ADVERTISEMENT

GAME 11 WAKE FOREST: Just what the doctored ordered

Turns out the raw “margin of victory” item is not used in the NET at all, not even up to 10 points. I took this from the NCAA website’s explanation of the NET:

Is there any notable data not included in the NET?​

Game date and game order were not included in the NET rankings so a team's first game counts the same as its 30th.

With the changes announced in May 2020, the NET will no longer use winning percentage, adjusted winning percentage and scoring margin. The change was made after the committee consulted with Google Cloud Professional Services, which worked with the NCAA to develop the original NET.
 
Turns out the raw “margin of victory” item is not used in the NET at all, not even up to 10 points. I took this from the NCAA website’s explanation of the NET:

Is there any notable data not included in the NET?​

Game date and game order were not included in the NET rankings so a team's first game counts the same as its 30th.

With the changes announced in May 2020, the NET will no longer use winning percentage, adjusted winning percentage and scoring margin. The change was made after the committee consulted with Google Cloud Professional Services, which worked with the NCAA to develop the original NET.
But is this still true?
While the NET may cap its measurement of victory margin at 10 points, its use of teams’ raw per-possession efficiency margins as a criterium means that even when a game’s outcome is not in doubt, a team has incentive to either run up the score or continue pushing for buckets.
 
Defensive and offensive efficiency is code name for scoring margin

Pretty much - which makes me wonder the weighting of different things. If you're already including off/deff efficiency, the scoring margin is almost like double-counting... so it makes sense that it's capped.

Turns out the raw “margin of victory” item is not used in the NET at all, not even up to 10 points. I took this from the NCAA website’s explanation of the NET:

Is there any notable data not included in the NET?​

Game date and game order were not included in the NET rankings so a team's first game counts the same as its 30th.

With the changes announced in May 2020, the NET will no longer use winning percentage, adjusted winning percentage and scoring margin. The change was made after the committee consulted with Google Cloud Professional Services, which worked with the NCAA to develop the original NET.
The original NET was a strange grab bag of things and the MOV it was using that was capped at 10 points was apparently UNADJUSTED margin of victory. Ditto the winning percentage, it was just a raw input that didn't adjust for SOS which is insane.

Thankfully they got rid of that and now it's just adjusted efficiency. Which is just tempo adjusted and opponent adjusted margin of victory. But it's uncapped.

It's a problem either way because the current methodology necessitates running up the score on cupcakes to keep your NET afloat. But capping it would make games against cupcakes automatically hurt your NET which is also bad. This is why, in my opinion, this kind of rating is bad to use for tournament selection.

The proper approach is something like Bart's wins above bubble or ESPNs strength of record. You use the full adjusted efficiency metrics to properly measure strength of schedule but then you compute a rating that only considers wins and losses against that schedule.
 
But is this still true?
While the NET may cap its measurement of victory margin at 10 points, its use of teams’ raw per-possession efficiency margins as a criterium means that even when a game’s outcome is not in doubt, a team has incentive to either run up the score or continue pushing for buckets.
The article I read is dated December 5, 2022.

Also, margin of victory is not included AT ALL, not even up to 10 points.

But I agree with what you and others have said in this thread, that the focus on offensive and defensive efficiency certainly causes margin of victory (or defeat) to be baked into the rankings.
 
  • Like
Reactions: RUChoppin
Good to know - thanks for looking that up!

So, there's no double-counting between off/def efficiency and scoring margin at all - it is just to the benefit of teams to remain efficient through the final whistle and not run the last several minutes with backups.

This isn't great for guys at the end of the bench or walkons - but it may do well for guys in the 8th-10th rotational spots to get more time with the primary rotation guys.
 
The original NET was a strange grab bag of things and the MOV it was using that was capped at 10 points was apparently UNADJUSTED margin of victory. Ditto the winning percentage, it was just a raw input that didn't adjust for SOS which is insane.

Thankfully they got rid of that and now it's just adjusted efficiency. Which is just tempo adjusted and opponent adjusted margin of victory. But it's uncapped.

It's a problem either way because the current methodology necessitates running up the score on cupcakes to keep your NET afloat. But capping it would make games against cupcakes automatically hurt your NET which is also bad. This is why, in my opinion, this kind of rating is bad to use for tournament selection.

