Because it's better at it than any humans are.
"Why use a machine that can do billions of calculations per second when I can just eyeball things?" say people who have no understanding of how this stuff works.
OFFICIAL NET Thread - 2022/23
- Thread starter fatsam98
- Start date
"Why use a machine that can do billions of calculations per second when I can just eyeball things?" say people who have no understanding of how this stuff works.
Clemson is 3-0 in one point games and also has a separate 2OT win. It's fair to say they've been lucky and their record is not fully indicative of their true talent. It's also fair when it comes to tournament seeding to say, well, wins are wins.
If selection was left entirely up to the computers, Ohio State would be in right now as an at large, because they are top 30 in NET, kenpon, and Bart. And St. Mary's would be a two-seed
No - but by default your basically saying:
OSU has underperformed so they should be out - but for everyone who beat them, the quality of that win should be based on how talented the “more reliable computer system” says they are and not what they’ve actually done consistently throughout the season. Yes?
How good a team is and their resume are not the same thing! The computer says Ohio State is better than Clemson. The computer is probably right. But we don't want to put the best teams in the tournament; we want to put the most deserving teams in the tournament.
The computers that try to measure "resumes" or "most deserving" don't have Ohio State ahead of Clemson. The two decent ones that try to do this (that I am aware of) are Bart's "Wins Above Bubble" and ESPN's "Strength of Record". Wins above bubble has Clemson #31 and Ohio State #82. Strength of record has Clemson #32 and Ohio State #71.
No, there are ways to use computers to measure resumes too.
No, you continue to misunderstand what these computers actually measure and do. Kenpom, Bart, NET, etc aren't measuring talent, they are measuring results. The same results that you want to measure. They just don't make the difference between a -1 point differential and a +1 point differential infinitely more important than the difference between a +1 and a +3.
Yes, the strength of a win against a team should be measured against how good that team actually is overall, not how they performed in a handful of high leverage possessions throughout the season.
If I average 215 in my bowling league and have a 70-30 record and some other dude averages 205 and has an 80-20 record.. no one is arguing that I should win the league, but I am still obviously the better bowler. And if someone else beats me they should get credit for beating the 215.
Right - and they’d “still” be in even when also using the new quad system tool for assessing their “signature” wins. On paper 5 Q1-Q2 wins (2 quad 1s) ought to be enough with computer numbers like theirs is the quad system accurately bucketed win caliber.
RPI / the old way of looking at resumes, was a much more reliable “weed out” combo tool in this regard. The pretender mid major teams with “good” RPIs of seasons past usually lacked the signature win element for serious at large consideration - not enough RPI top 50 wins.
teams are getting more credit for beating Ohio State then they should
Works for me - helps us for now.
If I recall many of Bracketology don’t have them in now anyway. So it’s a good guy for all in B1G.
NET updated early
www.ncaa.com
RU down just one spot to 21 flipping with Indiana who is now 20
Iowa moves up to 35 from 42
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
RU down just one spot to 21 flipping with Indiana who is now 20
Iowa moves up to 35 from 42
Great news. Weirdly I’m actually less worried about the NET than the B1G standings. We have given away our cushion at the hands of Iowa and MSU and so the margin of error is now very close from 2nd place to 10th. Piranhas biting our anklesNET updated early
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.
RU down just one spot to 21 flipping with Indiana who is now 20
Iowa moves up to 35 from 42
Bowling is an individual sport. It’s not the same thing at all. How disciplined a team plays when nursing a 25 point lead in hoops, or even how a team closes out a blow out loss has very little correlation to the outcome of an evenly matched 50/50 game played on a different day. Efficiency data blends everything together.
