Talent Comparison Game by Game 2014

[Biopsyman]

UTAH state will feel the pain the following reasons;

1. 105K fans
2. 80% humidity and 90 degrees
3. First game, everyone is healthy
4. Dooley is with the cowboys
5. Too much SEC power

 

[Madmartigan]

UTAH state will feel the pain the following reasons;

1. 105K fans
2. 80% humidity and 90 degrees
3. First game, everyone is healthy
4. Dooley is with the cowboys
5. Too much SEC power

Heard this a lot, what does it actually mean?

 

[thebjd]

Heard this a lot, what does it actually mean?

Nothing

 

[Catbone]

It looks like this: The Probability of winning at least X number of the 12 games is Y (assuming my math is correct).

at least X = Y%
1 = 100%
2 = 99.97%
3 = 99.67%
4 = 98.11%
5 = 92.65%
6 = 79.63%
7 = 58.09%
8 = 33.35%
9 = 13.95%
10 = 3.9%
11 = 0.64%
12 = 0.05%

Great, thanks!

 

[Quaint88]

If you do this sort of evaluation long enough you tend to find that attrition washes out relatively evenly.

So, in the long run attrition is included in the 30% of the time this type of analysis doesn't correctly pick the outcome. It's cool that it works out in the long run, but in the short term, it would be nice if there were some way to adjust it for attrition (both quality and quantity). Also, have you tried any methods of making the older recruiting classes who are upper classmen now have more weight in the calculation to eat into the 30% a little? I know you'd have to account for early enrollees in the draft, but it seems lke you could get a little closer. Either way, good stuff, guys.

 

[Biopsyman]

Heard this a lot, what does it actually mean?

SEC is the most dominate conference in terms of any metric over the past 20 years

 

[sjt18]

Correct...give or take. Individually each game against a team in red is a 70% chance of a win, and green is a 30% chance for a win. In aggregate it means that there is about a 58% chance that UT wins 7 or more games.

Are you sure? To me it would seem to suggest that UT will win 70% of 8 games... either 5 or 6 then 30% of 4 games or 1 additional win. Statistically shouldn't your numbers suggest that 7 is the "100%" prediction before you add the variables of injury et al?

 

[sjt18]

If you do this sort of evaluation long enough you tend to find that attrition washes out relatively evenly.

Great point. And according to what you've posted before... it wouldn't be that meaningful if they weren't exactly the same since experience doesn't move the needle that much, right?

 

[sjt18]

It looks like this: The Probability of winning at least X number of the 12 games is Y (assuming my math is correct).

at least X = Y%
1 = 100%
2 = 99.97%
3 = 99.67%
4 = 98.11%
5 = 92.65%
6 = 79.63%
7 = 58.09%
8 = 33.35%
9 = 13.95%
10 = 3.9%
11 = 0.64%
12 = 0.05%

can you compute the probability that UT wins 6-8 games? Be careful though... I've been shot at here for suggesting this team was at least a 6 win team.

 

[sjt18]

Look at last season. Some actually think that Tennessee had better players than Carolina. We don't win if Conner shaw doesn't get hurt.....probably. We beat them one out of 10 times and we were better that day, but we did not have the best players. The ratings say we did! Stupid.....

What? With all your wisdom and experience you weren't able to look at those two rosters and see how UT had some uniquely advantageous match ups?

No... what is abjectly stupid is that you don't seem to understand how the ratings generally relate to the talent of a given roster. Well... I suspect you DO understand but are so desperate to set the bar low that you reject that specific suggestion that the Vols should win some games this fall... or should have done so last fall.

 

[griffdog14]

HOW TO POST LIKE TENNESSEEDUKE: A VOLNATION TUTORIAL

Short optimistic comment ending with an exclamation mark!

Maybe one double-spaced below it!

:avatar:

Tennesseeduke

 

[sjt18]

Heard this a lot, what does it actually mean?

Bigger, faster, stronger, quicker, and experience playing against the best teams week in/week out in CFB.

