Interesting comments at the end of the Nebraska - Ohio State game last night about how the NIT tourney automatically takes the top two remaining B1G teams based on NET rankings. That would be Ohio State and Iowa as of today, assuming the six teams above them go to the NCAA tourney.
2023-24 NET ratings thread
- Thread starter Styre
- Start date
NU did not dig themselves a hole because they lost to CSU, they dug themselves a hole because they did not beat them by 20. If they won the game by 1, it wouldn’t move the Net much, if at all. The fact that the Net doesn’t factor in who actually won the game (you know, the whole point of playing), makes it borderline worthless.
Again, it's a predictive system. It cares about how you play against the caliber of opponent, it's not supposed to be tracking wins and losses. It is a tool, along with many others, to be used in assessing a team's quality.
I think the point many are making is it sucks at what it is supposed to be doing.
Again, I know what the Net is (this might be the longest thread in NU Rivals history). You are the one who said losing to Chicago st was a big problem but in reality, who won didn’t matter. I ( and manu others) are pointing out how flawed it is.
Last edited:
For what its worth... Baylor's worst game of the season (25.5 points of underperformance) occurred vs MSU.
Baylor's 2nd worst performance was a 6 point underperformance, losing to Duke by 8 the next time out.
The Baylor/MSU game is a gigantic outlier.
Fortunately for Michigan State, they have 2 more outliers that are almost as helpful. A 44 point blowout of Stony Brook and a 31 point blowout of Penn State. (24.5 point and 22 point outperformances)
Their NET rating is severely inflated because of those 3 games.
But, but, but they beat Stony Brook and Penn State by a lot after the holidays! Let’s crown them!
It's not margin of victory, it's net efficiency. We got a ton of credit in that game for our offensive performance - our second-best grade of the year, in fact - but we lost credit because the defense was awful and we gave up 100+. In fact, by that chart the loss at Purdue helped our rating more than ten of our actual wins, which to me sounds like we got a ton of credit for playing Purdue close.
The NET rankings *do* take into account wins and losses, and the quality of those wins and losses. The predictive element is part of NET, but there's also a descriptive (who have you beat/lost to) element.
KenPom, which is purely a predictive tool, does *not* take wins and losses into account.
I think that would account for why we are 48 in NET but 42 in KenPom (the drag of losing to CSU).
Torvik, which is another solely predictive metric, has us at 34. He uses the same approach as KenPom on most things, but he makes a couple key adjustments:
- Recent games count more than older games
- Blowouts against bad teams are "downgraded" in the system past a certain margin of victory so as to not allow them to have an outsized effect on adjusted efficiency numbers.
One thing I don't know is if NET's predictive part makes any such adjustment for blowouts in mismatches. Clemson's coach was complaining earlier this week that the Big 12 is gaming the NET by blowing out a lot of bad teams early and inflating their rankings so they get more credit for beating each other in the conference season.
One drum I like to keep beating is Strength of Record, which is a descriptive metric the committee takes strongly into account and ranks us very highly (19th as of today!).
I'm still enjoying this thread and am happy it's been going for so long.
Look, I think we can all agree that the NET is "flawed" in one way or another, but it's also possible that NU really is around the 48th best team in the country (which is probably not that different from 38th or 58th), but they know how to win, are a little lucky, or whatever. Some teams (like NU) manage to outplay their rating, while others (like Purdue in the NCAAs) don't. It's just one more stat you can use to paint the picture of the 2023-2024 season.
And in the end, the only metric that really matters to us now is are we in the tourny or not. Hopefully, a worse NET score doesn't come back to haunt us. But if we win 2 more games, I think we're pretty much assured a spot.
Joe Lunardi (I know, I know) made a Twitter post today about "teams better than first glance" which I assume means teams who are better than their NET once you take a closer look at their results. We're named there. I'm hoping the commitee looks beyond NET because that metric is STILL vastly underrating this Cats team.
This whole "efficiency" thing is pretty silly.
Offensive Efficiency = Points You Score / Possessions
Defensive Efficiency = Points You Allow / Possessions
KenPom (and NET) are both entirely dominated by margin of victory.
The OT loss at Purdue does not give NU enough credit because neither KenPom nor NET makes a reasonable adjustment to the points scored in overtime games, which should be represented as something closer to a tie.
