Could a "Moneyball"-type analysis help BYU football?

So, I'm basically a football fanatic in a nerd's body. Thankfully, I've also got nerd skills as well, so I have a way to pay the bills. But in reading some articles about our prep for the Wazzou game, and how we're going back to Bronco's experience with TT's offense, as well as reaching back to some old Wyo. games since WSU's DC was at Wyo. from '03 to '08, I started to wonder how predictable coaches are from year to year. From there, I wondered how much information on opponents' tendencies coaches generally have. Is it just film? Do they have play calls broken do by down and distance? Any statistical analysis?



The reason I wonder is that I did some work in my computer science studies with machine learning and AI that involved Bayesian statistics, and it seems like if you had a gigantic database of play calls with information about who was coaching, down and distance, time remaining, score, personnel available, previous results in similar situations, you could throw the right machine learning algorithms at it and do a very precise job of getting inside a play caller's head. If he's fairly diverse, it may not benefit you any. But if there are strong tendencies, it could really help your gameplan creation. If done correctly, you could even have a program that takes live game data, and spits out plays it thinks you should call and gives probabilities for what might happen, so that the OC / DC have more than just their gut behind their play calls. For example, such a program would have been screaming at Anae in '08 against Utah to run Harvey over and over again, just based on previous success in that game.



Anyway, I guess what I'm asking is, 1) does anyone know the extent of information that coaches currently have, and 2) is there any value in putting something like this together? I've always dreamed of making some kind of contribution to BYU football. When I couldn't get myself over 135 lbs., I figured that was limited to screaming my guts out in the stands and buying tickets. But I'm hoping that's not true. Could this form of "Moneyball" help our team?

I would hope they have that kind of analysis. At a minimum, I hope they have access to a spreadsheet/database similar to what I use.



Moneyball would be tough to implement at the college level, since you're talent pool is essentially from high school. That success doesn't necessarily translate.



In baseball, you're using it to evaluate other major leaguers and you have a good understanding of how AAA success might translate.

Wasn't Anae known for his analytics or at least his analytic mind. He struggled with the intuitive and creative mind--or at least what was staring him in the face (Unga, for example).

Some coaches run their way regardless of who they play. It is about execution of plays and if they do that they win. Where has that sounded familiar? :-)



Isn't the challenge sort of a multivariate problem? You can look at correlation but not causation. And by the time you figure that out, your opportunity to apply the analytic is over. Dr Phil might have a more effective approach: "how's that workin for ya?"

@Spiff, you're right, the Moneyball analogy breaks down when you're not trading players around and dealing with amateurs and students. I was mostly just thinking of the play calling aspect. My hypothesis is that with enough information, your play calling could offset serious disadvantages in speed and athleticism.



@dmcougar, There's a ton of variables to consider. The first step would be to figure out which ones matter, which machine learning can help with in the right hands. Once the data is there and you know what algorithms give you the results you want, though, modern computers are easily capable of taking live game data, adding it to the database, and giving you updated results as you play the game.



And I don't know about Anae's analytical abilities, but I can think of a few examples that indicate to me that could have used a little sharpening... ;)

It would seem the NCAA is a step ahead of me and has banned the use of any type of computer during the game. A shame... ;)

I do think a program like BYU needs to depend more on superior scheme than on superior athletes. We'll get the elite players every now and again but they're an occasional luxury at best. Better preparation, X's and O's, etc. is the key to beating the higher level teams that we can't match in terms of raw talent.

I'd love to see a coach holding a chart in his hand with a matrix of situations of when to call time outs for time management at the end of games and another one for when to go for 2.



(Think Utah game 2010 we kick an extra point to go up by 13 when the other team hasn't moved the ball. So we protect ourselves from them scoring 3 times with a TD and two field goals but open ourselves up to losing to a two scoring drive scenario. We lose by 1. 2011 Texas: Same thing, we kick an extra point to go up by 13 on a team not moving the ball on us. Lose by 1. Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier. The TCU game had us kicking an extra point to pull within 15 last year as well.)

Stirfry wrote:
I'd love to see a coach holding a chart in his hand with a matrix of situations of when to call time outs for time management at the end of games and another one for when to go for 2.



(Think Utah game 2010 we kick an extra point to go up by 13 when the other team hasn't moved the ball. So we protect ourselves from them scoring 3 times with a TD and two field goals but open ourselves up to losing to a two scoring drive scenario. We lose by 1. 2011 Texas: Same thing, we kick an extra point to go up by 13 on a team not moving the ball on us. Lose by 1. Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier. The TCU game had us kicking an extra point to pull within 15 last year as well.)

