So I read this Ringer article NFL Analytics Revolution And it got me wondering if:
A) Do we have an analytics person on our staff?
I know part of the GAs job is to chart plays from the all 22 film. After that data is entered who knows where it goes from there.
B) What teams in the NCAA use it? I have to think that Bama, Ohio State and Clemson have it some how.
The only time I have ever heard it being discussed was Army's defensive coordinator utilizes it on some level.
I'll add one more question. How do you feel analytics will impact College Football in the future? Personally schools have data scientists available, why not use the resources at your finger tips?
Forums:
DISCLAIMER: Forum topics may not have been written or edited by The Key Play staff.
Comments
Fuente is an analytics guy.
Edit: he brings it up all the time during GiT games, although now Chinballs is gone, maybe we won't hear about it.
He was too analytical during the GT games in my opinion.
The litany of ridiculous coaching blunders in the 2017 GT game gives me nightmares.
Yep. That game was the end of the honeymoon period.
I had originally typed up a long list of just the 2017 GiT game, where analytics failed. I chose to change the post.
Fuente is 0-3 against GT, so....
My honest gut feeling is I have a hard time seeing it happen at the college level. "Good" Data Scientists are just too expensive and hard to find. From reading the article you get the sense that even the NFL teams talking about it still aren't doing a whole lot.
No one is using analytics, they say they are, but then they make the wrong choice all the time. It seems like analytics means you can go for it on 4th down.
I finally got time to read the article. There are some cool things they talk about and I am finally glad to see things advancing, though I have always thought football will be the hardest since it has more variables per play. And as they said small sample size.
Metrics for drafting makes a huge amount of sense financially. I understand why people are working on that. I am glad to see the player work out analytics too.
My issue, which the article briefly touched on , is a much easier implementation. Play calling in the NFL sucks. How do coaches suck at clock management, cant they hire some one whose sole job is to do that, but instead we have a sheet of paper to remind everyone once not to make a mistake, but why would you have an analytics program that would just tell everyone what to do. There is probably some rule against that for the NFL, but if you better choices allow better analytics. When is it too early to go for two, when to go for it on 4th down based on your offense, your kicker, the score, the time, etc.
The motion tracking sound awesome, especially if you can update everything from the game in real time. Heck it could show a DE playing to far inside so you have an easy time sealing him off for an outside run. Or a DB is running slower than other games.
Also how was some one allowed to call a play in the super bowl that the team could execute. Where is the analytics team firing that coach. The game is on the line, you have one of the best power backs in the NFL and you call a play you have never successfully ran. You arent using analytics if you cant review your own play calling to see what works and what doesnt, this should be basic.
Probably the biggest problem with analytics and football is the lack of uniformity among teams. So given that you have two options. Only use numbers gathered from your own team which has a small sample or figure out how to adjust numbers across a larger number of teams (maybe all teams). Regardless, either case probably results in every call being a gamble. A good example is Paul Johnson's wrinkle. That wrinkle would not be applicable to any previously collected numbers because he had never used it before. Another one? Assuming Ohio State was using analytics in 2014. The bear front for which there are no numbers. There are so many nuances and variations that it would seem nearly impossible to control for. What you're really looking at then is a raw statistic. Basically gambling.
Football has too many critical real time decisions/adjustments that must be made as the snap approaches. You'd need a HUD (heads up display) type system, so you could signal all 11 players on your team simultaneously, to truly take advantage of the benefits of analytics.
Then you'd introduce "anti-analytic" processes, to deceive the models that predict your team's most likely behavior, for the play.
Kinda OT but what was the deal with the VR headset or something that the qb's were wearing a couple years ago? Are they still don't that?
Gimmick I feel like. Haven't heard anything about it since like you said.
A guy on the professional teams forum was looking for high schools coaches who wanted to test the VR/AR headsets he was working on.
Give it 10 years and AR/VR will be everywhere.
I'm sure to some level every team has someone in charge of whatever they call "analytics". This is really a question of how in-depth it goes and is utilized. To one team, analytics might just mean the guy who makes the sheet telling the coach whether to go for 2. To another, it might be player evaluations for every opponent and trends of play calling.
