind of go against one another. The teams with the best MOV are (generally!) the teams that win the most games! If you cap MOV at a certain number you're going to (probably! i havent sat down and done all that math, if I even could) move teams up that lose a lot of games because they are 1) more likely to lose by a lot, so you're helping their MOV and 2) when they win, not win by a lot, so they won't get hurt by capping their wins.

 

[Cusefan0307]

My gut says if they did some kind of a discount on point differential (like I mentioned above) it A) wouldnt really matter a whole lot, and B) would serve to just bring the best teams down and the worse teams higher.
I took a look at our schedule, our net rating this year is +1.8. We've outscored teams by 1.77 points per 100 possessions this season. That works out to 36 total points all season.

I went through our schedule and used a 15 point cap on MOV. So if we won a game by more than 15 points, then it would count for 15, if we lost by more than 15, it would count for 15. Our total point differential for all season increased from 36 points to 43 points, a difference of 0.3 points per 100 possessions. Our average game is around 71 possessions, so about 0.24 points per game.

The difference in margin between teams 21-100 and 1-20 is about the same. Michigan State is ranked 20 in KP and has one better win than us with a record of 17-11. Once you get past a certain point I would make it more about games won and overall record than some metric where there would be at most a one possession difference in spread on a neutral court.

 

[SWC75]

If you want to believe that it isn't that's fine. It's still dumb for a zillion variable reasons that take place every game.

Point differential against over-matched teams is dumb. Point differential against comparable teams is not.

 

[Sbucketball]

I've said this before, but maybe a weighted average scoring margin would be better than NET efficiency (which correlates strongly to margin of victory since it's just points scored per possession minus points let up per possession). Could break the game into 4 segments, with the first 10 minutes getting 1x weight, second 10 minutes getting 2x weight, third 10 minutes getting 3x weight, and final score getting 4x weight (or maybe just ignore the first 10 minutes and weight the final 3 segments or something, idk). That might better tell the story of a game than the current NET efficiency standard.

A 12 point loss could just as easily have been a no-contest blowout as it could have been a tight game that came down to the final two minutes.

The weighted system could tell the difference between a 12-point final that's a tight game where one team missed its shots the last few possession and the winning team hit free throws:

1st 10 mins: Ave difference 2
2nd 10 mins: ave difference 5
3rd 10 mins: ave difference -4
final score: 12

weighted average: ((2x1) + (5x2) + (-4x3) + (12x4)) / 10 units = 4.8

vs. a 12-point game blowout where maybe the losing team hit a couple late 3s v the walk-ons:

1st 10 mins: ave difference 8
2nd 10 mins: ave difference 14
3rd 10 mins: ave difference 22
final score: 12

weighted average: ((8x1) + (14x2) + (22x3) + (12x4)) / 10 units = 15

As I understand it, those two very different games -- one a blowout and one a tight game that slipped away at the end -- are treated the same under NET efficiency.

 

[domestu]

Really the lower majors, (if it's a one-bid conference, they aren't mid majors), belong n a basketball FCS - o Division 2. But a team that won the regular season deserves to be in the Big Dance more than a team that won the conference tournament - or a team that finished 7th or 8th in another conference.

I agree that a school that wins a regular season championship has done more to deserve a shot in the NCAAs than a team that wins a conference tournament, but after every conference went to money-making league tourneys, the NCAA had to install a selection system which limits the minor conferences to one bid.

The Atlantic 10 and American Athletic are considered mid-majors because they often have multiple teams in March madness. PAC-12 hoops have tumbled downward and are approaching mid-major level status in its final year of existence. You have to be good enough to earn an at-large bid.

Everyone is always going to disagree about the final bubble teams that get in or are left out of the Big Dance, but the main objective of the selection committee is to ensure that all of the most deserving teams get in. Fans want to see all of the best teams in the tournament.

I did a radio sports talk show for many years and will never forget a Sunday night program in March back in the 1990s when one of the co-hosts started arguing that every school should be allowed to participate in the NCAA Tournament. I realized that night that people can disagree about anything, no matter how absurd it is. Go Cuse!

 

[Knicks411]

The difference in margin between teams 21-100 and 1-20 is about the same. Michigan State is ranked 20 in KP and has one better win than us with a record of 17-11. Once you get past a certain point I would make it more about games won and overall record than some metric where there would be at most a one possession difference in spread on a neutral court.

I guess my question is what do you mean by "it"? Like selection criteria? Because its not like they go strictly by the NET, they do seem to mostly use it as a sorting tool and general ranking system. It's not like they just take the best at large teams ranked by NET. (Just checked last season, the last 4 in had NET rankings of 48,67,66, and 37 (with the last team in having the best net rating) and the first 4 out were 43,40,46, and 60. So if you just ranked them by NET then 3 of the first 4 out would've been in over 3 of the last 4 in)

I'm more talking about the NET as a blunt instrument to rank 350 or whatever teams who play 30 game regular seasons with very little crossover of games on a relative basis.

 

[domestu]

A bunch of teams from the majors conferences with its NET in the 40s did not get in last year. A number of schools with a poorer NET ranking jumped past other bubble teams because of their play late in the season and in their conference tournament. The NET is not the overriding factor used in selecting schools. How teams play on the road and at neutral sites is also looked at.

