Pool C -- 2014

[smedindy]

Oh, derp. I did it at work while waiting on exports and packing boxes.

[D3MAFAN]

I am waiting for the committee to use the "Body of Work" explanation. I translate that in terms of Division III, "Prior Year(s)?" results.

[D3MAFAN]

To be honest I think it would help everyone if we stopped putting geographics together with the seeds. Instead just do 4 top seeds, since the brackets aren't all that geographic anymore anyway.

1 -- UWW
2 -- UMU
3 -- UMHB
4 -- Wesley

I could easily see 2 and 3 being reversed, but if both teams are undefeated and you start going to that secondary criteria of "past playoff performance" you would put UMU second. This method also has the benefit of showing who we think will go on the road at the end.

All I know is that UWW and UMU will not bee seeing each other until Salem, which I agree. Therefore, seeds are going to be in favor of that relationship based on prior year history.

[ExTartanPlayer]

I think what we're all getting here is that we want to see Margin of Victory added to the data.

No, I don't think that's a good idea really at all.   

of course it is.  You don't have to give extra credit to Mount for winning 100-0 versus 70-0.  People love to point out exceptions to the margin of victory rule rather than point out all the cases where the margin gave us meaningful predictive data.
[/quote]

It's not just about 100-0 vs. 70-0. Even if you put a cap on it - something like any win over 28 points being equal - you're still opening a few cans of worms that I find pretty objectionable. Teams with a 14-21 point lead and 2 minutes to go would har incentive to keep trying to score instead of getting the backups a few snaps or just kneeling on it.  In general, teams would be more reluctant to put backups in during games near whatever the MOV cutoff is, lest the backups give up a score and hurt your MOV.

I know that you pretty much don't care about anyone else's feelings, but some of us do.

Plus, not all 17-point wins are created equal. Team A gets out to a 31-7 lead with ten minutes to play, puts in the JV defense & pulls the starting offense, and cruises home with a 31-14 win. Team B enters the fourth quarter in a 17-14 nailbiter before a touchdown and then a pick-six blow the margin up to 31-14. Team A dominated; Team B's game was in doubt til the end. They'd get the same credit because Team A took their foot off the gas, as any good sportsmen would do, and elected to get their young guys some experience rather than assuring the largest possible margin.

[AO]

I think what we're all getting here is that we want to see Margin of Victory added to the data.

No, I don't think that's a good idea really at all.   

of course it is.  You don't have to give extra credit to Mount for winning 100-0 versus 70-0.  People love to point out exceptions to the margin of victory rule rather than point out all the cases where the margin gave us meaningful predictive data.

It's not just about 100-0 vs. 70-0. Even if you put a cap on it - something like any win over 28 points being equal - you're still opening a few cans of worms that I find pretty objectionable. Teams with a 14-21 point lead and 2 minutes to go would har incentive to keep trying to score instead of getting the backups a few snaps or just kneeling on it.  In general, teams would be more reluctant to put backups in during games near whatever the MOV cutoff is, lest the backups give up a score and hurt your MOV.

I know that you pretty much don't care about anyone else's feelings, but some of us do.

Plus, not all 17-point wins are created equal. Team A gets out to a 31-7 lead with ten minutes to play, puts in the JV defense & pulls the starting offense, and cruises home with a 31-14 win. Team B enters the fourth quarter in a 17-14 nailbiter before a touchdown and then a pick-six blow the margin up to 31-14. Team A dominated; Team B's game was in doubt til the end. They'd get the same credit because Team A took their foot off the gas, as any good sportsmen would do, and elected to get their young guys some experience rather than assuring the largest possible margin.

Blowouts aren't given the same weight as the games against the best teams on the schedule but it's important that the 4th quarter of all games isn't "meaningless".  The worst thing that happens in sports is when the players are playing in "garbage" time.  The JV quarterback doesn't just want to play in "garbage" time, he wants to throw the ball and have it count for something.  You could leave your first team out there but you're risking injury for very little statistical gain.  Athletes will understand that they're not just trying to score at the end of the game to make you feel bad, they're just trying to continue playing.  One of the most crucial findings in sabermetrics, across virtually all sports, is that the average margin by which a team wins or loses conveys more information than wins and losses alone.

[wally_wabash]

I think what we're all getting here is that we want to see Margin of Victory added to the data.