Years ago when I played HS FB, we would always schedule a 4A team for our pre-season scrimmage. We played as 2A but were actually a 1A school. Generally, they beat us pretty good. They exposed every weakness and forced us to play at a higher level. We then went on to generally beat our regular opponents like dogs.

Later around the time Heath Shuler played there, they scheduled 3A teams in the regular season before conference. Very competitive games but my HS lost some of them.... then went on to win big against lower competition.

Playing against the best FORCES you to up your game even when you lose. It shows you what excellence looks like.

 

[ZoeVol]

The op informed us that overall Utah State is not as talented as some have let on with an average of #101.5.
We also know that a very experienced Sr QB can make a team more dangerous than would otherwise would be.

My evaluation of the game is that IF our defense can find a way to keep the QB in check that we should be able to beat Utah State. Nobody knows how well either side will play, so that's why we play the game.

Go Big Orange!

 

[BeattieFeathers]

If you do this sort of evaluation long enough you tend to find that attrition washes out relatively evenly.

Ok.....riidflr me this. Why rivals? Why not scout? Allen Wallace? Has rivals had any turnover in their evaluating staff in 10 years?
Surely you don't think there are 5 or 6 schools on Carolina's schedule that are more talented. I have watched Carolina play a lot and there is no way we have better players. Also, we probably have worse juniors and seniors than anyone we play, regardless of the ratings. We do not have many pro prospects at all in the upper classes.
Rivals panders to its subscription base, mainly in the south. You probably think that we would be evenly matched against UCLA based on your system. And for what it is worth 70 percent isn't that great.

 

[McDad]

Ok.....riidflr me this. Why rivals? Why not scout? Allen Wallace? Has rivals had any turnover in their evaluating staff in 10 years?
Surely you don't think there are 5 or 6 schools on Carolina's schedule that are more talented. I have watched Carolina play a lot and there is no way we have better players. Also, we probably have worse juniors and seniors than anyone we play, regardless of the ratings. We do not have many pro prospects at all in the upper classes.
Rivals panders to its subscription base, mainly in the south. You probably think that we would be evenly matched against UCLA based on your system. And for what it is worth 70 percent isn't that great.

Derek will eat your lunch (in a nice way) if you pursue this. He's smarter than you.

 

[daj2576]

So, in the long run attrition is included in the 30% of the time this type of analysis doesn't correctly pick the outcome. It's cool that it works out in the long run, but in the short term, it would be nice if there were some way to adjust it for attrition (both quality and quantity). Also, have you tried any methods of making the older recruiting classes who are upper classmen now have more weight in the calculation to eat into the 30% a little? I know you'd have to account for early enrollees in the draft, but it seems lke you could get a little closer. Either way, good stuff, guys.

Dave Bartoo (cfbmatrix.com) claims to sort his evaluations for attrition. Otherwise we do the same sort of thing. Our predictions are generally very close, meaning that while I haven't taken the time to sort for attrition, he has, and the predicted results are nearly identical, at least for UT.

 

[BeattieFeathers]

I agree with you in regards to Carolina having the better team last year. However, you are incorrect in saying that UT won because Shaw was hurt. Connor Shaw did not get hurt until the end of the game after being shut down all day. With the exception of the long touchdown pass early in the game. Vols won the game fair and square and without any excuses of this guy was hurt or that guy was hurt.

People do not realize how big of an upset that was. We likely knocked them out of the title game. Conner shaw was gimpy the whole game and at the end is where he did his best work. We beat a way better team last season but should have beaten Uga and vandy as well.

 

[BeattieFeathers]

Derek will eat your lunch (in a nice way) if you pursue this. He's smarter than you.

Recruiting services are good at predicting talent that is being brought in, but they miss a lot. Once they get on campus, coaches find out that some players are diamonds in the rough, while others like the big Carr kid from memphis weren't what we thought. Once they get there, the ratings are out the window. Especially after they become juniors and seniors when the picture becomes much clearer. We suffered a lot of attrition in the Dooley years and it doesn't make much sense to use statistics that don't account for 70 percent attrition. Furthermore, no way our talent was close to either Carolina or Missouri last season. They had way better players. I use the eyeball test. We scored a monumental upset when we beat Carolina and playing them with an injured qb was a factor. We likely don't win if he is healthy. Furthermore, I would say the more experienced qb wins 70 percent of the time.:hi:

 

[daj2576]

Derek will eat your lunch (in a nice way) if you pursue this. He's smarter than you.

