https://www.washingtonpost.com/sports/as...4ef3a812ae

Talks about the Astros and using analytics to an extreme to win the World Series.

Its only a matter of time before these things filter into football. Innovation will probably happen in the colleges before it hits pro football. Right now they only look at a limited set of figures.

(11-02-2017 01:31 PM)bullet Wrote:  https://www.washingtonpost.com/sports/as...4ef3a812ae

Talks about the Astros and using analytics to an extreme to win the World Series.

Its only a matter of time before these things filter into football. Innovation will probably happen in the colleges before it hits pro football. Right now they only look at a limited set of figures.

This isn't mentioned in the article, but it's interesting to me that another Houston team (the Rockets) is also probably the heaviest user of data analytics in its sports of basketball.

What will be the challenge for football is figuring out whether there is actually any data that has strong predictive value and, even if that exists, are there enough data points to create a large enough sample size to analyze. Both baseball and basketball have the benefit of long seasons with countless data points that can reasonably be isolated to individual players. I'm sure that someone will figure it out for football at a certain point, but it's definitely a different beast because there aren't as many data points and they're much more difficult to assign to individual players (outside of the quarterback). I do think it may need to come from the pro level, though (with a hypothetical caveat below). Colleges are good at experimenting with on-the-field schemes and techniques. However, data analytics require dedicated data science professionals that are expensive and in high demand in *every* industry. That would likely require pro-level financial investments.

Of course, athletic departments at schools with great computer science departments like Stanford, Cal, Illinois, Texas and Michigan could grab some professors and talented students at a lower cost and create some type of football data analytics research lab. That could be a win-win for athletic departments while giving CS students some high profile undergrad experience on fun projects.

Yeah football would be much more complex from a predictive analytics standpoint, thus no surprise there hasn't been as much analytics use in football yet.

Baseball is probably the easiest due to an extremely large data set and the individual match up of a pitcher vs a hitter that can then be overlaid with the different situations (i.e. 1 out, 2 outs. men on base or not, pitching ahead in the count vs. behind, etc...). Football would be a real monster to try to model.

what's funny about the series though was they were managed far less by the analytic book than LA was. LA did everything primarliy by the book. And they got burned to some degree during the series as a result.

(11-02-2017 01:58 PM)Frank the Tank Wrote:  

(11-02-2017 01:31 PM)bullet Wrote:  https://www.washingtonpost.com/sports/as...4ef3a812ae

Talks about the Astros and using analytics to an extreme to win the World Series.

Its only a matter of time before these things filter into football. Innovation will probably happen in the colleges before it hits pro football. Right now they only look at a limited set of figures.

This isn't mentioned in the article, but it's interesting to me that another Houston team (the Rockets) is also probably the heaviest user of data analytics in its sports of basketball.

What will be the challenge for football is figuring out whether there is actually any data that has strong predictive value and, even if that exists, are there enough data points to create a large enough sample size to analyze. Both baseball and basketball have the benefit of long seasons with countless data points that can reasonably be isolated to individual players. I'm sure that someone will figure it out for football at a certain point, but it's definitely a different beast because there aren't as many data points and they're much more difficult to assign to individual players (outside of the quarterback). I do think it may need to come from the pro level, though (with a hypothetical caveat below). Colleges are good at experimenting with on-the-field schemes and techniques. However, data analytics require dedicated data science professionals that are expensive and in high demand in *every* industry. That would likely require pro-level financial investments.

Of course, athletic departments at schools with great computer science departments like Stanford, Cal, Illinois, Texas and Michigan could grab some professors and talented students at a lower cost and create some type of football data analytics research lab. That could be a win-win for athletic departments while giving CS students some high profile undergrad experience on fun projects.

I think football will have several challenges when it comes to translating analytics to actual performance.

One being that there is no minor league system similar to baseball. The only thing they have is the college game and the college game is a far cry from the NFL. Too many read-option systems and spread offenses that don't require the QB's (or LB's) to make any reads. Everything is done by the coaches.

Another is that the careers of many NFL players are so short that there isn't a sufficient sample size to draw any conclusions from.

How many players came out of the combines with Tarzan numbers only to play like Jane?

I'm sure there is some formula that could work for football, but it seems to be a different beast.

I tend to be middle of the road on the whole thing. Sure numbers play into any decision, but I think too many of the "numbers" guys forget, as the great Hawk Harrelson would say, TWTW.

(11-02-2017 01:58 PM)Frank the Tank Wrote:  

(11-02-2017 01:31 PM)bullet Wrote:  https://www.washingtonpost.com/sports/as...4ef3a812ae

Talks about the Astros and using analytics to an extreme to win the World Series.