You're assuming, however, that we convert the 2-pt plays.

Stirfry wrote:
Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier.

If Riley's tipped pass wasn't caught it would have been 1st & goal from inside the 5-yard line.

Schmoe wrote:

I'd love to see a coach holding a chart in his hand with a matrix of situations of when to call time outs for time management at the end of games and another one for when to go for 2.



(Think Utah game 2010 we kick an extra point to go up by 13 when the other team hasn't moved the ball. So we protect ourselves from them scoring 3 times with a TD and two field goals but open ourselves up to losing to a two scoring drive scenario. We lose by 1. 2011 Texas: Same thing, we kick an extra point to go up by 13 on a team not moving the ball on us. Lose by 1. Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier. The TCU game had us kicking an extra point to pull within 15 last year as well.)

You're assuming, however, that we convert the 2-pt plays.

And lose by 2 instead of 1 if we don't? (Assuming the other team, which hasn't put 1 drive together let alone 3, scores 3 times and two of those times are field goals?) I'm not saying it's fool proof and it's easier in hindsight of course but it's way under used.

SpiffCoug wrote:

Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier.

If Riley's tipped pass wasn't caught it would have been 1st & goal from inside the 5-yard line.

And had 11 seconds...with the option to kick or be required to punch it in. I have no idea how many timeouts we had but I'd guess it wasn't many. Although maybe if we could use previous experiance to know when to use timeouts, say with a matrix, we'd have managed the time better at the end of the game? :)

Stirfry wrote:

Stirfry wrote:
Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier.

If Riley's tipped pass wasn't caught it would have been 1st & goal from inside the 5-yard line.

And had 11 seconds...with the option to kick or be required to punch it in. I have no idea how many timeouts we had but I'd guess it wasn't many. Although maybe if we could use previous experiance to know when to use timeouts, say with a matrix, we'd have managed the time better at the end of the game? :)

Couldn't possibly run more than one pass play from inside the 5 in 11 seconds.

Schmoe wrote:

SpiffCoug wrote:

Utah State 2011: We score to come within 5 with 10 minutes left and kick an extra point to pull within 4. If Reilly's tipped pass isn't caught with 11 seconds left we would have been in trouble. We could have kicked a field goal if we went for 2 earlier.

If Riley's tipped pass wasn't caught it would have been 1st & goal from inside the 5-yard line.

And had 11 seconds...with the option to kick or be required to punch it in. I have no idea how many timeouts we had but I'd guess it wasn't many. Although maybe if we could use previous experiance to know when to use timeouts, say with a matrix, we'd have managed the time better at the end of the game? :)

Couldn't possibly run more than one pass play from inside the 5 in 11 seconds.

True. We wouldn't want the option to kick anyways. We should always plan on our backup quarterback driving 90 yards. No reason to make it so he only has to drive 60 just incase.

redneckjedi wrote:
So, I'm basically a football fanatic in a nerd's body. Thankfully, I've also got nerd skills as well, so I have a way to pay the bills. But in reading some articles about our prep for the Wazzou game, and how we're going back to Bronco's experience with TT's offense, as well as reaching back to some old Wyo. games since WSU's DC was at Wyo. from '03 to '08, I started to wonder how predictable coaches are from year to year. From there, I wondered how much information on opponents' tendencies coaches generally have. Is it just film? Do they have play calls broken do by down and distance? Any statistical analysis?



The reason I wonder is that I did some work in my computer science studies with machine learning and AI that involved Bayesian statistics, and it seems like if you had a gigantic database of play calls with information about who was coaching, down and distance, time remaining, score, personnel available, previous results in similar situations, you could throw the right machine learning algorithms at it and do a very precise job of getting inside a play caller's head. If he's fairly diverse, it may not benefit you any. But if there are strong tendencies, it could really help your gameplan creation. If done correctly, you could even have a program that takes live game data, and spits out plays it thinks you should call and gives probabilities for what might happen, so that the OC / DC have more than just their gut behind their play calls. For example, such a program would have been screaming at Anae in '08 against Utah to run Harvey over and over again, just based on previous success in that game.



Anyway, I guess what I'm asking is, 1) does anyone know the extent of information that coaches currently have, and 2) is there any value in putting something like this together? I've always dreamed of making some kind of contribution to BYU football. When I couldn't get myself over 135 lbs., I figured that was limited to screaming my guts out in the stands and buying tickets. But I'm hoping that's not true. Could this form of "Moneyball" help our team?