But beyond this (and perhaps the most troubling aspect of the new system for me) is that I have been observing first hand how poor of an indicator individual performance against terrible teams is in predicting success against better competition. Believe me - I wish there was more correlation. My 8 year old son routinely drops 20+ in our town rec games. He’s an indisputable star there. Our rec team is undefeated. On our Town Travel team? Not so much. He’s a solid role player for sure - scores a basket or 2 each game but he’s far from an impact player. And yes - almost every kid who plays Travel also plays on a rec team. Bottom line is my kid is awesome at exploiting a terrible defense where 4 of 5 don’t know how to play at all and creating for himself. At the next level he can’t do that - at least not yet. I get that it’s not entirely the same 8U and college but I think the concept probably still applies to a degree.
Oddly I wasn’t that concerned about the loss. Iowa came off a dreadful shooting performance. I expected them to bounce back. We put up a solid scoring effort but our defense is built to limit teams in the paint and force them to beat us from the perimeter. Iowa was able to do that. Game in and out - I don’t think teams are going to be able to drop that many 3s on us (or any team). Our play has been consistent where I feel good about our chances when teams nail 7-8 threes a game.
Last edited:
The individual nature of the sport has nothing to do with the concept being demonstrated there. We've done this twice now sort of, the first time being where I described the end game of literally every close game Rutgers has played this season, said "well what if this small thing changed so we won/lost" and you replied "well no obviously all of THAT was luck but you know, the winner factor shows up other places".
You're correctly identifying the shortcomings and noise in these models, the issue is that
(1) you are waaay overexaggerating how much they matter and
(2) your solution of "throw away a shitload of information" isn't actually an improvement
An improvement in what sense. The shortcomings of the other system resulted in less anomalies on what is actually prioritized for field selection year in and out. At least it sure seems that way after 5 years worth of data.
My understanding from back when was that was supposed to be the point of these new sorting tool enhancements NET/quad vs RPI. The reason they went back to the drawing board in the first place.
But for every Southern Miss type team of the past with an RPI in the 30s that got routinely “snubbed” despite the old system in a vacuum saying they should be “in”, it seems like there are consistently far more teams in the NET/quad system with no chance for at large despite the “paper” credentials.
If the point of the system was to be more accurate as a ranking tool to assess the best teams in order - you’d be right. Rutgers and Illinois depict this. But it never seemed the case that a better RPI influenced getting a 2 seed over a 4 seed. That has always been resume based. RPI or NET - take your pick. 19 vs 38 isn’t that likely to impact where you sit in the bracket.
It’s built like a predictive system so I have to assume this is its purpose. It’s to have a good measure of team quality in order to do the human work of comparing resumes. It’s not to look at the NET as a major component in directly selecting or seeding teams.
Obviously I am not an insider and I don’t know exactly what they are doing. If they are using it directly for selection I disagree with it. I don’t think they are doing this because from looking at the fields and seeding it seems pretty clear they aren’t really.
I’m recalling the articles from 6 years ago. The beef always seemed to be that the RPIs weren’t well correlated with actual selection and mid majors were getting screwed. In effect it was hard for mid-majors to collect the amount of good RPI wins to get selected, but not that hard for them to have a pretty good RPI themselves with a good record.
There also used to be discussion about how those teams had to play more games on the road - which was when they created the quad systems to weight that. Except the new system actually helps major conference teams not mid majors in this regard. If you play mostly Q4s playing on the road doesn’t matter.
The RPI/NET metrics were never used directly for selection. I always thought they were supposed to be a sorting tool to a) provide as good an indicator as possible throughout the season of where teams stand b) quickly weed out teams with no chance at selection
You are both right, partially.
The NCAA for a long time has decided that they need some sort of algorithm to determine how strong teams are, to determine if a win is a good win or mediocre win, or if a loss is a good loss or bad loss. They do this to help eliminate big-name bias (claiming a win against Duke is a good win, just because it is Duke), and also to have a way to objectively look at mid-major schools that don't have as many opportunities to play top schools. Without this, the selection committee would be stuck in never-ending discussions and be unable to complete the bracket before April. Even Bac, when he posts his bracket predictions, references the number of Quad 1 wins or Quad 3/4 losses each school has.