The proper approach is something like Bart's wins above bubble or ESPNs strength of record. You use the full adjusted efficiency metrics to properly measure strength of schedule but then you compute a rating that only considers wins and losses against that schedule.
Is that what your model uses?
 
Someone needs to develop an efficiency model that disregards stats recorded once a game is safely in hand. I’m not sure exactly how that could work in terms of cutoffs for garbage time. Or maybe it could even be stats obtained while up by a certain margin?

The bottom line is if team is up by 30 points with 5 minutes to go in the game - data from those last few minutes of play isn’t indicative of anything and yet those minutes are 12.5% of the game and factoring in right now the same as the rest of the game time.
 
  • Like
Reactions: Scangg
Someone needs to develop an efficiency model that disregards stats recorded once a game is safely in hand. I’m not sure exactly how that could work in terms of cutoffs for garbage time. Or maybe it could even be stats obtained while up by a certain margin?

The bottom line is if team is up by 30 points with 5 minutes to go in the game - data from those last few minutes of play isn’t indicative of anything and yet those minutes are 12.5% of the game and factoring in right now the same as the rest of the game time.
I think Bart actually does discard garbage time.
 

T-Rank Methodology Update​


For the first time since 2014, I'm making significant changes to the T-Rank ratings. The output will be the same: adjusted offensive and defensive efficiencies, used to create the "Barthag" pythagorean expectancy. But I've changed how I create the adjusted efficiencies.

Specifically, I'm going to incorporate the "GameScript +/-" stat that I derive from the play-by-play (where available), and I'm going to alter the GameScript stat so that it disregards anything that happens during "garbage time." The resulting ratings will therefore more reflect how well teams actually perform when the outcome is still in question.



Not sure it fully disregards garbage time for everything but it is at least doing something to identify / lower it's effect
 

T-Rank Methodology Update​


For the first time since 2014, I'm making significant changes to the T-Rank ratings. The output will be the same: adjusted offensive and defensive efficiencies, used to create the "Barthag" pythagorean expectancy. But I've changed how I create the adjusted efficiencies.

Specifically, I'm going to incorporate the "GameScript +/-" stat that I derive from the play-by-play (where available), and I'm going to alter the GameScript stat so that it disregards anything that happens during "garbage time." The resulting ratings will therefore more reflect how well teams actually perform when the outcome is still in question.



Not sure it fully disregards garbage time for everything but it is at least doing something to identify / lower it's effect

Interesting. I wonder why NET doesn’t look to do this? Seems to me the current formula promotes poor sportsmanship.
 
  • Like
Reactions: Scangg
Interesting. I wonder why NET doesn’t look to do this? Seems to me the current formula promotes poor sportsmanship.
Exactly. Incentivizing blowouts and poor sportsmanship should not be part of the equation.

I wanted to see Miller Oskar and the walks on end of the 20+ point WF blowout. Instead we see Paul and Caleb coming back into the game. We had all upperclassmen in til basically a minute left
 
  • Like
Reactions: PSAL_Hoops and Groz
Anybody else look at the scoreboard in the 1St half when he had 43 points and thought that was the amount we scored in the entire game vs SHU
After our 2nd basket in 2H I told my son we just beat shu… what a difference 2 games make.
 
What happened to "Just what I needed"....lol

Great write up as always !

My thoughts on game:


- I believe Cliff is hampered by a right leg injury that is limiting his flexibility and ability to jump. You can see all of the padding around his right thigh / leg etc
Up close (both shu and wf games) it looks like Cliff is wearing a brace on his right knee under his legging. He barely jumped for the opening tap against wf.
 
Exactly. Incentivizing blowouts and poor sportsmanship should not be part of the equation.

I wanted to see Miller Oskar and the walks on end of the 20+ point WF blowout. Instead we see Paul and Caleb coming back into the game. We had all upperclassmen in til basically a minute left
Yup. And then to make matters worse - as the season goes on, the incentive for doing this becomes less and less apples to apples for everyone as teams accumulate losses and fall out of the mix for post season consideration.

UofL isn’t getting an At large bid or an NIt invite so up or down 30 at the end of a game there’s no longer incentive for them to keep their starters in. More teams will fall into this category over time and the teams who play them are impacted by that.
 
  • Like
Reactions: fluoxetine
ADVERTISEMENT