Adjusted D Down to #41 | Page 5 | The Boneyard

Adjusted D Down to #41

caw said:
Wonder how high a fully healthy team would have been. Have to think top 5 defense.

3rd on offense still.
Expand ...
I posted this in another thread, but we’re #2 on defense and #1 overall on Bart Torvik’s site since Clingan returned on 1/17.

IMG_4343.jpeg


 
NJHusky said:
why that even matters is ridiculous...who cares if you play the #175 v #325 team??
Expand ...
The issue is that we played teams that are considered very bad and then underperformed the expectations. You don’t get credit for beating a really bad team if you couldn’t cover the spread. These models are all stat based and the numbers from those games don’t flatter us compared to what we’ve done versus better opposition.
 
yungdowork said:
The issue is that we played teams that are considered very bad and then underperformed the expectations. You don’t get credit for beating a really bad team if you couldn’t cover the spread. These models are all stat based and the numbers from those games don’t flatter us compared to what we’ve done versus better opposition.
Expand ...
Thx for the info…but that shows it’s a broken model. So beating a bad team by 50 is good but letting your walk ons play and only winning by 35 is bad??? Do they track whose playing ? If not, That’s a model that shows nothing.
 
NJHusky said:
Thx for the info…but that shows it’s a broken model. So beating a bad team by 50 is good but letting your walk ons play and only winning by 35 is bad??? Do they track who’s playing ? If not, That’s a model that shows nothing.
Expand ...
There’s plenty of models that track lineups. Every single model on the planet is flawed, that’s why there are so many of them. If we had a perfect model, we wouldn’t have the rankings determined by sports writers who haven’t watched the majority of the games in a given week.

Mississippi Valley State is 0-22. They’ve lost 12 games by 30+ points. Teams a lot worse than us also beat them with their bench. The bottom line is that we have absolutely nothing to gain by playing teams that are that non-competitive. If we win that game by 100, it doesn’t make our resume look any better.
 
NJHusky said:
Thx for the info…but that shows it’s a broken model. So beating a bad team by 50 is good but letting your walk ons play and only winning by 35 is bad??? Do they track whose playing ? If not, That’s a model that shows nothing.
Expand ...
Yes, some of the models like Torvik take that into account and ignore the late game walk on minutes when a game is considered analytically over
 
yungdowork said:
There’s plenty of models that track lineups. Every single model on the planet is flawed, that’s why there are so many of them. If we had a perfect model, we wouldn’t have the rankings determined by sports writers who haven’t watched the majority of the games in a given week.

Mississippi Valley State is 0-22. They’ve lost 12 games by 30+ points. Teams a lot worse than us also beat them with their bench. The bottom line is that we have absolutely nothing to gain by playing teams that are that non-competitive. If we win that game by 100, it doesn’t make our resume look any better.
Expand ...
Your posts in this thread are so far off base I don't even know where to start, so I'll just go with a broad suggestion of ignore it all
 
navery12 said:
Your posts in this thread are so far off base I don't even know where to start, so I'll just go with a broad suggestion of ignore it all
Expand ...
I’m genuinely interested in what you mean.

I’m definitely not an expert on Torvik but aren’t the adjusted offensive and defensive efficiency of your opponent pretty heavily weighed into the equation? Obviously the line about beating them by 100 is hyperbole, but we’d have to massively outperform our average efficiency stats to not have that game be a net negative. Is that not accurate? MVSU is 362nd in adjusted offensive efficiency and 359th in adjusted defensive efficiency.
 
yungdowork said:
I’m genuinely interested in what you mean.

I’m definitely not an expert on Torvik but aren’t the adjusted offensive and defensive efficiency of your opponent pretty heavily weighed into the equation? Obviously the line about beating them by 100 is hyperbole, but we’d have to massively outperform our average efficiency stats to not have that game be a net negative. Is that not accurate? MVSU is 362nd in adjusted offensive efficiency and 359th in adjusted defensive efficiency.
Expand ...
Basically yes, but I think you're overstating the impact of the games vs a 300+ team over 150-200. It's all about how you perform relative to expectations, so the expectations change depending on opponent. You don't need to outperform lower rated opponents any more than you do middle or high level opponents.

As a fan I get it, I'd love to watch us play Yale or Harvard for entertainment value, but the impact on metrics is not really there. Unless I'm misunderstanding what you're saying
 
navery12 said:
Basically yes, but I think you're overstating the impact of the games vs a 300+ team over 150-200. It's all about how you perform relative to expectations, so the expectations change depending on opponent. You don't need to outperform lower rated opponents any more than you do middle or high level opponents.

As a fan I get it, I'd love to watch us play Yale or Harvard for entertainment value, but the impact on metrics is not really there. Unless I'm misunderstanding what you're saying
Expand ...
That definitely could be true. I assumed if the difference in the adjusted rating between #150ish and MVSU was bigger than between us and #150, it'd have a pretty substantial impact, but I am definitely always learning. I mostly used AdjEM but I did run the numbers of both offensive and defensive efficiency and it was true for both of those as well.