Previously the NCAA used RPI as its sorting algorithm. But this formula was too heavily weighted to strength of schedule (presumably to prevent schools from loading their schedules with cupcakes to boost their W-L record). Also, once you get into conference play, the conference SOS for all conference schools is fairly similar. So your RPI gets baked during OOC play, without a lot of opportunity to move it once you enter conference play.
The NCAA moved to NET as their algorithm to determine strength of each school. NET uses adjusted offensive/defensive efficiency and some secret team value index. Because NET seems to be very closely correlated to other efficiency algorithms like KenPom and Bart Torvik, it seems efficiency is the dominant factor in NET.
Fluox is right that efficiency has shown itself to be a fairly good (though not perfect) predictor of who will win games. That is because if you have a strong offensive efficiency you tend to score more points and if you have a strong defensive efficiency you tend to give up fewer points. And teams that score more points than they give up usually win more games than they lose.
And Bac is also right when he points out how efficiency doesn't care about W-L and winning matters. He is right when he points to too many outliers as examples of how the NET is flawed. We all knew that last year when Rutgers knocked off ranked team after ranked team, and still had a NET in the 70s. No one seriously considered Wisconsin's quad 3 loss to Rutgers to be a bad loss for them, even though that is what the NET said.
The NET is only as good as the assumptions that went into designing it. And there are enough obvious flaws that we've seen last season and this, that I think the assumptions need to be changed. And since the selection committee uses the NET as a tool to evaluate strength of wins and losses, it is important to fix the NET to provide a better tool for that evaluation.
If I were to change the NET, here are some of the changes I would incorporate:
- Only look at efficiency for the first 37 or 38 minutes of the game. Since efficiency doesn't care about W-L, you don't need to go for the entire game. Teams employ different end-game strategies which impact their normal efficiency numbers. Plus it would allow teams to give bench players an opportunity to play without impacting a team's NET ranking.
- Reduce the weighting of efficiency and increase the weighting of W-L. Winning matters. UConn this season has won 3 games since Christmas and still have an unbelievable NET ranking of 7. Sure they have a lot of lopsided wins and close losses, which makes them look efficient. But their actual record does not look like the 7th best team in the country. Maybe they have just been unlucky, or maybe they have a flaw that prevents them from closing out close games. But at the end of the day, actually winning the games has to count.
- Run 2 versions of the NET, NET-A and NET-B, simultaneously with different assumptions and weightings. Sure it will complicate the nitty-gritty sheets. But it will reinforce to the Selection Committee that the rankings are dependent on the assumptions built into them. If a team is ranked 12 in NET-A and 52 in NET-B, that is a reminder that they need to look more closely at the team. Likewise a loss to a Quad 1 team in NET-A that is a Quad 3 loss in NET-B (perhaps like Wisconsin's loss to Rutgers last year) also means closer examination is warranted.
Bubbletology update:
Ironclad 1-bid leagues: 21
No new additions. Saint Louis won the road at Davidson to get to 7-1 in the league to go with non-conference wins over Memphis and Providence. At one point they were 9-6, they're now 15-6. We're monitoring the situation but it's possible they resurrect their resume to at-large quality and we bump the A-10 to the next section.
21 bids
Possible 2-bid leagues: 4
This is the AAC (Houston/Memphis), CAA (Charleston), C-USA (FAU), and WCC (Gonzaga/SMC).
UCF has exited the discussion in the AAC with four straight losses. Temple is 8-2 in the league but they have some terrible losses and got taken to OT by South Florida in between road wins at Houston and UCF. If they can beat Houston at home and Memphis on the road then their bizarre resume will get another look, but for now this is a two bid league.
Charleston lost at home to Hofstra but I'm not quitting on them yet. If they went 29-2 and lost in the conference tournament to Hofstra again that would only be a Q2 loss. One more regular season loss of any kind will do them in, though. We will charitably allow for a bid stealer here and call it seven.