No, I don't think that's a good idea really at all.   

of course it is.  You don't have to give extra credit to Mount for winning 100-0 versus 70-0.  People love to point out exceptions to the margin of victory rule rather than point out all the cases where the margin gave us meaningful predictive data.

It's not just about 100-0 vs. 70-0. Even if you put a cap on it - something like any win over 28 points being equal - you're still opening a few cans of worms that I find pretty objectionable. Teams with a 14-21 point lead and 2 minutes to go would har incentive to keep trying to score instead of getting the backups a few snaps or just kneeling on it.  In general, teams would be more reluctant to put backups in during games near whatever the MOV cutoff is, lest the backups give up a score and hurt your MOV.

I know that you pretty much don't care about anyone else's feelings, but some of us do.

Plus, not all 17-point wins are created equal. Team A gets out to a 31-7 lead with ten minutes to play, puts in the JV defense & pulls the starting offense, and cruises home with a 31-14 win. Team B enters the fourth quarter in a 17-14 nailbiter before a touchdown and then a pick-six blow the margin up to 31-14. Team A dominated; Team B's game was in doubt til the end. They'd get the same credit because Team A took their foot off the gas, as any good sportsmen would do, and elected to get their young guys some experience rather than assuring the largest possible margin.

Blowouts aren't given the same weight as the games against the best teams on the schedule but it's important that the 4th quarter of all games isn't "meaningless".  The worst thing that happens in sports is when the players are playing in "garbage" time.  The JV quarterback doesn't just want to play in "garbage" time, he wants to throw the ball and have it count for something.  You could leave your first team out there but you're risking injury for very little statistical gain.  Athletes will understand that they're not just trying to score at the end of the game to make you feel bad, they're just trying to continue playing.  One of the most crucial findings in sabermetrics, across virtually all sports, is that the average margin by which a team wins or loses conveys more information than wins and losses alone.

MOV in systems where there isn't incentive to score as much as humanly possible is probably very useful for predictive purposes.  But you can't make it a criteria for selection or seeding because then it stops being the organic thing that your hypothetical predictive model needs it to be to be useful.

[AO]

MOV in systems where there isn't incentive to score as much as humanly possible is probably very useful for predictive purposes.  But you can't make it a criteria for selection or seeding because then it stops being the organic thing that your hypothetical predictive model needs it to be to be useful.

Care to explain like I'm 5?  Organic?

[bashbrother]

for me as well Wally? 

http://youtu.be/B1QpyGa61zs

[ITH radio]

I am waiting for the committee to use the "Body of Work" explanation. I translate that in terms of Division III, "Prior Year(s)?" results.

That's correct. On our show last season that was basically admitted / alluded to by the Chair, i.e., prior playoff experience / results are factored in.

[wally_wabash]

MOV in systems where there isn't incentive to score as much as humanly possible is probably very useful for predictive purposes.  But you can't make it a criteria for selection or seeding because then it stops being the organic thing that your hypothetical predictive model needs it to be to be useful.

Care to explain like I'm 5?  Organic?

MOV, in the absence of incentive to push the score differential as far as you can, is useful.  And it could absolutely be a useful data point.  However, once you attach that particular incentive to the MOV, the usefulness of that number is gone because now teams are making the effort to skew the stat and it doesn't mean the thing that you think it means anymore. 

[AO]

MOV in systems where there isn't incentive to score as much as humanly possible is probably very useful for predictive purposes.  But you can't make it a criteria for selection or seeding because then it stops being the organic thing that your hypothetical predictive model needs it to be to be useful.

Care to explain like I'm 5?  Organic?

MOV, in the absence of incentive to push the score differential as far as you can, is useful.  And it could absolutely be a useful data point.  However, once you attach that particular incentive to the MOV, the usefulness of that number is gone because now teams are making the effort to skew the stat and it doesn't mean the thing that you think it means anymore.

Didn't I address this point by saying it wouldn't have to award the blowouts?  Use the law of diminishing returns like Sagarin does.  Teams should make every effort to "skew" all stats.  Doesn't mean they're going to be successful as the defense and the statistician will be there to cut out the noise.  Have we already forgotten about our discussion about how difficult it is for a coach to guarantee an upgrade to their SOS through scheduling?  Wouldn't it be nice to have a criteria that they could affect in-game?

[wally_wabash]

MOV in systems where there isn't incentive to score as much as humanly possible is probably very useful for predictive purposes.  But you can't make it a criteria for selection or seeding because then it stops being the organic thing that your hypothetical predictive model needs it to be to be useful.

