Talent Analysis: A-State

[Face]

I put together a basic talent analysis for the USU team and predicted (as did 99% of people) we would win big based on talent levels. Team talent is one of, if not THE determining factor of game outcomes at the college level.

To give people a snapshot of our talent vs. weekly opponents, I'm going to put together one of these threads each week. They are not intended to be absolute predictions, but a basic expectation based on raw talent stats. So here we go for stAte...


The recruiting class rankings for Arky state according to Rivals are as follows:
2014 – 86
2013 – 108
2012 – 84
2011 – 96
The average stAte recruiting class for that time interval is 93.5.

The % (ignoring the effect of JUCOs; i.e. JUCOs count for the year they signed) of players on the roster corresponding with each recruiting year are:
2014 – 26.6%
2013 – 21.1%
2012 – 26.5%
2011 – 25.6%
The weighted average recruiting rank by % of class on roster is 92.7.

I made one more weighted average calculation. Assuming half the players starting are SRs, 25% are JRs, 15% are SOs, and 10% are FRs, the w/a recruiting class is 93.8.

The composite rank according to the above measures is 93.3.

Comparing this to our numbers:
Rivals Rank
2014 – 5
2013 – 21
2012 – 17
2011 – 13
Average: 14.0.

The % (ignoring the effect of JUCOs) of players on the roster corresponding with each recruiting year are:
2014 – 31.7%
2013 – 21.0%
2012 – 21.0%
2011 – 26.7%
The weighted average recruiting rank by % of class on roster is 13.0.

Our roster has been pretty wonky, so i made the measure a little different in W/A by starter in each class.

Assuming 30% of the players starting are SRs, 20% are JRs, 30% are SOs, and 30% are FRs, the w/a recruiting class is 13.8.

Composite: 13.6


Our composite talent advantage (or deficit) is: 79.7.

Based on the above numbers, we should comfortably handle the Red Wolves.

Against USU, we had a CTA (Composite Talent Advantage) of 86.2. This lead to an effective 0.36 point victory margin (PVM) per talent per CTA point (31/86.2).

If that ratio holds true with a 10% margin of error, UT should beat Arky State by 25 and 32 points.

I'm interested to see how these numbers turn out when the rating is closer and how the PVM/CTA ratio changes week to week.

 

[Sgt_Nick_Fury]

Interested to see these "numbers" with Sooners and Dawgs.
As well as how they stack up against those finals scores.

 

[Freak]

I'm interested to see how these numbers look against Oklahoma. I hope you'll post this for that game as well.

*Nick beat me to it.

 

[Sgt_Nick_Fury]

My guess is as useful as these numbers are against lower tier schools, when we get into top 10 teams, things like experience are going to become a larger factor. I think our raw talent easily matches the Sooners or Georgia....but experience wise, and state of program still needs to catch up.

This is an incredibly talented young team, but it may need a season in the SEC oven before it's really done. Then again it is college ball......shizz happens! Some kids grow up faster then others.

 

[daj2576]

I put together a basic talent analysis for the USU team and predicted (as did 99% of people) we would win big based on talent levels. Team talent is one of, if not THE determining factor of game outcomes at the college level.

To give people a snapshot of our talent vs. weekly opponents, I'm going to put together one of these threads each week. They are not intended to be absolute predictions, but a basic expectation based on raw talent stats. So here we go for stAte...


The recruiting class rankings for Arky state according to Rivals are as follows:
2014 – 86
2013 – 108
2012 – 84
2011 – 96
The average stAte recruiting class for that time interval is 93.5.

The % (ignoring the effect of JUCOs; i.e. JUCOs count for the year they signed) of players on the roster corresponding with each recruiting year are:
2014 – 26.6%
2013 – 21.1%
2012 – 26.5%
2011 – 25.6%
The weighted average recruiting rank by % of class on roster is 92.7.

I made one more weighted average calculation. Assuming half the players starting are SRs, 25% are JRs, 15% are SOs, and 10% are FRs, the w/a recruiting class is 93.8.

The composite rank according to the above measures is 93.3.

Comparing this to our numbers:
Rivals Rank
2014 – 5
2013 – 21
2012 – 17
2011 – 13
Average: 14.0.

