FCS Ranking Composite

[JohnStOnge]

You know what? This time I'm not going to bother re-typing. He's got 49 ranking systems there now. Take a look:

http://masseyratings.com/cf/compare1aa.htm" onclick="window.open(this.href);return false;

OVC still ahead of the Big Sky in the conference rankings. Youngstown State quietly creeping up. Eastern Illinois now at #2 and Coastal Carolina at #4. Southern Conference having an awful year with the highest rated Southern Conference team at 22 and the conference rated 7th. So on and so forth.

[89Hen]

Asburn
2. EIU
3. Tennessee State
14. Murray St
15. UT-Martin

No... there's no such thing as self implosion, right Massey?

[JohnStOnge]

Asburn
2. EIU
3. Tennessee State
14. Murray St
15. UT-Martin

No... there's no such thing as self implosion, right Massey?

All Massey is doing is doing a composite of a bunch of rating systems. If he throws one out because he doesn't agree with the rankings he's not being objective. 49 systems are included and the results "smooth" out. Thus right now when you look at those four teams composite ratings they are:

2. EIU
10. Tennessee State
36. Murray State
31. UT Martin

[89Hen]

Asburn
2. EIU
3. Tennessee State
14. Murray St
15. UT-Martin

No... there's no such thing as self implosion, right Massey?

All Massey is doing is doing a composite of a bunch of rating systems. If he throws one out because he doesn't agree with the rankings he's not being objective. 49 systems are included and the results "smooth" out. Thus right now when you look at those four teams composite ratings they are:

2. EIU
10. Tennessee State
36. Murray State
31. UT Martin

Garbage in, garbage out. Massey will defend computer models 'til the cows come home. I've had the discussion with him over these models not having enough data to be accurate and them imploding on themselves and he was very arrogant about it and dismissed even the possibility of it happening. A 12 year old in pre-algebra can look at the rankings I posted from Ashburn and see this is absolutely the case.

You can take cow manure, dry it out and make stuff with it... but I'm not buying it.

[JohnStOnge]

Garbage in, garbage out. Massey will defend computer models 'til the cows come home. I've had the discussion with him over these models not having enough data to be accurate and them imploding on themselves and he was very arrogant about it and dismissed even the possibility of it happening. A 12 year old in pre-algebra can look at the rankings I posted from Ashburn and see this is absolutely the case.

You can take cow manure, dry it out and make stuff with it... but I'm not buying it.

One of these years I'm going to remember to start at the beginning and have you pick randomly selected games from among all the games and see how you do against one or more of the better models in predicting outcomes. Maybe I'll pick Massey. And you will almost certainly lose the contest. As I've said before you'll have a shot at beating it in terms of percent winners picked correctly. But the odds will be against you and if you do win it will be close. When it comes to predicting the point spread you will have no shot.

They are about as "accurate" as one can expect anything to be in terms of picking football. They're usually correct in picking the winner somewhere between 70 and 80 percent of the time. And you're not going to find a "football expert" who is going to do significantly better than that when trying to pick every single game over the season. I couldn't do it and you couldn't do it either. Not saying we're experts. Just referring to us because we're the ones having this conversation.

[JohnStOnge]

BTW, how do you judge that a model has "imploded?" I hope it's not by saying it comes up with rankings you don't think are correct.

The way you judge a model is to see what it predicts each time and look at all of the results. It's not reasonable to expect any method of predicting the outcomes of football games to be right every time. So you have to ask yourself if being right in terms of who wins and loses, say, 77 percent of the time over 372 games as Massey's model has so far this season is a good performance.

Clearly the model has predictive value. If one were to just flip a coin each time to pick 372 games you'd be 95% confident that they'd get no more than 54 percent of the winners correct. So you have to ask yourself how 77 percent compares to other ways of doing it. And I know from having been through this in the 90s that you can ask people who think they know about football to try to beat the models and they can't. By that I mean they can't achieve a result that is "significantly" better and are probably going to do worse.

And what I'm talking about is a scenario in which they have to either pick all the games or a randomly selected sample of them. No cherry picking.

[89Hen]

One of these years I'm going to remember to start at the beginning and have you pick randomly selected games from among all the games and see how you do against one or more of the better models in predicting outcomes. Maybe I'll pick Massey. And you will almost certainly lose the contest.

Bullshit. I asked you 5 straight years to do this and you never take me up on it. Who gives a shit if the computer does better on Valpo vs Campbell games? For years, the gohens pool has completely out performed the compter models.

[89Hen]

BTW, how do you judge that a model has "imploded?" I hope it's not by saying it comes up with rankings you don't think are correct.

Simple, when it has 4 OVC teams ranked in the top 15. It's obviously imploding on itself... giving too much credit for beating another OVC team which falsely inflates them which makes them give further credit for beating them next week and so on...