Power Rank

[Mavric]

Also, unless I'm not reading it right, they say 70% on who's going to win the game. But that doesn't take into account how close the prediction was. If they said Nebraska would beat Southern Mississippi by 24 points but we only win by 3, their model is "correct" on picking the winner but not a very accurate indicator on how the game went.

 

[Sargon]

http://www.thepredictiontracker.com/ncaaresults.php

Can check the link for model performances vs. line open, line updated, etc. for 70ish models.

The power rank results were poor in 14, 13, didn't see it for 12, stopped looking.

That it uses the prior 4 years in the current model tells me enough to turn away.

 

[BigRedBuster]

I never have understood power rankings and why people pay so close attention to them. To me, it always seems like a feeble attempt at comparing a large number of programs in a manner that really doesn't tell you much more than you knew before you started.

 

[cg_8]

Also, unless I'm not reading it right, they say 70% on who's going to win the game. But that doesn't take into account how close the prediction was. If they said Nebraska would beat Southern Mississippi by 24 points but we only win by 3, their model is "correct" on picking the winner but not a very accurate indicator on how the game went.

So wait... For top ranked teams who win more than 70% of their games...

Is it possible that 70% is a terrible predictor considering that the top 10 teams are going to win around 10 games anyway? I mean, "predicting" Alabama, Oregon, Ohio State, FSU, etc. are going to win 11 of their games, then saying "I was 70% correct..." wouldn't you call me an idiot? (although, a lot of people would call me an idiot either way...)

It's not hard to be 70% correct on us beating Southern Miss (the 2004 version of myself says hello...) but even this model predicted Ohio State beating VT, it still predicted Ohio State winning all of their games. That's a 91%.

I'm thinking too much about it...

Anyway, the season needs to start for a guy like me...

 

[Mavric]

http://www.thepredictiontracker.com/ncaaresults.php

Can check the link for model performances vs. line open, line updated, etc. for 70ish models.

The power rank results were poor in 14, 13, didn't see it for 12, stopped looking.

That it uses the prior 4 years in the current model tells me enough to turn away.

Great find.

What is interesting (scary?) is how close all those models are to 50% against the spread.

 

[NUance]

Don't get why Wisconsin is ahead of MSU.

Me too. Seems like every poll has MSU underrated, and Wisc overrated. Didn't the Badgers just lose their coach and Heisman candidate RB?

 

[Sargon]

http://www.thepredictiontracker.com/ncaaresults.php

Can check the link for model performances vs. line open, line updated, etc. for 70ish models.

The power rank results were poor in 14, 13, didn't see it for 12, stopped looking.

That it uses the prior 4 years in the current model tells me enough to turn away.

Great find.

What is interesting (scary?) is how close all those models are to 50% against the spread.

Sure the tight grouping around 50% success against the spread is always the case. A tool consistently much worse will be abandoned (or ignored anyway). A tool much better would over time have increasingly more weight in the minds of line-setters and bettors. The lines will shift to be in better agreement with the "smartest" predictors. (True, lines are meant to yield 50/50 action however they do end up being great predictors year in year out. Folks try to stop betting or change their failed ways if they have been "stupid" forecasters.)