Centreville also passed Ocean Lakes--this might have surprised some people.
I was thinking about this and I used to rate teams and I think this situation probably gets VHSL-Reference some complaints, but I think I understand how this might happen.
There are a gazillion ways for computers to rank football teams (you can actually get a master's degree from N.C. State in this stuff, if you want--I want), and of the systems that seem thought out pretty well there's not really more than a dime's worth of difference. So, please don't think I'm putting down their system over there. Their system works just as well as any I've seen, and I have never seen any of the algorithms they use, but I've watched it enough to know it's not like my algorithm was and it's not like Kenneth Massey's or Jeff Self who used to rate them. I think Cal Preps uses something close to what I used to do and I have even wondered on occasion if they took my system as a starting point and modified it (not a problem for me, that's what I did to get mine; I took a system I saw in a 1970's pre-season NFL magazine and modified it).
But here's what I think is likely happening:
So, I had commented on how close it was over there before with the difference in difference between our four semi-finalists being minuscule, but this is how they were ranked going into this week's action:
1. Tuscarora
2. Massaponax
3. L.C. Bird
4. Salem
Again, note that they were super close.
Most of us might think it's a no-brainer that Tuscarora should stay on top of Bird. Why? Tuscarora was rated higher going in, Tuscarora was playing a higher rated team than Bird, Tuscarora won by more points than Bird. (Bird did play a better game defensively than Tuscarora purely form a points allowed stand point and some systems take that into account, but generally not enough to make this switch occur).
Tuscarora was 13-0 for a 1.000 winning average. Tuscarora was 14-0 afterwards; still 1.000.
Bird was 11-2 for a .846 winning average. Bird was 12-2 for a .857 winning average. This has gone up .011
Other variables in the formula were enough to draw Bird very close to Tuscarora before the game. So, if this formula uses winning average as one of it's factors, Bird could have passed Tuscarora and Centreville could have passed Ocean Lakes due to the fact that the teams with the worse record got a little boost in the winning percentage category. As close as the teams were it would definitely be likely.
So, I assume that any system that had Bird go by Tuscarora this week or Centreville go by Oscar Smith this week uses winning percentage as a factor (it may not even be in there explicitly, but folded into the formula in a way the creator didn't even realize. THIS IS NOT A FLAW in the system. As you get more evidence you might change how much you want to weigh previous evidence, and using this as a factor is one way of saying, "This win tells me I don't want to hold those old losses against this team now, quite as much as I did last week (see Ohio State)." It is counter intuitive to us as we are used to polls that like to lock teams in spots until they lose or at least until they win a squeaker and somebody below them just looks like a monster beast.
I just think it's interesting to watch these things and speculate about how math formulas might lead to this (it could be a totally different reason for all I know). In my old system this couldn't have happened, but I just went about getting stuff a different way, and I doubt my system was any better than the VHSL-Reference system.
Of course, if you can do better, just waste a few months of your life coming up with your own system and writing all the code so that people can tell you what an idiot your are. It's nice to be appreciated.
I love VHSL-Reference, and this post is a pre-emptive strike and a plea to everyone to try and understand that what he's doing is a lot of work, he's making no money off of any of us and he gives a wealth of data we can't easily find anywhere else. Please don't write him, or any of these guys, any hate mail. It'll all be sorted out on the field in 5 days.
I was thinking about this and I used to rate teams and I think this situation probably gets VHSL-Reference some complaints, but I think I understand how this might happen.
There are a gazillion ways for computers to rank football teams (you can actually get a master's degree from N.C. State in this stuff, if you want--I want), and of the systems that seem thought out pretty well there's not really more than a dime's worth of difference. So, please don't think I'm putting down their system over there. Their system works just as well as any I've seen, and I have never seen any of the algorithms they use, but I've watched it enough to know it's not like my algorithm was and it's not like Kenneth Massey's or Jeff Self who used to rate them. I think Cal Preps uses something close to what I used to do and I have even wondered on occasion if they took my system as a starting point and modified it (not a problem for me, that's what I did to get mine; I took a system I saw in a 1970's pre-season NFL magazine and modified it).
But here's what I think is likely happening:
So, I had commented on how close it was over there before with the difference in difference between our four semi-finalists being minuscule, but this is how they were ranked going into this week's action:
1. Tuscarora
2. Massaponax
3. L.C. Bird
4. Salem
Again, note that they were super close.
Most of us might think it's a no-brainer that Tuscarora should stay on top of Bird. Why? Tuscarora was rated higher going in, Tuscarora was playing a higher rated team than Bird, Tuscarora won by more points than Bird. (Bird did play a better game defensively than Tuscarora purely form a points allowed stand point and some systems take that into account, but generally not enough to make this switch occur).
Tuscarora was 13-0 for a 1.000 winning average. Tuscarora was 14-0 afterwards; still 1.000.
Bird was 11-2 for a .846 winning average. Bird was 12-2 for a .857 winning average. This has gone up .011
Other variables in the formula were enough to draw Bird very close to Tuscarora before the game. So, if this formula uses winning average as one of it's factors, Bird could have passed Tuscarora and Centreville could have passed Ocean Lakes due to the fact that the teams with the worse record got a little boost in the winning percentage category. As close as the teams were it would definitely be likely.
So, I assume that any system that had Bird go by Tuscarora this week or Centreville go by Oscar Smith this week uses winning percentage as a factor (it may not even be in there explicitly, but folded into the formula in a way the creator didn't even realize. THIS IS NOT A FLAW in the system. As you get more evidence you might change how much you want to weigh previous evidence, and using this as a factor is one way of saying, "This win tells me I don't want to hold those old losses against this team now, quite as much as I did last week (see Ohio State)." It is counter intuitive to us as we are used to polls that like to lock teams in spots until they lose or at least until they win a squeaker and somebody below them just looks like a monster beast.
I just think it's interesting to watch these things and speculate about how math formulas might lead to this (it could be a totally different reason for all I know). In my old system this couldn't have happened, but I just went about getting stuff a different way, and I doubt my system was any better than the VHSL-Reference system.
Of course, if you can do better, just waste a few months of your life coming up with your own system and writing all the code so that people can tell you what an idiot your are. It's nice to be appreciated.
I love VHSL-Reference, and this post is a pre-emptive strike and a plea to everyone to try and understand that what he's doing is a lot of work, he's making no money off of any of us and he gives a wealth of data we can't easily find anywhere else. Please don't write him, or any of these guys, any hate mail. It'll all be sorted out on the field in 5 days.