C'mon Bayesian statistics only works in the classroom. Stick to Classical statistics! ;)



Personally I would think the greatest use for statistics would be to predict play calling in various situations (including the chance of an onside kick when they are up 35pts on their opponent).

in baseball



1. you have 162 games which is a much greater sample AND where it works is over the course of an entire season where being .600 is monumental.



A. .600 in football only works in the Pac 12 when USC is down and the rest of the teams suck. .600 in a good conference gets you nowhere but the toilet bowl.



B. Moneyball might get you to the playoffs where anything can happen (at times) but there are no playoffs for the Cougars to attempt to get into.



2. Moneyball is also based on getting value out of players that were undervalued... so the most obvious place that BYU can make hay there is in recruiting, which sometimes they do.



3. Bill James never had to consider our honor code, or our academics.

imuakahuku wrote:

So, I'm basically a football fanatic in a nerd's body. Thankfully, I've also got nerd skills as well, so I have a way to pay the bills. But in reading some articles about our prep for the Wazzou game, and how we're going back to Bronco's experience with TT's offense, as well as reaching back to some old Wyo. games since WSU's DC was at Wyo. from '03 to '08, I started to wonder how predictable coaches are from year to year. From there, I wondered how much information on opponents' tendencies coaches generally have. Is it just film? Do they have play calls broken do by down and distance? Any statistical analysis?



The reason I wonder is that I did some work in my computer science studies with machine learning and AI that involved Bayesian statistics, and it seems like if you had a gigantic database of play calls with information about who was coaching, down and distance, time remaining, score, personnel available, previous results in similar situations, you could throw the right machine learning algorithms at it and do a very precise job of getting inside a play caller's head. If he's fairly diverse, it may not benefit you any. But if there are strong tendencies, it could really help your gameplan creation. If done correctly, you could even have a program that takes live game data, and spits out plays it thinks you should call and gives probabilities for what might happen, so that the OC / DC have more than just their gut behind their play calls. For example, such a program would have been screaming at Anae in '08 against Utah to run Harvey over and over again, just based on previous success in that game.



Anyway, I guess what I'm asking is, 1) does anyone know the extent of information that coaches currently have, and 2) is there any value in putting something like this together? I've always dreamed of making some kind of contribution to BYU football. When I couldn't get myself over 135 lbs., I figured that was limited to screaming my guts out in the stands and buying tickets. But I'm hoping that's not true. Could this form of "Moneyball" help our team?

C'mon Bayesian statistics only works in the classroom. Stick to Classical statistics! ;)



Personally I would think the greatest use for statistics would be to predict play calling in various situations (including the chance of an onside kick when they are up 35pts on their opponent).

The odds of whether someone will onsides kick while up 35 are dependent entirely on one variable: Is your coach a classless :utard: , or is he a regular human being? Non-zero in the first case, zero in the second. See how helpful this is already? :) What next?

I wish Bronco would step on the gas sometimes, but I am even more delighted that he would never do anything so classless.

Redneckjedi:

You may be having a "Steve Jobs" moment so don't be too quick to dismiss your idea. Just because the existing consumer does not see the value doesn't mean that that won't change. It may be that the end application is not specifically for coaches on game day but there may be other user applications that prove to be just as valuable if not more so.



If someone hasn't already done it or hasn't already started working on it there are possibilities for enhancing training and game preparation as players study "film" by making it more interactive. i.e., if I make this move or call this play what will be the result.



How about viewer participation during live games? Especially when combined with DVR. i.e., Okay this is the situation you make the call. The computer tells you how you did then you watch the actual results of what the coach or qb called.



All I am saying is that the end user application may not be fully known yet but that doesn't mean millions of dollars couldn't be made. Or if it ends up improving position mastery and and makes for smarter players, that would still bring the end result you are looking for.

I think the biggest upside would be that I would have to do a TON of "research" (watching and classifying college football plays full time). I would love for that to be my job!

Back to the original poster, yes, the football team does break down its opponents plays by down, distance, pass vs. run, and I believe situation (time remaining, etc.). When I was in the coaching football class, they showed us a little of the technology they use to break down game film and it showed down and distance. Gary Crowton brought the technology with him. It probably is still in use, but I'm not sure it has all of the variables you might want to look at. It certainly doesn't have the applications you and others have mentioned.



For this information, just cut me 10 percent of the stock you issue in your IPO. ;)