Initially I chose Rivals because their site was easy to navigate, and included JUCO players in their evaluations. This began as a shot in the dark to test an hypothesis about Derek Dooley.

When I began to see how strongly the numbers correlated, I didn't see a reason to change. The longer I do this sort of thing, tweaking for certain variables, reading sports statistics books, evaluating systems that others have created, I find that Rivals is a great starting point. Others who do similar evaluations find ways to average recruiting services to wash out bias inherent in all of them.

I agree that a 70% prediction rate leaves room for improvement. 70% does tend to elucidate some pretty interesting points, like which coaches tend to perform outside of reasonable expectations (Muschamp, Dooley, Kiff, Brown, were all negative performers whereas Spurrier, Petrino, Jones are all pretty solid positive performers, and guys like Freeze, Malzahn, and Miles tend to win the number of games they should, regardless of their perception.

But simply saying it is 70% is too broad a picture. The better two teams recruit, the better the prediction rate that the higher rated team will win. The less talent two teams recruit, the less stable that relationship becomes. When one team recruits very well, and one team doesn't, the prediction rate skyrockets. The prediction rate in BCS title games was 90%.

Basically it goes like this: If two teams are evenly rated numbers suggest the home team wins about 55% of the time. The prediction rate for any SEC game is about 70%. The prediction rate for any SEC team versus third tier opponent approaches 80% or greater.

If you look at seasonal predictions in the SEC, that is how many games over-all any single team should win versus how many teams they do win, it is basically that 50% of the teams finish within 1 game of predictions, and 80% finish within 2 game of predictions. While you might not find that a strong correlation, it tends to mean that there isn't much variation in results from expectations based on talent.

EDIT: Sorry this was meant for Mr. Feathers.

 

[daj2576]

Recruiting services are good at predicting talent that is being brought in, but they miss a lot. Once they get on campus, coaches find out that some players are diamonds in the rough, while others like the big Carr kid from memphis weren't what we thought. Once they get there, the ratings are out the window. Especially after they become juniors and seniors when the picture becomes much clearer. We suffered a lot of attrition in the Dooley years and it doesn't make much sense to use statistics that don't account for 70 percent attrition. Furthermore, no way our talent was close to either Carolina or Missouri last season. They had way better players. I use the eyeball test. We scored a monumental upset when we beat Carolina and playing them with an injured qb was a factor. We likely don't win if he is healthy. Furthermore, I would say the more experienced qb wins 70 percent of the time.:hi:

Actually your assertion about QBs has been recently studied. According to the initial findings by Mr. Bartoo (I am going from memory here so don't quote me, but it is on his site), going from an experienced QB to an inexperienced QB only accounted for about a .2 game a year difference in wins. The most important position, according to this same study, was going from an experienced to inexperienced kicker. That was something like a 1.5 game a year swing.

 

[BeattieFeathers]

Initially I chose Rivals because their site was easy to navigate, and included JUCO players in their evaluations. This began as a shot in the dark to test an hypothesis about Derek Dooley.

When I began to see how strongly the numbers correlated, I didn't see a reason to change. The longer I do this sort of thing, tweaking for certain variables, reading sports statistics books, evaluating systems that others have created, I find that Rivals is a great starting point. Others who do similar evaluations find ways to average recruiting services to wash out bias inherent in all of them.

I agree that a 70% prediction rate leaves room for improvement. 70% does tend to elucidate some pretty interesting points, like which coaches tend to perform outside of reasonable expectations (Muschamp, Dooley, Kiff, Brown, were all negative performers whereas Spurrier, Petrino, Jones are all pretty solid positive performers, and guys like Freeze, Malzahn, and Miles tend to win the number of games they should, regardless of their perception.