The % (ignoring the effect of JUCOs) of players on the roster corresponding with each recruiting year are:
2014 – 31.7%
2013 – 21.0%
2012 – 21.0%
2011 – 26.7%
The weighted average recruiting rank by % of class on roster is 13.0.

Our roster has been pretty wonky, so i made the measure a little different in W/A by starter in each class.

Assuming 30% of the players starting are SRs, 20% are JRs, 30% are SOs, and 30% are FRs, the w/a recruiting class is 13.8.

Composite: 13.6


Our composite talent advantage (or deficit) is: 79.7.

Based on the above numbers, we should comfortably handle the Red Wolves.

We had a CTA (Composite Talent Advantage) of 86.2. This lead to an effective 0.36 point victory margin (PVM) per talent per CTA point (31/86.2).

If that ratio holds true with a 10% margin of error, UT should beat Arky State by 25 and 32 points.

I'm interested to see how these numbers turn out when the rating is closer and how the PVM/CTA ratio changes week to week.

You say we "had a CTA of 86.2" the tense of this sentence confuses me. Can you explain this more.

Having done this very same thing for a couple years, I too am interested to see what your results are on PVM/CTA ratio. What I have found, to date, is that using this sort of analysis alone does not tend to provide a stable way to predict score differentials, just wins and losses. For example, Texas A&M out ranked SCAR by about 3 points, but won by about 10 points per talent difference. Or, Oregon vs. UT last year, both teams had the same talent average as did Oklahoma and Notre Dame. Neither score was particularly close. I understand the problem with trying to make a rule out of the exceptions, but my data has a far lower correlation of talent difference to score difference.

 

[tnprogrammer]

I would be interested to go back and see what these numbers would have looked like for the Utah St game, now that we know the outcome.

 

[daj2576]

My guess is as useful as these numbers are against lower tier schools, when we get into top 10 teams, things like experience are going to become a larger factor. I think our raw talent easily matches the Sooners or Georgia....but experience wise, and state of program still needs to catch up.

This is an incredibly talented young team, but it may need a season in the SEC oven before it's really done. Then again it is college ball......shizz happens! Some kids grow up faster then others.

Our raw talent matches the Sooners, but not UGA. UGA is a far more talented team than you would expect.

Here is a spreadsheet of the SEC using these sorts of talent averages as a guide. So far this season (all of two games) the talent averages have predicted all of the games correctly. By the end of the year, roughly 70% of the games will likely be predicted by these talent averages.

https://docs.google.com/spreadsheet/ccc?key=0AkwyQgwl-hyfdEU2WkFralNPVTE1VG5xa0Ffak5OMmc&usp=sharing

 

[Face]

I'm interested to see how these numbers look against Oklahoma. I hope you'll post this for that game as well.

*Nick beat me to it.

I will do one then and I'm just as interested. After we see a 2nd game, I'll probably try to work in a home field advantage adjustment to the numbers as well.

 

[daj2576]

I would be interested to go back and see what these numbers would have looked like for the Utah St game, now that we know the outcome.

Not the OP, but a similar analysis. If I Bleed Orange, I'm Bled Out: Don't Sweat Utah State.

 

[volinSmyrna]

I put together a basic talent analysis for the USU team and predicted (as did 99% of people) we would win big based on talent levels. Team talent is one of, if not THE determining factor of game outcomes at the college level.

To give people a snapshot of our talent vs. weekly opponents, I'm going to put together one of these threads each week. They are not intended to be absolute predictions, but a basic expectation based on raw talent stats. So here we go for stAte...


The recruiting class rankings for Arky state according to Rivals are as follows:
2014 – 86
2013 – 108
2012 – 84
2011 – 96
The average stAte recruiting class for that time interval is 93.5.

The % (ignoring the effect of JUCOs; i.e. JUCOs count for the year they signed) of players on the roster corresponding with each recruiting year are:
2014 – 26.6%
2013 – 21.1%
2012 – 26.5%
2011 – 25.6%
The weighted average recruiting rank by % of class on roster is 92.7.