Its only a matter of time before these things filter into football. Innovation will probably happen in the colleges before it hits pro football. Right now they only look at a limited set of figures.

This isn't mentioned in the article, but it's interesting to me that another Houston team (the Rockets) is also probably the heaviest user of data analytics in its sports of basketball.

What will be the challenge for football is figuring out whether there is actually any data that has strong predictive value and, even if that exists, are there enough data points to create a large enough sample size to analyze. Both baseball and basketball have the benefit of long seasons with countless data points that can reasonably be isolated to individual players. I'm sure that someone will figure it out for football at a certain point, but it's definitely a different beast because there aren't as many data points and they're much more difficult to assign to individual players (outside of the quarterback). I do think it may need to come from the pro level, though (with a hypothetical caveat below). Colleges are good at experimenting with on-the-field schemes and techniques. However, data analytics require dedicated data science professionals that are expensive and in high demand in *every* industry. That would likely require pro-level financial investments.

Of course, athletic departments at schools with great computer science departments like Stanford, Cal, Illinois, Texas and Michigan could grab some professors and talented students at a lower cost and create some type of football data analytics research lab. That could be a win-win for athletic departments while giving CS students some high profile undergrad experience on fun projects.

That was what I was thinking. Some tech savvy coach at Stanford at some point bringing in the computer science department. It will work better in the pros because the number of players are more limited and they don't turn over as much. But a lot of innovation in football seems to start at the college level where coaches generally have 3 years to get things done without getting fired.

There are a lot of possible metrics in recruiting, alignments, play calling and player rotations. I get the impression its all very crude and simple right now. 40 speeds, bench press, going for 2 at a certain point in the game....

I remember there being a coach in Arkansas who took a tiny private high school to state dominance with a lightning-paced ultra-aggressive style of play in which they almost always went for it on 4th down and attempted onside kicks. He had some big database that showed the odds of making either of them and the potential points upside weighed versus the potential downside if they failed, though obviously that hasn't caught on in mainstream football.

In a way, the spread offense is a bit of 'Moneyball' in that its really spread like wildfire among smaller programs trying to gain a competitive advantage versus much wealthier programs. No one could beat Oklahoma or OSU at the wishbone or pro formation football where they could just get the biggest guys out there and manhandle you (especially before weight training was as advanced as it is today), so you spread the field and throw a bunch of confusing formations at them and you have a chance.

(11-02-2017 09:16 PM)Love and Honor Wrote:  I remember there being a coach in Arkansas who took a tiny private high school to state dominance with a lightning-paced ultra-aggressive style of play in which they almost always went for it on 4th down and attempted onside kicks. He had some big database that showed the odds of making either of them and the potential points upside weighed versus the potential downside if they failed, though obviously that hasn't caught on in mainstream football.

In a way, the spread offense is a bit of 'Moneyball' in that its really spread like wildfire among smaller programs trying to gain a competitive advantage versus much wealthier programs. No one could beat Oklahoma or OSU at the wishbone or pro formation football where they could just get the biggest guys out there and manhandle you (especially before weight training was as advanced as it is today), so you spread the field and throw a bunch of confusing formations at them and you have a chance.

I'd heard of that high school coach but I think the application of that kind of analysis is VERY limited. For example, in college, do you use the average 4th down FBS conversion % in a game between Ohio State and Kent State? LOL

(11-02-2017 09:16 PM)Love and Honor Wrote:  I remember there being a coach in Arkansas who took a tiny private high school to state dominance with a lightning-paced ultra-aggressive style of play in which they almost always went for it on 4th down and attempted onside kicks. He had some big database that showed the odds of making either of them and the potential points upside weighed versus the potential downside if they failed, though obviously that hasn't caught on in mainstream football.

In a way, the spread offense is a bit of 'Moneyball' in that its really spread like wildfire among smaller programs trying to gain a competitive advantage versus much wealthier programs. No one could beat Oklahoma or OSU at the wishbone or pro formation football where they could just get the biggest guys out there and manhandle you (especially before weight training was as advanced as it is today), so you spread the field and throw a bunch of confusing formations at them and you have a chance.

Actually the wishbone was more about speed. Texas, the national champs in 1969, only had one offensive lineman over 210 (small even in that era). They had to get out and block. But your point is the same-they could recruit the speed and the great running QB. That's why schools like Rice in the wishbone era used pro style passing offenses.

Coach Kelly is still at Pulaski Academy.
The problem with his approach is fan and media buy-in.
You can find several cases where coaches went for it on fourth down for example and failed to convert even though it was statistically the right call, the fans and the media were livid.
Take a semi-common situation. Team is down 15 and scores a TD. You absolutely should go for two. If you are successful you kickoff knowing you need a TD and PAT to tie. If you fail you kickoff knowing you must score two more times. BUT if you kick the PAT you still need a TD and 2pt conversion, in other words you need a TD and then need to convert a single play with roughly a 45% to 52% chance of being successful. If you get the second TD and fail on the conversion you need another score.