Lastly, calling Ken Pom "predictive" is somewhat like looking at the last 20 years of Chicago weather on March 1st, taking the high temperature, averaging those and "predicting" the high this year. KenPom is a backfitting algorithm for a dataset of games that have already been played. It uses averages and treats all games the same.
NU went up 4 spots and their efficiency margin increased from 14.91 to 15.38 after the OT loss to Purdue. How much credit should we have gotten instead? You seem pretty dismissive of a system that has been pretty successful for 20 years and is utilized by countless coaches let alone fans.
Apologies, if some of us are dismissive, it’s merely because it sucks. Just because it’s been around 20 years doesn’t mean it is good.
Would we have been better off beating Chicago State by 30 and losing to Purdue by 1?
From a predictive standpoint? Yes, absolutely. From a strength of record standpoint? Likely no.
Said differently, would you rather play a team that lost to Purdue by 1 at home and beat Chicago State by 30 at home, or a team that lost to Chicago State by 2 and beat Purdue by 4, both at home?
I’d prefer to play the latter, all else equal. They can hang with the best team in the country and if you’re not on your game they will destroy you vs. they have wide swings in outcome and even if you’re a crappy team they’ve shown the ability to crap the bed. Speaks to lapses and composure.
As an opponent,I would prefer to play the team with the lower ceiling, higher floor. NU clearly has a very high ceiling (beat the #1 overall seed once and took them to OT second game, bear a very good Illinois team, smoked a good Nebraska team). It shows that they can play with the best.
A team that can hang with good teams but ultimately can’t get over the hump doesn’t scare anyone.
I was specifically talking about the scenario I shared. And NU’s floor is lower than low. The bottom fell off against one of the worst teams in college basketball. That’s the point.
Taking to a great team to OT and losing by a few points are close to the same from a “they can hang with anyone” standpoint.
NU took Purdue to OT and also beat them. There is a big difference to beating someone and almost beating someone. NU was the king of hanging close with good teams (at home) but ultimately losing. Moral victories don’t count.
Keep winning is what is most important. But in the weeds of our NET numbers heading into today's games:
- Minnesota is still floating at that #75 number (currently #77). Maryland also has fallen to close to #75 (currently #71). On paper, it's nicer to have Quad 1 and 2 wins vs. Quad 2 and 3 wins. But can only control what we can control at this point.
- Speaking of...Iowa is a Quad 2 game today.
I don't know what you mean.
Successful for 20 years? What does that mean? How accurate are KenPom's predictions?
Within 1% 3% 10% ??? The errors are quite large.
This game is a toss-up, but one third of the time either team could win by 15 points?
Thats not predictive.
First there was Sagarin.
He implemented a method to assign each team a rating based on the games that had been played.
Team A's score - Team B's score, with an adjustment for the neutrality of the site.
As simple as it gets. All games weighted equally.
People used those ratings to set point spreads and justify bets on one team or another.
Next came Ken Pomeroy.
He implemented a method to assign each team an offensive and defensive rating using the results of previously played games. This is essentially the same technique Sagarin used, except that it is based on points scored and point allowed, instead of margin.
Everything beyond that for KenPom (and the NET) has been bells and whistles, including using Points/Possession (so-called "Efficiency") instead of Points/Minute. He created something he called "Tempo" to estimate how many possessions each team typically played in a 40 minute game. That enabled him to talk about tempo and justified his use of "Efficiency."
I recognize that Sagarin was better than nothing and had a sort of simple, unbiased beauty to it... and that KenPom was slightly better than Sagarin because you could take a better guess at the actual score, not just the margin.
But it didn't take me long, just a few hours, to create comparable algorithms and start addressing their deficiencies. Again, thanks to Bart Torvik for providing free data via his website.
This stuff is not rocket science and, in my opinion, is biased against winning teams from lesser conferences... by design. So I dismiss the reverence that coaches and fans show to algorithms that are soft and (undoubtedly) lose money consistently to bettors armed with superior information and systems.
As far as determining which teams deserve to be invited to a tournament to declare the national champion, well thats the NCAA's thing - and that is controlled almost entirely by money. They have an interesting habit of putting the top mid-majors in 8/9 games and 7/10 games, often against each other. They do not want those capable teams to advance.
Last edited:
Successful in the sense that he's made a living off of his website and been the inspiration for countless others to try and study basketball in similar ways. Torvik has always credited Pomeroy as the best, btw.
Seems like you should create your own website and charge for it since your model is superior, you must be leaving a lot of money on the table.