I made one more weighted average calculation. Assuming half the players starting are SRs, 25% are JRs, 15% are SOs, and 10% are FRs, the w/a recruiting class is 93.8.

The composite rank according to the above measures is 93.3.

Comparing this to our numbers:
Rivals Rank
2014 – 5
2013 – 21
2012 – 17
2011 – 13
Average: 14.0.

The % (ignoring the effect of JUCOs) of players on the roster corresponding with each recruiting year are:
2014 – 31.7%
2013 – 21.0%
2012 – 21.0%
2011 – 26.7%
The weighted average recruiting rank by % of class on roster is 13.0.

Our roster has been pretty wonky, so i made the measure a little different in W/A by starter in each class.

Assuming 30% of the players starting are SRs, 20% are JRs, 30% are SOs, and 30% are FRs, the w/a recruiting class is 13.8.
Composite: 13.6


Our composite talent advantage (or deficit) is: 79.7.

Based on the above numbers, we should comfortably handle the Red Wolves.

We had a CTA (Composite Talent Advantage) of 86.2. This lead to an effective 0.36 point victory margin (PVM) per talent per CTA point (31/86.2).

If that ratio holds true with a 10% margin of error, UT should beat Arky State by 25 and 32 points.

I'm interested to see how these numbers turn out when the rating is closer and how the PVM/CTA ratio changes week to week.

Am I missing something? That adds up to 110%. Not sure if it was just a typo.

 

[Izzy]

"So your telling me they have a chance!!"
Loyd Christmas

 

[sechamp2012]

Do you factor in players that where high star players that have left the team for one reason or the other. Our 2011 class does not seem that strong in reality Rivals 13.

 

[DERELICK_MY_VOLS]

Good work dude. Interesting stuff. With that said what is life like in The Matrix?

Just kidding good stuff.

 

[Face]

[/B]

You say we "had a CTA of 86.2" the tense of this sentence confuses me. Can you explain this more.

Having done this very same thing for a couple years, I too am interested to see what your results are on PVM/CTA ratio. What I have found, to date, is that using this sort of analysis alone does not tend to provide a stable way to predict score differentials, just wins and losses. For example, Texas A&M out ranked SCAR by about 3 points, but won by about 10 points per talent difference. Or, Oregon vs. UT last year, both teams had the same talent average as did Oklahoma and Notre Dame. Neither score was particularly close. I understand the problem with trying to make a rule out of the exceptions, but my data has a far lower correlation of talent difference to score difference.

Sorry for the confusion. I meant to say, "Against USU, we had a..."

I made an edit.

I agree that the precision of the analysis probably will not 100% correlate to the score. I'm putting these together as more of a bench mark than a prediction. I include my prediction for fun :hi:

 

[Volview]

As big as the gap is in over all talent, it's the talent gap of the 2nd teams that will cause the game to get out of hand in the 2nd half. Much like the USU game.

I know it would take more research...but 2nd team star level might be even more useful.

 

[daj2576]

I will do one then and I'm just as interested. After we see a 2nd game, I'll probably try to work in a home field advantage adjustment to the numbers as well.

I am interested to see how that comes out.

I haven't done comprehensive studies on quantifying home field advantage (Phil Steele has some data, Neyland is a 4.5 FYI). What I have found is that ignoring any other factors, the win rate for better talent is about 70%.

Having done this for a couple years, I have been urged to try to adjust for many factors. What I have found is that the more complicated you tend to make this analysis, the less stable it becomes. It is my observation that much of what we believe actually effects wins doesn't. It's like saying that UT is 0-2 in the greys, so while there is a correlation, let us try to adjust for it like it is causative.

Home field advantage is a different beast altogether. For some interesting reading on the topic, I suggest the books "Scorecasting", "Mathletics" and "Stumbling on Wins."

http://www.amazon.com/Scorecasting-Hidden-Influences-Behind-Sports/dp/0307591808

Mathletics: How Gamblers, Managers, and Sports Enthusiasts Use Mathematics in Baseball, Basketball, and Football: Wayne L. Winston: 9780691154589: Amazon.com: Books

http://www.amazon.com/Stumbling-Bon...=1409850085&sr=1-1&keywords=stumbling+on+wins

 

[Face]

Do you factor in players that where high star players that have left the team for one reason or the other. Our 2011 class does not seem that strong in reality Rivals 13.