Care to explain like I'm 5?  Organic?

MOV, in the absence of incentive to push the score differential as far as you can, is useful.  And it could absolutely be a useful data point.  However, once you attach that particular incentive to the MOV, the usefulness of that number is gone because now teams are making the effort to skew the stat and it doesn't mean the thing that you think it means anymore.

Didn't I address this point by saying it wouldn't have to award the blowouts?  Use the law of diminishing returns like Sagarin does.  Teams should make every effort to "skew" all stats.  Doesn't mean they're going to be successful as the defense and the statistician will be there to cut out the noise.  Have we already forgotten about our discussion about how difficult it is for a coach to guarantee an upgrade to their SOS through scheduling?  Wouldn't it be nice to have a criteria that they could affect in-game?

By incentivizing poor sportsmanship?  No.  Not that way. 

It's easy to say that you can include MOV as a criteria, but only in certain, non-blowout cases.  I don't know how you can do that.  Where is the line?  How do you decide what points in a game are meaningful and what points aren't?  And when do you decide when teams are piling on just to pile on and when teams are scoring late with a lot of duplicate number guys?  It's impossible to corral all of that because every game is a case by case basis and nobody on these committees has time to scrub every single game to "cut out the noise" as you say. 

We can use MOV in models to predict results and to predict maybe which at-large teams have performed better than others, but you can't attach selection incentive to it or else it loses its meaning.  It's like when the subjects of some kind of social study become aware that they are subjects of some kind of social study and their behavior changes.  Whatever it was that you were trying to learn about the way people behave naturally is destroyed. 

[AO]

MOV in systems where there isn't incentive to score as much as humanly possible is probably very useful for predictive purposes.  But you can't make it a criteria for selection or seeding because then it stops being the organic thing that your hypothetical predictive model needs it to be to be useful.

Care to explain like I'm 5?  Organic?

MOV, in the absence of incentive to push the score differential as far as you can, is useful.  And it could absolutely be a useful data point.  However, once you attach that particular incentive to the MOV, the usefulness of that number is gone because now teams are making the effort to skew the stat and it doesn't mean the thing that you think it means anymore.

Didn't I address this point by saying it wouldn't have to award the blowouts?  Use the law of diminishing returns like Sagarin does.  Teams should make every effort to "skew" all stats.  Doesn't mean they're going to be successful as the defense and the statistician will be there to cut out the noise.  Have we already forgotten about our discussion about how difficult it is for a coach to guarantee an upgrade to their SOS through scheduling?  Wouldn't it be nice to have a criteria that they could affect in-game?

By incentivizing poor sportsmanship?  No.  Not that way. 

It's easy to say that you can include MOV as a criteria, but only in certain, non-blowout cases.  I don't know how you can do that.  Where is the line?  How do you decide what points in a game are meaningful and what points aren't?  And when do you decide when teams are piling on just to pile on and when teams are scoring late with a lot of duplicate number guys?  It's impossible to corral all of that because every game is a case by case basis and nobody on these committees has time to scrub every single game to "cut out the noise" as you say. 

We can use MOV in models to predict results and to predict maybe which at-large teams have performed better than others, but you can't attach selection incentive to it or else it loses its meaning.  It's like when the subjects of some kind of social study become aware that they are subjects of some kind of social study and their behavior changes.  Whatever it was that you were trying to learn about the way people behave naturally is destroyed.

I don't think you have to draw a line.  Every additional point helps less than the previous point.  You could make the curve especially steep at the higher margins.  I'm not sure how Sagarin works it but it's said that 70-0 is nearly equal to 35-0.

I don't see anyone crying foul in D3 about 50 point losses.  They happen all the time.

[hickory_cornhusker]

MOV could skew it at lower levels as well. If I have first and ten from the opponents 30 with a minute left and up 7, I'm not taking two knees and going home. I'm going to try to get more points. When people are playing for MOV of victory it changes how games end.

[AO]

MOV could skew it at lower levels as well. If I have first and ten from the opponents 30 with a minute left and up 7, I'm not taking two knees and going home. I'm going to try to get more points. When people are playing for MOV of victory it changes how games end.

Go ahead and risk a turnover when you've already won the game if you just kneel.  The other team will be more excited about the possibilities than the coach just trying to pump up the margin.