Another Win. Another Drop in the NET?

[djcon57]

I have attempted to defend my assertions with hard data on here before. See the Kadary thread, if you care to. Even presented with hard data that satisfied every stipulation these posters asked, they would not budge an inch. This is just one example. Such people cannot be dealt with, and why would I attempt to, at this point?

I'm not complaining, here. I am simply defending against your characterization of my posting which is really off-the-wall brother

There's a Kadary thread on this board in 2024?

 

[Wrinkle]

There's a Kadary thread on this board in 2024?

Special player it still hurts

 

[Niastri]

I have attempted to defend my assertions with hard data on here before. See the Kadary thread, if you care to. Even presented with hard data that satisfied every stipulation these posters asked, they would not budge an inch. This is just one example. Such people cannot be dealt with, and why would I attempt to, at this point?

I'm not complaining, here. I am simply defending against your characterization of my posting which is really off-the-wall brother

Much better

 

[Wrinkle]

Much better

Much better? I didn't do anything differently. I've been speaking civilly with you even as you constantly condescend and call me a child... This board...

 

[jncuse]

AGAIN...name the actual teams beaten. You are missing my point, it seems

MWC's 6 Q1 wins:
SD State - St Mary's (N), Gonzaga (A)
Boise St - St Mary's (N)
Colorado St - Creighton (N)
Nevada - Washington (A) ...WASHINGTON!!!, TCU (N)

Whoop-de-dam-doo.

I get the numbers. I'm taking about the context behind the numbers.

1 of the 6 wins is a win over a terrible Washington team
1 of 10 ACC wins is Clemson winning at Alabama

Yet they are both "Q1" wins

And you are missing my point. My point is that as a whole - be it Q1, Q2, Q3, Q4 games, the ACC and MWC performed at similar levels and had similar win %'w, which is why the MWC moved forward in conference play with some good NETS. The ACC had a few more big wins, in many more games. It does nothing to prove they are better from my perspective.

Of course the ACC is going to get more chances at big wins OOC - its one of the advantages the top conferences have.

I don't think this disagreement will get anywhere, because neither will convince the other!

But you did show you are not happy with the MWC!

 

[djcon57]

And you are missing my point. My point is that as a whole - be it Q1, Q2, Q3, Q4 games, the ACC and MWC performed at similar levels and had similar win %'w, which is why the MWC moved forward in conference play with some good NETS. The ACC had a few more big wins, in many more games. It does nothing to prove they are better from my perspective.

Of course the ACC is going to get more chances at big wins OOC - its one of the advantages the top conferences have.

I don't think this disagreement will get anywhere, because neither will convince the other!

But you did show you are not happy with the MWC!

Wasn't about convincing, was a friendly debate on my end. I'm cool with differences of opinion.

I don't buy the "advantages of top conference" stuff in 2024, either. San Diego State was in a Final 4 last year. Mix in a Pac 12 team

 

[atlanticuse]

We've won a lot of close games, only had a very few blowout wins. But every time we lose we get blown out. The bad overall margin of victory shows up in our key metrics that are used to predict win-loss records.

Our actual record is significantly better than our predicted record says we should be.

They call this effect "luck"

As noted in my other post, our win margin is much less than our loss margin.
Yes Wake is getting stronger, but 29, 36, and other point losses have hurt mightily.

 

[jncuse]

Wasn't about convincing, was a friendly debate on my end. I'm cool with differences of opinion.

I don't buy the "advantages of top conference" stuff in 2024, either. San Diego State was in a Final 4 last year. Mix in a Pac 12 team

We look at the same data differently. Which makes sense, since there is not one number or piece of data that dictates all.

 

[CbusCuse]

It's a bit of a holdover from baseball SABR metrics.

In baseball, everything can be boiled down to single events, in a way no other sport can.

With sophisticated ball teaching tech, even things like spin on the ball is being quantified.

Basically, if you have a teams pitching results and hitting results, you have their run differential. How many runs they score and how many they allow. If you have those, you can predict their record almost exactly over a full season. Yet, sometimes teams are significantly better or worse than their predicted record. While some teams are aberrations, 95% of records are within a few games of the prediction. Once in a while, a team is far outside this predicted range.

If you can't determine why this is, you call it luck.

Syracuse has far more wins than they should when looking at how many points we score vs. how many we give up.

The obvious answer is "who cares? We beat who we beat!"

Apparently, huge amount of weight to margin of victory and these efficiency metrics when calculating the NET.

Every year, some teams are given preferential treatment by the NET, and some are screwed over.

The NET has always sucked this way, we just weren't good enough to have to worry about it. This year, we have to worry about it, and simultaneously, WE are the team getting screwed over. Historically. I imagine in the history of the net no team has been ranked as far below their actual quality as Syracuse is right now.