You are better off playing knowing whether you need one or two scores.

Invariably if a team tries the early two and fails the complaint is the coach went for two too early.

Or consider a team going for it on fourth down at the opponent 40 while leading by 4 to 7 points. Fail and there is a universal cry that the coach should have punted so the opponent has to go further. Probability is a punt makes a 20 to 30 yard difference. That just doesn't make a big difference in the probability of the opponent converting, but if you get the first down, your probability of winning shoots up because you are holding the ball longer and increased your chance of scoring.

Unless fans and media buy in or the coach is just so successful he can withstand the heat, following probability over conventional wisdom can get you fired.

Back to Coach Kelly, he is only all onside, all go for it when he has had the superior teams. Before their improved enrollment push for athletes kicked in, he regularly kicked deep and punted. A school like Bama could adopt the philosophy and increase their success rate even more. I don't think Texas State is likely to add wins.

(11-03-2017 09:04 AM)arkstfan Wrote:  Coach Kelly is still at Pulaski Academy.
The problem with his approach is fan and media buy-in.
You can find several cases where coaches went for it on fourth down for example and failed to convert even though it was statistically the right call, the fans and the media were livid.
Take a semi-common situation. Team is down 15 and scores a TD. You absolutely should go for two. If you are successful you kickoff knowing you need a TD and PAT to tie. If you fail you kickoff knowing you must score two more times. BUT if you kick the PAT you still need a TD and 2pt conversion, in other words you need a TD and then need to convert a single play with roughly a 45% to 52% chance of being successful. If you get the second TD and fail on the conversion you need another score.

You are better off playing knowing whether you need one or two scores.

Invariably if a team tries the early two and fails the complaint is the coach went for two too early.

Or consider a team going for it on fourth down at the opponent 40 while leading by 4 to 7 points. Fail and there is a universal cry that the coach should have punted so the opponent has to go further. Probability is a punt makes a 20 to 30 yard difference. That just doesn't make a big difference in the probability of the opponent converting, but if you get the first down, your probability of winning shoots up because you are holding the ball longer and increased your chance of scoring.

Unless fans and media buy in or the coach is just so successful he can withstand the heat, following probability over conventional wisdom can get you fired.

Back to Coach Kelly, he is only all onside, all go for it when he has had the superior teams. Before their improved enrollment push for athletes kicked in, he regularly kicked deep and punted. A school like Bama could adopt the philosophy and increase their success rate even more. I don't think Texas State is likely to add wins.

OT, but I'd disagree on the 15 point scenario. If you fail, the psychie of the team is destroyed. If you kick, its still a one TD game.

(11-03-2017 09:07 AM)bullet Wrote:  

(11-03-2017 09:04 AM)arkstfan Wrote:  Coach Kelly is still at Pulaski Academy.
The problem with his approach is fan and media buy-in.
You can find several cases where coaches went for it on fourth down for example and failed to convert even though it was statistically the right call, the fans and the media were livid.
Take a semi-common situation. Team is down 15 and scores a TD. You absolutely should go for two. If you are successful you kickoff knowing you need a TD and PAT to tie. If you fail you kickoff knowing you must score two more times. BUT if you kick the PAT you still need a TD and 2pt conversion, in other words you need a TD and then need to convert a single play with roughly a 45% to 52% chance of being successful. If you get the second TD and fail on the conversion you need another score.

You are better off playing knowing whether you need one or two scores.

Invariably if a team tries the early two and fails the complaint is the coach went for two too early.

Or consider a team going for it on fourth down at the opponent 40 while leading by 4 to 7 points. Fail and there is a universal cry that the coach should have punted so the opponent has to go further. Probability is a punt makes a 20 to 30 yard difference. That just doesn't make a big difference in the probability of the opponent converting, but if you get the first down, your probability of winning shoots up because you are holding the ball longer and increased your chance of scoring.

Unless fans and media buy in or the coach is just so successful he can withstand the heat, following probability over conventional wisdom can get you fired.

Back to Coach Kelly, he is only all onside, all go for it when he has had the superior teams. Before their improved enrollment push for athletes kicked in, he regularly kicked deep and punted. A school like Bama could adopt the philosophy and increase their success rate even more. I don't think Texas State is likely to add wins.

OT, but I'd disagree on the 15 point scenario. If you fail, the psychie of the team is destroyed. If you kick, its still a one TD game.

Depends on the clock. If you only have time for one more possession? Sure I'll be the psychological argument but usually when I see it happen there is time for more possessions and you need to know you time management strategy.