7 bids
That leaves 40 up for grabs among the top seven conferences. Eyeballing it (2.5 basically means they’ll get either 2 or 3 bids):
MWC: 4 (San Diego State, Utah State, Boise State, New Mexico, Nevada)
Not much new to report here with five teams in the NET top 50. Utah State's resume (0-3 in Q1 with two Q4 losses) is still too dinged-up to bump this league to 4.5. They get New Mexico at home this week. Nevada has alternated wins and losses in their last six, including dropping one at UNLV. They find themselves as the second to last team in the field per Bracket Matrix, so they could use the Q1 win.
Pac-12: 3.5 (UCLA, Arizona, USC, Arizona State, Utah)
USC surges to the third spot as Arizona State has now lost four in a row. The Sun Devils are better in Q1/Q2 than Utah, but they also have a Q4 loss. Oregon has a run of three Q1 games in their next four, so they can enter the picture.
Big East: 5 (UConn, Creighton, Providence, Marquette, Xavier)
The Big East continues to be boring. Seton Hall is lurking but their next two games (at St. John's, vs. DePaul) can only hurt, not help.
ACC: 6.5 (Duke, Virginia, UNC, Miami, Clemson, Pitt, NC State, Wake Forest)
I'm late bumping them to 6.5. The first seven above are all in right now -- Pitt is the first team out at Bracket Matrix but not all have updated to reflect their wins and Arizona State's losses. A potential last hurrah for Wake Forest on Tuesday as they play at Duke.
Big 12: 6.5 (Kansas, Iowa State, Kansas State, Texas, TCU, Baylor, pick one of the next 3)
Sticking with 6.5 even though everyone who needed an SEC Challenge win -- West Virginia, Oklahoma, and Oklahoma State -- got one. This week brings us Oklahoma at Oklahoma State, then the Sooners are on the road at West Virginia. TCU could be the real MVP here as they host WVU and then are at Oklahoma State. If the Horned Frogs go 0-2 then this gets bumped to seven.
SEC: 6.5 (Tennessee, Alabama, Auburn, Arkansas, Missouri, Kentucky, Texas A&M, Mississippi State)
Bumping to 6.5. Somehow the SEC also got wins it needed on Saturday with Missouri and Mississippi State, plus Texas A&M surviving Vanderbilt.
Big Ten: 8.5 (Everyone except Nebraska and Minnesota)
Close to lowering this one to eight. Michigan and Ohio State are 11-10, and Wisconsin has lost 6 of 7. But there's enough elasticity elsewhere to still get to nine.
Total: 40 bids
Currently projecting two of the leagues at x.5 to go OVER and three to go UNDER.
Ironclad 1-bid leagues: 21
No new additions. Saint Louis won the road at Davidson to get to 7-1 in the league to go with non-conference wins over Memphis and Providence. At one point they were 9-6, they're now 15-6. We're monitoring the situation but it's possible they resurrect their resume to at-large quality and we bump the A-10 to the next section.
21 bids
Possible 2-bid leagues: 4
This is the AAC (Houston/Memphis), CAA (Charleston), C-USA (FAU), and WCC (Gonzaga/SMC).
UCF has exited the discussion in the AAC with four straight losses. Temple is 8-2 in the league but they have some terrible losses and got taken to OT by South Florida in between road wins at Houston and UCF. If they can beat Houston at home and Memphis on the road then their bizarre resume will get another look, but for now this is a two bid league.
Charleston lost at home to Hofstra but I'm not quitting on them yet. If they went 29-2 and lost in the conference tournament to Hofstra again that would only be a Q2 loss. One more regular season loss of any kind will do them in, though. We will charitably allow for a bid stealer here and call it seven.
7 bids
That leaves 40 up for grabs among the top seven conferences. Eyeballing it (2.5 basically means they’ll get either 2 or 3 bids):
MWC: 4 (San Diego State, Utah State, Boise State, New Mexico, Nevada)
Not much new to report here with five teams in the NET top 50. Utah State's resume (0-3 in Q1 with two Q4 losses) is still too dinged-up to bump this league to 4.5. They get New Mexico at home this week. Nevada has alternated wins and losses in their last six, including dropping one at UNLV. They find themselves as the second to last team in the field per Bracket Matrix, so they could use the Q1 win.
