For those of you who know me or have followed my talent evaluation threads, you also know that I was one of the few who was trying to illustrate that the UT v. Oregon game, based on those evaluations, would be a close game.
Tie me up and give me my lashings. I was wrong.
Last night was an "aha!" moment for me.
I went back over some of these predictions and came to another interesting conclusion: talent can be used to predict outcomes of games (I am running about 70% this year) but it is an awful indication of score differential.
Scores seem to be distributed wildly all over the board, with some teams who have a great talent disparity having very close games, and some teams who are evenly matched having blow-out games. The bottom line remains that talent CAN be used to predict the outcome of games but that predicting the score, or score differential, using only those ratings is inconclusive.
The question I have is this: why does score matter? In predicting a win loss rate across seasons, it doesn't. Huge score differentials have never helped Oregon win a national championship any more than squeaking past Arkansas (a far inferior team talent wise) didn't keep UT from winning one in 1998.
Basically scores are good for gamblers and to make fans feel really good, or really bad, about their team.
Think about last year for a minute. Georgia beats Vanderbilt 48-3 in a blow out. Georgia beats UT 51-44 in a shoot out. If you try to use scores to see how good a team is based on performance against a common opponent you could conclude that if Georgia beats Vanderbilt by 45 and UT by 7, then UT is better than Vanderbilt by 38, right? That is if scores matter like the way we are trying to say they do after yesterday.
Wrong. Vandy beats UT 41-18. That is a 68 point swing from thinking UT is 38 points better to losing by 23.
The talent predictions in those games were right, viewed only through the W or the L, 2 out of 3 times. Shocker!
Even with evenly matched talent (on paper) I had to give the W to Oregon yesterday based solely off of the fact that Oregon is playing at home (another quantifiable predictor in games). So we lost to Oregon, that was predicted by most.
It doesn't matter what the score was, it really doesn't. It only counts for 1 loss. If we had beat Oregon by 100 it would only count for 1 win, and honestly has no bearing on any other game down the pipe.
Next week we play a Florida team that is having some struggles but is still, on paper, an enormously talented team. It has to be counted in the L category, but that shouldn't create a panic. I will admit that UT has been a huge talent under-performer in recent years (-4 games or so a year) so either UT is singularly exempt from talent evaluations that predict almost 70% of the other games, or the triumvirate of Fulmer, Kiffin and Dooley could not utilize the talent that they had.
If it makes you feel better, both Dooley and Kiffin have a history at other schools of consistently under utilizing talent. It should make you feel better that Jones has never dropped below the talent predicted rate of his teams, even when he went 4-8 at Cincinnati.
Here is how talent should predict the remainder of the UT season. That is operating off of the presumption that Jones continues his historical ability of simply winning the games that talent predicts he should win. Note that in year two and three at Cincy, Jones over-performed by 3 wins a year while increasing the recruiting averages.
Now get to whippin' and Go Vols!
Tie me up and give me my lashings. I was wrong.
Last night was an "aha!" moment for me.
I went back over some of these predictions and came to another interesting conclusion: talent can be used to predict outcomes of games (I am running about 70% this year) but it is an awful indication of score differential.
Scores seem to be distributed wildly all over the board, with some teams who have a great talent disparity having very close games, and some teams who are evenly matched having blow-out games. The bottom line remains that talent CAN be used to predict the outcome of games but that predicting the score, or score differential, using only those ratings is inconclusive.
The question I have is this: why does score matter? In predicting a win loss rate across seasons, it doesn't. Huge score differentials have never helped Oregon win a national championship any more than squeaking past Arkansas (a far inferior team talent wise) didn't keep UT from winning one in 1998.
Basically scores are good for gamblers and to make fans feel really good, or really bad, about their team.
Think about last year for a minute. Georgia beats Vanderbilt 48-3 in a blow out. Georgia beats UT 51-44 in a shoot out. If you try to use scores to see how good a team is based on performance against a common opponent you could conclude that if Georgia beats Vanderbilt by 45 and UT by 7, then UT is better than Vanderbilt by 38, right? That is if scores matter like the way we are trying to say they do after yesterday.
Wrong. Vandy beats UT 41-18. That is a 68 point swing from thinking UT is 38 points better to losing by 23.
The talent predictions in those games were right, viewed only through the W or the L, 2 out of 3 times. Shocker!
Even with evenly matched talent (on paper) I had to give the W to Oregon yesterday based solely off of the fact that Oregon is playing at home (another quantifiable predictor in games). So we lost to Oregon, that was predicted by most.
It doesn't matter what the score was, it really doesn't. It only counts for 1 loss. If we had beat Oregon by 100 it would only count for 1 win, and honestly has no bearing on any other game down the pipe.
Next week we play a Florida team that is having some struggles but is still, on paper, an enormously talented team. It has to be counted in the L category, but that shouldn't create a panic. I will admit that UT has been a huge talent under-performer in recent years (-4 games or so a year) so either UT is singularly exempt from talent evaluations that predict almost 70% of the other games, or the triumvirate of Fulmer, Kiffin and Dooley could not utilize the talent that they had.
If it makes you feel better, both Dooley and Kiffin have a history at other schools of consistently under utilizing talent. It should make you feel better that Jones has never dropped below the talent predicted rate of his teams, even when he went 4-8 at Cincinnati.
Here is how talent should predict the remainder of the UT season. That is operating off of the presumption that Jones continues his historical ability of simply winning the games that talent predicts he should win. Note that in year two and three at Cincy, Jones over-performed by 3 wins a year while increasing the recruiting averages.
Now get to whippin' and Go Vols!