With our current win streak, ever since beating UNC, we haven't moved up at all. We were probably slightly under ranked a week or two ago, but now it's getting ridiculous.

It would be humorous if it didn't matter so much.

Which makes sense if your dealing with determining predictions like gamblers are. But this is a system that is not supposed to predict, but quantify and compare the actual results, aka Wins and Losses that occurred in relation to how good the opponents were. Not what should have occurred based on the raw data of scoring differentials. So the committee has to make exceptions for the NET. The NET may be great if you want to predict what we will do moving forward. But it doesn't give a fare take on what actually happened results wise regarding W/L. They need to drastically reduce the factor of score differential in the results. Its flawed.

 

[kingtidge]

Our NET is baffling...

I think the weighting on the point margin on wins vs. losses is higher than you'd expect.

I am assuming we want to be 70 or under in NET when all is said and done to make our bubble case (assuming we get to 22 or 23 wins...)

So how do we stack up against a sub-70 team currently??

Let's take a look at Maryland at 67...

For the life of me, I can't understand how they are there and we're at 82.

Let's compare...

 

[djcon57]

Our NET is baffling...

I think the weighting on the point margin on wins vs. losses is higher than you'd expect.

I am assuming we want to be 70 or under in NET when all is said and done to make our bubble case (assuming we get to 22 wins...)

So how do we stack up against a sub-70 team currently??

Let's take a look at Maryland at 67...

For the life of me, I can't understand how they are there and we're at 82.

Let's compare...

Maryland is ranked 6 in AdjD. Look at results like 73-51 over Nebraska. Right in the NET sweet spot.

 

[kingtidge]

Maryland is ranked 6 in AdjD. Look at results like 73-51 over Nebraska. Right in the NET sweet spot.

They're 2 games over 500 tho...

 

[SUskibum]

Luckily (no pun intended) we aren’t the only team that will end up with whacky numbers, as many pointed out a bunch of teams with very strong NET’s will show to be otherwise weak on a variety of metrics and a few others like us will have weak NET’s but strong across most other metrics. Strong being relative to other bubble teams of course.

Once the committee see a bunch of NET’s that do look indicative, it will underweight (or throw out) that metric anytime it doesn’t seem to align with a variety of other important metrics. Obviously I could be wrong but I have often (not always) observed this dynamic with the committee over time, but don’t see it as much with the talking heads (lunardi etc)

Things I think the Committe will care more about than NET

1. Total winning percentage (for Power 6 teams) (this will largely be baked in by the time you count up the rest of these points)
2. Wins against any top 20-50 programs (not NET based)
3. Strength of schedule (not NET based)
4. Wins on the road vs programs that are decent (top 100ish, not NET based)
5. Winning percentage in Conference (for Power 6 teams)
6. Bad/weak losses (to teams below 150ish)

 

[djcon57]

They're 2 games over 500 tho...

Right, I'm explaining why, not that I agree.

Why they are ahead of us in NET is they have great AdjD numbers and same amount of Q1 and Q2 of wins.

 

[Niastri]

Our NET is baffling...

I think the weighting on the point margin on wins vs. losses is higher than you'd expect.

I am assuming we want to be 70 or under in NET when all is said and done to make our bubble case (assuming we get to 22 or 23 wins...)

So how do we stack up against a sub-70 team currently??

Let's take a look at Maryland at 67...

For the life of me, I can't understand how they are there and we're at 82.

Let's compare...

Just... Wow!

It's like the NET has some secret "special sauce" that makes actual game results wins and losses irrelevant.

 

[pokercuse08]

Regarding the luck factor, perhaps this poker analogy will help. If you get it all-in with AA vs 22, you're 82%. So if you win the pot, you actually got "lucky" to win 100%. Or the better way to put it, you outperformed your expected pot share. If you got it in five times in that scenario, you'd expect to lose (roughly) one pot.

In sports, if you go into a game with a 75% chance to win and do win, you got "lucky" by .25 wins. If you play four such games and win all four, you got "lucky" by a full win.

Likewise if you have three 50-50 games, you're going to get lucky or unlucky because you cannot possibly split them 1.5 to 1.5.

Now, where the efficiency numbers can be misleading is that a computer is going to look at a basket of games that were +/- 2 points, and say you should win roughly half of them. So it's going to count a game that was close the whole way that finished within 2 points the same as one where you had a 20 point lead and hung on by 2. By the eye test, those wins are not equal.

So a team that has a bunch of the second type of games is going to look "lucky" on KenPom. It may mean they actually were! Maybe they just caught some lucky bounces this year in close games, the calls went their way, etc. That's going to happen to a handful of teams. But maybe the data is just not giving a good representation.

I'd argue in our case the data is not giving a good representation.