Pac-12: 3.5 (UCLA, Arizona, USC, Arizona State, Utah)
USC surges to the third spot as Arizona State has now lost four in a row. The Sun Devils are better in Q1/Q2 than Utah, but they also have a Q4 loss. Oregon has a run of three Q1 games in their next four, so they can enter the picture.
Big East: 5 (UConn, Creighton, Providence, Marquette, Xavier)
The Big East continues to be boring. Seton Hall is lurking but their next two games (at St. John's, vs. DePaul) can only hurt, not help.
ACC: 6.5 (Duke, Virginia, UNC, Miami, Clemson, Pitt, NC State, Wake Forest)
I'm late bumping them to 6.5. The first seven above are all in right now -- Pitt is the first team out at Bracket Matrix but not all have updated to reflect their wins and Arizona State's losses. A potential last hurrah for Wake Forest on Tuesday as they play at Duke.
Big 12: 6.5 (Kansas, Iowa State, Kansas State, Texas, TCU, Baylor, pick one of the next 3)
Sticking with 6.5 even though everyone who needed an SEC Challenge win -- West Virginia, Oklahoma, and Oklahoma State -- got one. This week brings us Oklahoma at Oklahoma State, then the Sooners are on the road at West Virginia. TCU could be the real MVP here as they host WVU and then are at Oklahoma State. If the Horned Frogs go 0-2 then this gets bumped to seven.
SEC: 6.5 (Tennessee, Alabama, Auburn, Arkansas, Missouri, Kentucky, Texas A&M, Mississippi State)
Bumping to 6.5. Somehow the SEC also got wins it needed on Saturday with Missouri and Mississippi State, plus Texas A&M surviving Vanderbilt.
Big Ten: 8.5 (Everyone except Nebraska and Minnesota)
Close to lowering this one to eight. Michigan and Ohio State are 11-10, and Wisconsin has lost 6 of 7. But there's enough elasticity elsewhere to still get to nine.
Total: 40 bids
Currently projecting two of the leagues at x.5 to go OVER and three to go UNDER.
Feel like it's trending to 8 for the B1G (and more for SEC and Big12).
Can't see Michigan, Ohio State or Wiscy getting in at this point based on how they've all looked recently. Michigan's remaining schedule is brutal, Ohio's is tough as well.
Think it ends up as Purdue, Rutgers, Sparty, Iowa, Illinois, Indiana... And two of Maryland, Northwestern and Penn State. None of those last 3 have awful remaining schedules.
This would be my guess as well. There's precedent for OSU/Mich/Wiscy though: 2020 Wisconsin.
They lost non-conference games to New Mexico and Richmond and were 13-10 (6-6) on February 6. They reeled off eight straight wins to finish 21-10 (14-6) and were a projected 4 seed before covid happened.
Not going to update the schedule rankings list today, like I usually do after a game. In bed with covid and a 102.4 fever.
The teams which gave all three conference loses to Northwestern appear on their remaining schedule. Rutgers gets them the last regular season game late on that Sunday.
Last edited:
Keep chopping against that virus! Hope you feel better soon.
haha they were favorites actually...it wont move them up that much, maybe 10 spots and they still will have to win at least 5 of their next 9 conference games to realistically have a shot at a bid
its not their net that is their issue, its that they are now 1-9 in Q1 and 1-10 in Q1/2
They got a gift on a blatant hack that wasn't called, play ended up forcing the turnover that led to TT tying it up at the end of regulation. I didn't see much of the game, but we would have been livid if it happened against us. Kind of like officials missing a guy coming from out of bounds to hit a game winner.
Feel better.
Just over a bout and still with a bit of fatigue and hard to concentrate on work.
Good luck
Some good ideas in here Upstream, well done.
Two suggestions I have for the selection committee:
1) Consider evaluating teams only on their wins.
2) Consider (again) weighting recent performance.