No. That would make for a more accurate adjustment. Maybe when I have a little more time, I'll factor it in next week

 

[daj2576]

As big as the gap is in over all talent, it's the talent gap of the 2nd teams that will cause the game to get out of hand in the 2nd half. Much like the USU game.

I know it would take more research...but 2nd team star level might be even more useful.

I really hate to continue stepping all over the OP, but I have similar data readily available. Apologies OP, I love this stuff.

Here is the total roster, showing the change from last year to this year.



Here is the change in the two deep from last year to this year. (the numbers to the right aren't stars but rivals scores, stars are denoted by "*").



Neither of these charts leads me to believe that the drop off in talent is worse than last year. In fact, I believe that the talent is actually better AND deeper in general.

 

[daj2576]

No. That would make for a more accurate adjustment. Maybe when I have a little more time, I'll factor it in next week

This is always a huge criticism of this system. First it should be pointed out that without adjusting for attrition that the win/loss correlation is still exceedingly high. For instance, last year Tennessee's schedule was predicted correctly 10 out of 12 games. That is 83%.

Before last season, I took the time to adjust out all of the attrition across all SEC teams. It was labor intensive and changed nothing that would effect Tennessee's schedule.

The problem is that there was enough of a gap between Georgia, Tennessee and SCAR that Tennessee could have a great deal of attrition and it would not effect talent averages in a way that would make them fall below SCAR, just as UGA would require a great deal of attrition to fall below UT. The wiggle room is relatively large on an 85 man roster, to lose players, and not change the overall average.

 

[daj2576]

Sorry for the confusion. I meant to say, "Against USU, we had a..."

I made an edit.

I agree that the precision of the analysis probably will not 100% correlate to the score. I'm putting these together as more of a bench mark than a prediction. I include my prediction for fun :hi:

I hope you find a more meaningful correlation than I did. I would love to see it. From the data I have reviewed, I just don't think we can get there from here.

I will stop interjecting myself. If you ever want to discuss what you have found, or compare notes, email me at derek@mybloodisorange.com.

 

[StarRaider]

Here is a spreadsheet of the SEC using these sorts of talent averages as a guide. So far this season (all of two games) the talent averages have predicted all of the games correctly. By the end of the year, roughly 70% of the games will likely be predicted by these talent averages.

https://docs.google.com/spreadsheet/ccc?key=0AkwyQgwl-hyfdEU2WkFralNPVTE1VG5xa0Ffak5OMmc&usp=sharing

yes Mr. White math! Nice work!

 

[Sheik Yerbouti]

My guess is as useful as these numbers are against lower tier schools, when we get into top 10 teams, things like experience are going to become a larger factor. I think our raw talent easily matches the Sooners or Georgia....but experience wise, and state of program still needs to catch up.



This is an incredibly talented young team, but it may need a season in the SEC oven before it's really done. Then again it is college ball......shizz happens! Some kids grow up faster then others.

I bet it would be possible to come up with a factor to account for experience. I wonder what it would be though? Is a three-star with three years of college experience and weight training 30% better than when he was a freshman? More? Less? Inquiring minds want to know.

 

[VolsnRavens]

After all of my mathematical equations factored in for talent and star quality and what the players eat for breakfast on Saturday at Calhouns. My final analysis is....

We gone whoop dey azz.

 

[Orange_Crush]

I'm interested to see how these numbers look against Oklahoma. I hope you'll post this for that game as well.

*Nick beat me to it.

Why wait til game week? I can't wait.

 

[wildorange]

I think I will just watch the games, all the projections and everybody's predictions are kind of redundant. Nothing against the poster with these stats at all, just a overall feeling. Kinda like espn giving percentages for a win, it really is kind of silly or hold a lot of merit. Really how did they come up with 56% chance that Utah State beats Tennessee, we are young but if you looked at the rosters and skill level I just don't get it. But I realize that predictions are just in peoples nature, but sometimes they get old. GO VOLS!!