Well your first sentence is a little off. If you want to credit anybody, credit Jeff Sagarin. Torvik has been gracious to Ken Pomeroy. Torvik is very collaborative and deserves praise for that alone.. I don't think Pomeroy has been the same toward Sagarin.
The 2nd sentence (which I sort of expected) is missing the point. I think Ken Pomeroy, Bart Torvik or anything I come up with cannot compete with better informed professionals who bet on games. All of these score-based approaches are just too simplistic, even if they are unbiased.
I noticed you didn't respond to any of my criticisms of Ken Pom's ratings, but if his stuff doesn't stand up to scrutiny, we're better off knowing that, right?
I had a professor who developed the system with Jeff Sagarin. I forgot how to do it now, but it was, as PWB says, shockingly simplistic. They were just better at Excel or its predecessors.
“How do you know how good a team is? Just look at their scores.”
He was also working with the Mavs / Cuban on plus-minus and lineup combination stuff in the early/mid 00s.
“How do you know how good a team is? Just look at their scores.”
He was also working with the Mavs / Cuban on plus-minus and lineup combination stuff in the early/mid 00s.
Sure, but I don't have any issues with his ratings. He posts his error margins and bias transparently for anyone to observe, along with how his model performs against expectations. So again, presumably if yours is better, there seems to be a market for people to demonstrate the superiority of their systems, which I would certainly encourage you to explore.
No model can account for injuries that professional gamblers are privy to, such as when the models expected NU to be a 2-3 point dog at Maryland, but the line ended up being 5-6 points because the sharp gamblers knew that Langborg was injured before the public did.
Heading into the final week, NU finds itself staying level at #53.
QUAD 1 (5-5)
H #2 Purdue - W
A #2 Purdue - L
H #15 Illinois - W
A #15 Illinois - L
H #21 Dayton - W
H #22 Michigan State - W
A #23 Wisconsin - L
N #32 Mississippi State - L
A #43 Nebraska - L
A #74 Maryland - W
QUAD 2 (5-3)
H #43 Nebraska - W
H #57 Iowa - L
H #59 Ohio State - W
H #74 Maryland - W
A #77 Minnesota - L
A #91 Rutgers - L
A #94 Penn State - W
A #101 Indiana - W
QUAD 3 (3-0)
H #94 Penn State - W
N #124 Arizona State - W
H #131 Michigan - W
QUAD 4 (7-1)
N #212 Rhode Island - W
H #271 Binghamton - W
H #287 Jackson State - W
H #289 Western Michigan - W
H #300 Chicago State - L
H #301 Northern Illinois - W
A #321 DePaul - W
H #359 Detroit Mercy - W
In the final regular season week, NU has a Quad 1 road game at Michigan State and a Quad 3 home game against Minnesota.
QUAD 1 (5-5)
H #2 Purdue - W
A #2 Purdue - L
H #15 Illinois - W
A #15 Illinois - L
H #21 Dayton - W
H #22 Michigan State - W
A #23 Wisconsin - L
N #32 Mississippi State - L
A #43 Nebraska - L
A #74 Maryland - W
QUAD 2 (5-3)
H #43 Nebraska - W
H #57 Iowa - L
H #59 Ohio State - W
H #74 Maryland - W
A #77 Minnesota - L
A #91 Rutgers - L
A #94 Penn State - W
A #101 Indiana - W
QUAD 3 (3-0)
H #94 Penn State - W
N #124 Arizona State - W
H #131 Michigan - W
QUAD 4 (7-1)
N #212 Rhode Island - W
H #271 Binghamton - W
H #287 Jackson State - W
H #289 Western Michigan - W
H #300 Chicago State - L
H #301 Northern Illinois - W
A #321 DePaul - W
H #359 Detroit Mercy - W
In the final regular season week, NU has a Quad 1 road game at Michigan State and a Quad 3 home game against Minnesota.
The current selection committee is made up of 12 people: 3 conference commissioners and 9 athletic directors.
The conferences represented by commissioners: SWAC (chairperson), Sun Belt, Big Sky
The schools represented by ADs: Iowa St, Oregon St, North Carolina, Minnesota, Alabama, Butler, Samford, Temple, Santa Clara
So it's 6 representatives from power conferences, 6 from mid/low majors.
I'm not really making any sort of argument, just thought it's interesting how it's put together, and that the chair is the commissioner of the weakest conference.
Similar threads