 

[Red2003]

Luckily (no pun intended) we aren’t the only team that will end up with whacky numbers, as many pointed out a bunch of teams with very strong NET’s will show to be otherwise weak on a variety of metrics and a few others like us will have weak NET’s but strong across most other metrics. Strong being relative to other bubble teams of course.

Once the committee see a bunch of NET’s that do look indicative, it will underweight (or throw out) that metric anytime it doesn’t seem to align with a variety of other important metrics. Obviously I could be wrong but I have often (not always) observed this dynamic with the committee over time, but don’t see it as much with the talking heads (lunardi etc)

Things I think the Committe will care more about than NET

1. Total winning percentage (for Power 6 teams) (this will largely be baked in by the time you count up the rest of these points)
2. Wins against any top 20-50 programs (not NET based)
3. Strength of schedule (not NET based)
4. Wins on the road vs programs that are decent (top 100ish, not NET based)
5. Winning percentage in Conference (for Power 6 teams)
6. Bad/weak losses (to teams below 150ish)

solid post. They’ve always said the Net is used for sorting purposes. The Net puts you in the room and in the conversation with Committee.

All your points then get brought up. I would add as #7 what have you done lately. No way a Virginia team should be making the tourney if they can’t score over 50 points and close the season like 2-5 even if they have 22 wins.

 

[atlanticuse]

Our NET is baffling...

I think the weighting on the point margin on wins vs. losses is higher than you'd expect.

I am assuming we want to be 70 or under in NET when all is said and done to make our bubble case (assuming we get to 22 or 23 wins...)

So how do we stack up against a sub-70 team currently??

Let's take a look at Maryland at 67...

For the life of me, I can't understand how they are there and we're at 82.

Let's compare...

I hear you.
But again, non-scientific - their win margin is nearly 13. Loss margin around 8.
(we're upside down - win by 11.89; lose by 17.8)
Couple overtime games. Purdue and UCLA help.

 

[atlanticuse]

Regarding the luck factor, perhaps this poker analogy will help. If you get it all-in with AA vs 22, you're 82%. So if you win the pot, you actually got "lucky" to win 100%. Or the better way to put it, you outperformed your expected pot share. If you got it in five times in that scenario, you'd expect to lose (roughly) one pot.

In sports, if you go into a game with a 75% chance to win and do win, you got "lucky" by .25 wins. If you play four such games and win all four, you got "lucky" by a full win.

Likewise if you have three 50-50 games, you're going to get lucky or unlucky because you cannot possibly split them 1.5 to 1.5.

Now, where the efficiency numbers can be misleading is that a computer is going to look at a basket of games that were +/- 2 points, and say you should win roughly half of them. So it's going to count a game that was close the whole way that finished within 2 points the same as one where you had a 20 point lead and hung on by 2. By the eye test, those wins are not equal.

So a team that has a bunch of the second type of games is going to look "lucky" on KenPom. It may mean they actually were! Maybe they just caught some lucky bounces this year in close games, the calls went their way, etc. That's going to happen to a handful of teams. But maybe the data is just not giving a good representation.

I'd argue in our case the data is not giving a good representation.

The reality of "luck" is that we've got a lot of misses right at the basket. If more of those went in, not sure it would move the needle a lot, but definitely could have swung more games in our favor or further in our favor.

 

[FireballPhil]

Wasn't about convincing, was a friendly debate on my end. I'm cool with differences of opinion.

I don't buy the "advantages of top conference" stuff in 2024, either. San Diego State was in a Final 4 last year. Mix in a Pac 12 team

There is always going to be advantages. It doesn't matter the metric either, there are and always will be built in biases. At the start of the year everyone is 0-0, yet Vegas still sets odds on the game, and there are mythical rankings. These are pulled from preexisting bias. NC is always going to be favored. Arizona. Kansas. Kentucky. We all know the names. The media feeds these biases-it fits their narrative. Metrics are now skewed. But reality is we need something. The net is stupid because of the preseason flaws, but the overall theory behind it is solid. There are opinions on how to fix it, I just don't know if there is truly one way to get past the bias. Maybe some combination of last 10 games of the prior season with some adjustments made based on personnel movement. A team that has gelled and returns their core group should be expected at the start of the next season to be leaders of the pack. Higher turnover teams should be expected to start slower. And then play their way into contention, much like our team is doing this year. Who knows, but there has to be a better way.

 

[atlanticuse]

Much better? I didn't do anything differently. I've been speaking civilly with you even as you constantly condescend and call me a child... This board...

At least you ain’t shadow banned yet.