On the first point, the parity that started evolving with the 3-point shot has now been enhanced by the transfer portal. And in any case all losses are bad because the object of the game is to win. There are plenty of stats that will fairly compare team performance and relative strength if only wins are utilized, along with strength of schedule.
On the second point, I’ll use a player analogy: come tournament time, as a head coach I’m gonna play the bench guys who have been outplaying the other bench guys in the most recent 7 to 10 games, because they give us the best chance to win and therefore make us a better tournament team. So, recent performance should count for something, imo.
Last edited:
Feeling a lot better, but that may be the Theraflu talking... fingers crossed.
RU is now #22
Results:
Q1 (1-30 Home, 1-50 Neutral, 1-75 Away)
3 - @Purdue (W)
20 - Indiana (W)
29 - @Ohio St (L*)
29 - Ohio St (W)
38 - @Iowa (L)
40 - @Northwestern (W)
47 - @Miami (L)
48 - @MSU (L)
Q2 (31-75 Home 51-100 Neutral, 76-135 Away)
36 - Iowa (L)
38 - Maryland (W)
53 - Penn St (W)
57 - Seton Hall (L)
72 - Wake Forest (W)
Q3 (76-160 Home, 101-200 Neutral, 136-240 Away)
115 - (N) Temple (L)
151 - UMass-Lowell (W)
Q4 (161+ Home, 201+Neutral, 241+ Away)
206 - Rider (W)
307 - Bucknell (W)
325 - Sacred Heart (W)
328 - Coppin St (W)
333 - Central CT St (W)
341 - Columbia (W)
Upcoming
Q1 (1-30 Home, 1-50 Neutral, 1-75 Away)
20 - @Indiana
26 - @Illinois
48 - (N)MSU
58 - @Penn St
74 - @Wisconsin
Q2 (31-75 Home, 51-100 Neutral, 76-135 Away)
40 - Northwestern
Q3 (76-160 Home, 101-200 Neutral, 136-240 Away)
83 - Michigan
99 - Nebraska
220 - @Minnesota
Q4 (161+ Home, 201+Neutral, 241+ Away)
220 - Minnesota
Notes:
- Ohio State still inexplicably a Q1 win
- Temple is now healthy and may be making a push for Q2?
- UMass-Lowell is diving toward Q4
- Bucknell dips down below 300, making 5 teams with 300+ NET
RU is now #22
Results:
Q1 (1-30 Home, 1-50 Neutral, 1-75 Away)
3 - @Purdue (W)
20 - Indiana (W)
29 - @Ohio St (L*)
29 - Ohio St (W)
38 - @Iowa (L)
40 - @Northwestern (W)
47 - @Miami (L)
48 - @MSU (L)
Q2 (31-75 Home 51-100 Neutral, 76-135 Away)
36 - Iowa (L)
38 - Maryland (W)
53 - Penn St (W)
57 - Seton Hall (L)
72 - Wake Forest (W)
Q3 (76-160 Home, 101-200 Neutral, 136-240 Away)
115 - (N) Temple (L)
151 - UMass-Lowell (W)
Q4 (161+ Home, 201+Neutral, 241+ Away)
206 - Rider (W)
307 - Bucknell (W)
325 - Sacred Heart (W)
328 - Coppin St (W)
333 - Central CT St (W)
341 - Columbia (W)
Upcoming
Q1 (1-30 Home, 1-50 Neutral, 1-75 Away)
20 - @Indiana
26 - @Illinois
48 - (N)MSU
58 - @Penn St
74 - @Wisconsin
Q2 (31-75 Home, 51-100 Neutral, 76-135 Away)
40 - Northwestern
Q3 (76-160 Home, 101-200 Neutral, 136-240 Away)
83 - Michigan
99 - Nebraska
220 - @Minnesota
Q4 (161+ Home, 201+Neutral, 241+ Away)
220 - Minnesota
Notes:
- Ohio State still inexplicably a Q1 win
- Temple is now healthy and may be making a push for Q2?
- UMass-Lowell is diving toward Q4
- Bucknell dips down below 300, making 5 teams with 300+ NET