 

[Cusefan0307]

I guess my question is what do you mean by "it"? Like selection criteria? Because its not like they go strictly by the NET, they do seem to mostly use it as a sorting tool and general ranking system. It's not like they just take the best at large teams ranked by NET. (Just checked last season, the last 4 in had NET rankings of 48,67,66, and 37 (with the last team in having the best net rating) and the first 4 out were 43,40,46, and 60. So if you just ranked them by NET then 3 of the first 4 out would've been in over 3 of the last 4 in)

I'm more talking about the NET as a blunt instrument to rank 350 or whatever teams who play 30 game regular seasons with very little crossover of games on a relative basis.

That did not type out like my brain wanted it to. I would weigh wins more and scoring margin less in the formula. Wins matter. I get how it is used, but like the RPI it's starting to be covered in the media more than it should.

TCU has played 8 quad 4 games, Texas 9, Oklahoma 8, Texas Tech 9 and we have played 4. Those schools all have similar records as us. It's going to encourage this going forward. Why play a top 10 school in November if you can go beat Houston Christian by 40. Quad 4 having 200 teams is a bit absurd. There's a big difference between Houston Christian and Louisville.

 

[Millhouse]

The main point that I am probably struggling to illustrate is the takes of "we should reward actually winning games" and "we should cap MOV" or whatever kind of go against one another. The teams with the best MOV are (generally!) the teams that win the most games! If you cap MOV at a certain number you're going to (probably! i havent sat down and done all that math, if I even could) move teams up that lose a lot of games because they are 1) more likely to lose by a lot, so you're helping their MOV and 2) when they win, not win by a lot, so they won't get hurt by capping their wins.

The teams with the best MOV are generally the teams that you can expect will win the most games in the future. It will tell you who has been lucky or unlucky.

In pro leagues, it doesn't matter if you're lucky or unlucky, it just matters what your record is. I get why college basketball can't be like that because everyone plays different schedules but i wish record mattered more

 

[Millhouse]

I get splitting hairs on records because I do the same, but I don't think Clemson is a good comparison. They have beat a projected 2,3,7, and 9 seed. 5 of their 8 losses by 3 points or less.

We have one great win and that's it. I'm comfortable with the committee thinking they're better than us.

On the other hand teams like Iowa, Utah and Texas A&M being ahead of us make no sense.

they have 1 more win against a similar schedule. if they have a more good wins, then they must have losses against teams that aren't good.

all the criteria people care about have to mean caring less about something else.

if you pick one team over another with the same record because of road wins, it has to mean you care less about home losses, etc.

 

[upperdeck]

I've said this before, but maybe a weighted average scoring margin would be better than NET efficiency (which correlates strongly to margin of victory since it's just points scored per possession minus points let up per possession). Could break the game into 4 segments, with the first 10 minutes getting 1x weight, second 10 minutes getting 2x weight, third 10 minutes getting 3x weight, and final score getting 4x weight (or maybe just ignore the first 10 minutes and weight the final 3 segments or something, idk). That might better tell the story of a game than the current NET efficiency standard.

A 12 point loss could just as easily have been a no-contest blowout as it could have been a tight game that came down to the final two minutes.

The weighted system could tell the difference between a 12-point final that's a tight game where one team missed its shots the last few possession and the winning team hit free throws:

1st 10 mins: Ave difference 2
2nd 10 mins: ave difference 5
3rd 10 mins: ave difference -4
final score: 12

weighted average: ((2x1) + (5x2) + (-4x3) + (12x4)) / 10 units = 4.8

vs. a 12-point game blowout where maybe the losing team hit a couple late 3s v the walk-ons:

1st 10 mins: ave difference 8
2nd 10 mins: ave difference 14
3rd 10 mins: ave difference 22
final score: 12

weighted average: ((8x1) + (14x2) + (22x3) + (12x4)) / 10 units = 15

As I understand it, those two very different games -- one a blowout and one a tight game that slipped away at the end -- are treated the same under NET efficiency.

yes. There is little logic built into show the flow of the game in these metrics. you can lead by 20 all game and play subs to win by 10 or lead by 10 all game and play starters the last 2 min to win by 20.. vastly different game control but the metrics would rate one much better.

 

[atlanticuse]

18-10 against the 20th best schedule and we are 82nd in SRS
Florida State is 14-13 against the 22nd schedule and has a nicer 69th SRS.
Clemson 19-8 against 25th schedule, SRS is 24

Scoring Margin is a great predictor but if you're just evaluating a season, wins and losses have to matter more

Are ncaa berths predictions or rewards?

I wish we could model "what-if" like if we only lost by 3-5 points instead of 20+.
Apologies, because this is not scientific, but I calculated win and loss margins for a couple of schools:

Pitt - 18-9
13.56 LOSS MARGIN
18.83 WIN MARGIN

Mich State - 17-11
7.55 LOSS MARGIN
18.71 WIN MARGIN

UNC - 21-6
4.83 LOSS MARGIN
18.67 WIN MARGIN

Wake Forest - 18-9
7.11 LOSS MARGIN
18.33 WIN MARGIN

Syracuse - 18-10
17.80 LOSS MARGIN
11.89 WIN MARGIN

 

[Knicks411]

That did not type out like my brain wanted it to. I was saying is I would weigh wins more and scoring margin less in the formula. Wins matter. I get how it is used, but like the RPI it's starting to dominate the RPI by the media more than it should.