Here's some stats from KenPom in case anyone is curious what I'm referencing
Adjusted Efficiency Margins:
28.65 - UConn (#3)
1.23 - UMass Lowell (#150)
-32.90 - MVSU (#362)
 
yungdowork said:
That definitely could be true. I assumed if the difference in the adjusted rating between #150ish and MVSU was bigger than between us and #150, it'd have a pretty substantial impact, but I am definitely always learning. I mostly used AdjEM but I did run the numbers of both offensive and defensive efficiency and it was true for both of those as well.


Here's some stats from KenPom in case anyone is curious what I'm referencing
Adjusted Efficiency Margins:
28.65 - UConn (#3)
1.23 - UMass Lowell (#150)
-32.90 - MVSU (#362)
Expand ...
I think it's actually a level deeper than that. So it's not just that MVSU has a worse defense than UMass Lowell, but we'd be expected to score more points against MVSU. So if we're expected to score 70 and go for 70 against UMass Lowell that's the same as scoring 95 if that's what we're expected to get against MVSU
 
Sounds like from what you guys are saying, you basically need to embarrass a bad team to not negatively impact your numbers....seems stupid to me...what's the benefit of beating a bad team by 75 instead of 50? Because someone's model 'expects' it?
 
NJHusky said:
Sounds like from what you guys are saying, you basically need to embarrass a bad team to not negatively impact your numbers....seems stupid to me...what's the benefit of beating a bad team by 75 instead of 50? Because someone's model 'expects' it?
Expand ...
In real life, there's no benefit. It's a bad line of code in an otherwise good algorithm.
Not exactly sure how to fix it, but i believe Torvik just cuts off after a certain point. Like once a team is up/down 30 or so, the stats no longer count. But honestly, i have no idea how difficult something that is to implement. Sounds difficult to me given how many possible variables there are.
A 43 point victory against Xavier should count fully given Xavier's expectations, but a 35 point win against MVSU should count significantly less because of the different expectations. Maybe the spread should be taken into account. idk ???
 
NJHusky said:
Sounds like from what you guys are saying, you basically need to embarrass a bad team to not negatively impact your numbers....seems stupid to me...what's the benefit of beating a bad team by 75 instead of 50? Because someone's model 'expects' it?
Expand ...
You don't have to embarrass the other team, the advanced models are never looking for a 75 point win. The MVSU game was about as drastic a difference as you'll see and that was projected at 41 points. And FWIW we played our worst game of the season, won by "only" 34 and our advanced metrics barely moved. On KenPom we went from 26.87 to 26.61
 
willie99 said:
To state the obvious which has probably already been stated. Our defense is a whole lot better with Clingan playing.

With him, we're national title contenders. Without him, we're longshots. The second weekend may be our ceiling
Expand ...
TRUE
 
yungdowork said:
The issue is that we played teams that are considered very bad and then underperformed the expectations. You don’t get credit for beating a really bad team if you couldn’t cover the spread. These models are all stat based and the numbers from those games don’t flatter us compared to what we’ve done versus better opposition.
Expand ...
We play too many of these games. Theyre bad to watch, bad for team development, bad for our budget.
 
navery12 said:
Yes, some of the models like Torvik take that into account and ignore the late game walk on minutes when a game is considered analytically over
Expand ...
Maybe Torvik or KenPom track this already, correct me if they do...

For these blowout buy games, it would be more valuable to know when a game is "analytically over." For example if Team A runs another team out of the gym with 12 minutes to go, that should bump their metrics more than Team B who reaches that threshold with 8 minutes to go. Even if Team A puts in walkons and ends up winning by 18 while Team B leaves starters in and runs it up to 30
 
Drame said:
We play too many of these games. Theyre bad to watch, bad for team development, bad for our budget.
Expand ...
Bad for team development? Not sure I can say this. Hurley needed to integrate three new players into the starting lineup and develop depth with only one player, Hassan, having significant experience as a role player. And yet we’re #1 with a better record than last season at this juncture.

Bad for budget? We’re selling out not only our arenas but opposition arenas as well. Prices for tickets have risen. Yes it’s partially due to the NC hype. But that only carries into this season for a short time unless this season is successful.

Bad to watch? I partially agree with you. But losing is typically worse to watch and I personally loathed watching the Providence game to any of those cupcakes including Missouri Valley State.
 
Tyler2Rings said:
Maybe Torvik or KenPom track this already, correct me if they do...

For these blowout buy games, it would be more valuable to know when a game is "analytically over." For example if Team A runs another team out of the gym with 12 minutes to go, that should bump their metrics more than Team B who reaches that threshold with 8 minutes to go. Even if Team A puts in walkons and ends up winning by 18 while Team B leaves starters in and runs it up to 30
Expand ...
This is what Haslametrics does. He basically ends the game for scoring margin when it goes analytically final.

Torvik adjusts final margin based on the average lead throughout.

KenPom just weights mismatch games less than other games. The others might do this too.
 
You must log in or register to reply here.

Online statistics

Members online
205
Guests online
1,284
Total visitors
1,489
Totals may include hidden visitors.

Share This Page

Forum statistics

Threads
163,962
Messages
4,376,850
Members
10,168
Latest member
CTFan142


.
..
Top Bottom