But simply saying it is 70% is too broad a picture. The better two teams recruit, the better the prediction rate that the higher rated team will win. The worst two teams recruit, the less stable that relationship becomes. When one team recruits very well, and one team doesn't, the prediction rate skyrockets. The prediction rate in BCS title games was 90%.

Basically it goes like this: If two teams are evenly rated numbers suggest the home team wins about 55% of the time. The prediction rate for any SEC game is about 70%. The prediction rate for any SEC team versus third tier opponent approaches 80% or greater.

If you look at seasonal predictions in the SEC, that is how many games over-all any single team should win versus how many teams they do win, it is basically that 50% of the teams finish within 1 game of predictions, and 80% finish within 2 game of predictions. While you might not find that a strong correlation, it tends to mean that there isn't much variation in results from expectations based on talent.

EDIT: Sorry this was meant for Mr. Feathers.

I consider two and four game swings a very wide gulf, if that is what you mean in the last paragraph concerning 80 percent. Big difference in 4-8 and 8-4 for a team predicted to finish 6-6.

 

[daj2576]

I consider two and four game swings a very wide gulf, if that is what you mean in the last paragraph concerning 80 percent. Big difference in 4-8 and 8-4 for a team predicted to finish 6-6.

Look at it the other way: there is only a 20% chance that any team will finish three or more games from their predictions.

I tend to picture it like this: talent is a vicious Rottweiler that is leashed to a stake in the back yard. The leash is only 2 feet long tethered to the stake. Half of those dogs spend all day perched on top of the stake, and don't move. For the other half that move around, it doesn't matter which direction the Rottweiler goes, it is very unlikely that he will ever get farther than 2 feet from his stake. When 20% of the Rottweilers get farther than two feet from the stake, the variable is the competency of the owner (in our case, the coach).

That is how important talent is, that regardless of all of the other variables many people consider, it is basically a stake by which any team's expectations can be set. Few owner's have any Rottweiler that goes beyond 2 feet. Those that do, tend to do so with consistency and are either incredibly valuable (they've taught the rottie how to unchain himself and he goes for a daily stroll and comes home to wash the owners car), or incredibly negligent (the rotty gets lose and mauls a child or gets hit by a car).

Either way, 80% of the rotties will stay within two feet of that tether, regardless of anything else, with the 20% who got off divided among the incredibly well trained, or the incredibly neglected.

Interesting note: a four game a year negative trend is the threshold at which many coaches tend to lose their job.

 

[daj2576]

I consider two and four game swings a very wide gulf, if that is what you mean in the last paragraph concerning 80 percent. Big difference in 4-8 and 8-4 for a team predicted to finish 6-6.

Another example, UT last year was predicted to go 7-5 but went 5-7. That means that talent predicted 10 of 12 games, or 83%. Does it matter if UT was predicted to go 5-7 and went 7-5? That would still mean that talent predicted 10 of 12 games, or 83%. So while you see a four game swing, the reality is that talent either predicted 83%, 92% or 100% of games. Or to visualize it along a line, it would be (-2) 83%, (-1) 92%, (all correct) 100%, (+1) 92%, (+2) 83%.

 

[BeattieFeathers]

Another example, UT last year was predicted to go 7-5 but went 5-7. That means that talent predicted 10 of 12 games, or 83%. Does it matter if UT was predicted to go 5-7 and went 7-5? That would still mean that talent predicted 10 of 12 games, or 83%. So while you see a four game swing, the reality is that talent either predicted 83%, 92% or 100% of games. Or to visualize it along a line, it would be (-2) 83%, (-1) 92%, (all correct) 100%, (+1) 92%, (+2) 83%.

If that's the case, I'm not sure that the analysis tells us much at all. We will celebrate at 8-4 and remain concerned at 4-8.

 

[McDad]

If that's the case, I'm not sure that the analysis tells us much at all. We will celebrate at 8-4 and remain concerned at 4-8.

I tried to warn you. Be thankful Derek is blessed with abundant patience when explaining.