TCU has played 8 quad 4 games, Texas 9, Oklahoma 8, Texas Tech 9 and we have played 4. Those schools all have similar records as us. It's going to encourage this going forward. Why play a top 10 school in November if you can go beat Houston Christian by 40. Quad 4 having 200 teams is a bit absurd. There's a big difference between Houston Christian and Louisville.

Ok I hear what you're saying.

To the bolded point, as a general rule, I would be in favor of anything that results in more good teams playing each other, but I am not sure focusing on wins would get you that result. If wins matter more than you should go play Houston Christian instead of playing the top 10 team. I think wins over good teams should count the most.

It's kind of funny that this has come nearly full circle, I remember the days when we were really good (it was a long time ago, sadly) and played a bad OOC and the argument was it didnt matter we had a bad OOC because of how many good teams we'd play in conf play and now its the teams doing that who are the bad guys.

 

[Knicks411]

yes. There is little logic built into show the flow of the game in these metrics. you can lead by 20 all game and play subs to win by 10 or lead by 10 all game and play starters the last 2 min to win by 20.. vastly different game control but the metrics would rate one much better.

I guess I just kind of quibble with the "much better". You're talking about 10 points over the course of a 30 game season, and you're only assuming it happens in one direction and not the other

Cleaning the Glass, which covers the NBA, does a thing where they filter out stats in garbage time. Would be a lot harder to implement at the college level I think

 

[Cusefan0307]

Ok I hear what you're saying.

To the bolded point, as a general rule, I would be in favor of anything that results in more good teams playing each other, but I am not sure focusing on wins would get you that result. If wins matter more than you should go play Houston Christian instead of playing the top 10 team. I think wins over good teams should count the most.

It's kind of funny that this has come nearly full circle, I remember the days when we were really good (it was a long time ago, sadly) and played a bad OOC and the argument was it didnt matter we had a bad OOC because of how many good teams we'd play in conf play and now its the teams doing that who are the bad guys.

I posted on here recently that if the NET existed in 02, 06 and 07 we probably get in. 2007 we definitely get in.

 

[pokercuse08]

Say one team is good and shoots 50% every 20 secs and another team is good and shoots 50% every 30 secs. One is going to score more often but both teams pretty much do the same thing on offense.

Assuming they are equally skilled/deep, play the same schedule, and are better than most of their opponents, the higher tempo team should win more. If they're worse than most of their opponents, the higher tempo team should lose more.

Increasing the number of possessions, all else staying equal, reduces the variance (luck factor).

On the other hand, take team number 3. They're another carbon copy of these two teams. They play the faster tempo, but if they get a certain sized lead (and they've calculated this inflection point well), they switch to the slower tempo. If they fall behind by a certain amount, they play even faster. This team would win the most often of the three.

The issue with NET is that Team #3, which is actually the best version of this team, may well have the worst NET ranking. They're going to have fewer blowout wins, because they take the air out of the ball with a big lead. When they're trailing, they press harder to maximize their chances at winning, but they thus lose by a bigger margin more often.

 

[NorSyr66]

If the tourney expanded by 50%, my interest in CBB would decrease proportionally. Maybe that’s just me.

I can hear what you are saying but after the first weekend there would only be 25% of the teams remaining! Tournament of 64 teams has 16 remaining and tournament of 96 teams (your 50% expansion) would have 24 team remaining. It would sort of be like the "top 25" remaining to play for the championship

 

[Knicks411]

I posted on here recently that if the NET existed in 02, 06 and 07 we probably get in. 2007 we definitely get in.

I'm gonna use Ken Pom because it's A) available and B) probably close enough to the NET.

In 2007, we were 38th in KP before the tournament started. I remember there was a lot of debate about us vs Arkansas (they were one of the last teams in, 12 seed) was one spot behind us. There were 11 teams who made it as at large teams who were lower than us in the final ken pom. Additionally, Gonzaga was an AQ as a 10 seed, they were also behind us. UVA was 16 spots lower than us, and they were a 4 seed! So I would think yes, we make it in 2007 probably with room to spare. (Also, there were 6 teams ahead of us who didnt make it in the field)

2008, we were 46th in the final KP before the tournament. There were 5 teams ahead of us who also didnt make it. There were 6 teams below us in the rankings who made it as at large teams. (I initially had 5 here, I had to keep scrolling to add Kentucky, who made it as an 11 and were 37 spots behind us). This year I think is closer than 2007.

2002: 60th in Ken Pom, there were 16 teams ahead of us who didnt make the field. There were 2 teams behind us who made it as at large teams. St Johns made it at 77, which isnt all that much better than where we rank for the 23-24 season.

 

[Cstate]

Hopefully all the ACC coaches speak out about it!

 

[orangenauburn]