 

[jncuse]

There is always going to be advantages. It doesn't matter the metric either, there are and always will be built in biases. At the start of the year everyone is 0-0, yet Vegas still sets odds on the game, and there are mythical rankings. These are pulled from preexisting bias. NC is always going to be favored. Arizona. Kansas. Kentucky. We all know the names. The media feeds these biases-it fits their narrative. Metrics are now skewed. But reality is we need something. The net is stupid because of the preseason flaws, but the overall theory behind it is solid. There are opinions on how to fix it, I just don't know if there is truly one way to get past the bias. Maybe some combination of last 10 games of the prior season with some adjustments made based on personnel movement. A team that has gelled and returns their core group should be expected at the start of the next season to be leaders of the pack. Higher turnover teams should be expected to start slower. And then play their way into contention, much like our team is doing this year. Who knows, but there has to be a better way.

February 27th or year end NET or KP**, isn't impacted by preseason rankings. They are calculated purely based on data in the current year.

Your win over a team is not set in stone at a point of time. Say you play someone on November 15th, the win will initially be measured but he value of that win constantly changes over the course of the year based on how that other team is doing.

NET only starts to get published in late November. It can't be properly run until that date.

** KP does run preseason rankings, which he layers on top of his system that uses current year data. The weight of those preseason rankings continues to drop until early January, until it has no impact on the rankings.

 

[FireballPhil]

February 27th or year end NET or KP**, isn't impacted by preseason rankings. They are calculated purely based on data in the current year.

Your win over a team is not set in stone at a point of time. Say you play someone on November 15th, the win will initially be measured but he value of that win constantly changes over the course of the year based on how that other team is doing.

NET only starts to get published in late November. It can't be properly run until that date.

** KP does run preseason rankings, which he layers on top of his system that uses current year data. The weight of those preseason rankings continues to drop until early January, until it has no impact on the rankings.

You misunderstand. Or are ignoring the obvious. Net, or any ranking is 100% affected by early rankings or biases. Who determines what quad teams start in? A game on Dec 22 can be a quad one vs quad 3. Why? Who determined in December that say NC is a quad one in Dec and their opponent, say Seton Hall is a quad three? The built in bias did that. Now we have late February and we are looking at a net all jacked up because the quads were not set correctly in December. Not all teams in quad one should have them. Not all quad three teams should be quad three, some should probably be at least a two. Even within the quad there are issues. Should a team really be 73rd when all metrics appear and show they should be 56th? Every single metric is set by preseason bias. Every metric tries to iron out its flaws through the season. None are perfect, none will ever truly be. Even come November, what information do they use to run the net? Preseason tournaments? If that was the case why isn't the ACC as a conference and thus the teams in it rated higher than the Big 12, since it went 9-3 against him? If the results of the tournaments were held against them, you would never get marquee match-ups because why join a tournament for if you lose you are going to be deemed starting in Q3 and now have a massive hill to climb? And again, how does Vegas set the odds? Who is to say at the start of the year that Kansas should be favored by 15 over Arkansas State? Bias has a huge impact on initial seedings which carries over throughout the year. The goal is to level out and diminish that impact, but the system, as designed, is to flawed to do that.

 

[Niastri]

Much better? I didn't do anything differently. I've been speaking civilly with you even as you constantly condescend and call me a child... This board...

I am sorry I came at you so hard yesterday. I was unusually unkind in my words. My apologies.

On the other hand, there is specific behavior that I (and other posters) have asked you to modify.

As I said, I think you have potential to be a good poster, or I wouldn't have responded.

We have an ignore button for a reason

 

[pokercuse08]

You misunderstand. Or are ignoring the obvious. Net, or any ranking is 100% affected by early rankings or biases. Who determines what quad teams start in? A game on Dec 22 can be a quad one vs quad 3. Why? Who determined in December that say NC is a quad one in Dec and their opponent, say Seton Hall is a quad three? The built in bias did that. Now we have late February and we are looking at a net all jacked up because the quads were not set correctly in December. Not all teams in quad one should have them. Not all quad three teams should be quad three, some should probably be at least a two. Even within the quad there are issues. Should a team really be 73rd when all metrics appear and show they should be 56th? Every single metric is set by preseason bias. Every metric tries to iron out its flaws through the season. None are perfect, none will ever truly be. Even come November, what information do they use to run the net? Preseason tournaments? If that was the case why isn't the ACC as a conference and thus the teams in it rated higher than the Big 12, since it went 9-3 against him? If the results of the tournaments were held against them, you would never get marquee match-ups because why join a tournament for if you lose you are going to be deemed starting in Q3 and now have a massive hill to climb? And again, how does Vegas set the odds? Who is to say at the start of the year that Kansas should be favored by 15 over Arkansas State? Bias has a huge impact on initial seedings which carries over throughout the year. The goal is to level out and diminish that impact, but the system, as designed, is to flawed to do that.

The point is that wins can change quadrants over the course of the year. We have a few games on the precipice of moving up a quadrant. So early on, the NET is reliant on the early season data because that's all it has. By the end of the year, everything is weighted evenly.

This is the reason you'll see bigger swings after results earlier in the year. Each game makes up a bigger chunk of the data.