I think looking to a computer to tell you exactly which play to run is a bad idea unless things get really advanced. You're always going to deal with small sample sizes there, and often the most successful plays will be those that you have never run before but take advantage of some tendency by the other team. Maybe a computer will someday help identify those potential tendencies, but even then it's hard to imagine it's not simply prioritizing what coaches watch on film. that would be great value - imagine how much film there is on an opponent and having to watch for so many different things. Now imagine an algorithm telling you ahead of time that a tendency it has identified with high confidence is a certain corner playing outside leverage against tall wide receivers much more often than shorter ones. It also spits out a list of plays by opponent, down and distance and time on clock, and the height of the opposing receiver along with whether he played outside leverage. That gives your coaches some very specific things to watch for and exploit, and enhances good film review rather than trying to replace it.
Also imagine the motion tracking - a WR has been hobbled but claims to be fine. Would you rather trust your eyes trying to watch him on the field, or a highly-accurate motion tracking system letting you know that his top end speed is no longer faster than every DB on your next opponent's team, but rather is slower than both starting CB's and a S. Maybe you don't bench him, but maybe this info helps you decide to play him in the slot and not send him on deep routes. That 3% slow down in speed may not be immediately obvious in practice but would have been really exposed in a game where suddenly your team can't spread the field.
I agree a computer calling a specific play is not realistic for reasons you said. However, today a computer could know the score, time left in the game, average drive time, defensive efficiencies and tell you whether to go for it on 4 th or punt. Also, doesn't need a computer, but some one needs to tell coaches when to go for two. Last bowl season a team scored with like 2.5 minutes left and was up by 7 before the xp. And the kicked the XP instead of going for 2. So down 7 points the other team scores, goes for 2, forces OT and I think wins it. The team lost because of basic math.
Except, 2-pt conversions are converted at less than 50%. I don't know about the specifics of those teams, and so no team-specific data, but at the most general level of analytics, you would actually take the 8-pt lead, and rely on your D to get a stop on 4th down or on a 2-pt conversion. If both of those fail, still have OT.
You can find an "analytic" data point that says VT's second half offense was decent in 2018 and our hoops team has good size. It's total grain of salt for me. I'd rather be able to line up under center andget 2 tough yards again vs. studying our offensive S&P.
why would any good analyst select data points to prove a preconceived conclusion? that's the antithesis of analytics.
They are preparing statistical data for use in a law suit? Or for a sales pitch for their company? Or to justify a raise? Or to prove their opinion is correct on the internet? Maybe only the last applies to football, but they all answer the question.
Fair point. Not every analyst is paid to reach true actuality.
These are all sadly true, especially the law suit part. It is sad how facts and data are exploited in the legal world to argue your legal conclusion. It is also said that this reasoned to be "fair" because there are two sides who are equal in the blind eyes of justice (even if 1 side is a corporation that spends $1M in legal analytics to argue their point and the other side is a poor individual who can't even afford their own attorney).
How is there a choice between lining up and getting two yards and reading an S&P+ (or some other metric) number? Is anyone suggesting Justin Fuente pull players off the practice field and go stand around a computer screen?
Rating systems like S&P and FPI exists to allow anyone to compare any two teams. They don't exist for an NCAA head coach to go scout a team. What information would be relevant to Fuente is very different than what is relevant to a casual fan, and I've never seen anyone credible suggest even remotely that analyzing data should, in any way, replace practice and coaching.
And just because there are bad metrics out there doesn't mean all metrics are useless. Baseball teams had a bunch of old school scouts as well who made broad, baseless statements about sabermetrics until you couldn't make the playoffs without utilizing them (believe me, I'm a Braves fan). Most of those scouts are either unemployed or marginalized within their organizations because they shunned more information and became ineffective at their jobs because of it. Everyone - GM, coach, player, whoever - gets to decide what metrics provide them with valuable information and which don't, but the ones who ignore all metrics tend to get left behind. For a player a metric might be their 40 time or max squat. It doesn't have to be complicated.
I see the best use of analytics as applying to the off season. Build a playbook based on what plays are most effective against your opponents, and cull out those least likely to be effective, looking both at when you ran them and when others did. Then compare results on what you ran offensively and defensively to how others did on the same plays to identify what plays to practice more often, or what to change on the plays.
The number of variables involved and how quickly they can change make play to play analytics unpractical to implement, and the best coordinators at calling plays seem to those who can watch game film and tell you what the analytics should say if done correctly.
Like in poker, knowing the odds is always a good thing.
Doesn't mean you can't buck them if it's indicated, but you should still know them.
Have to be careful about analytics, though. As much as I like it in baseball, where its a pretty basic cause and effect sport, all things considered, but when you are talking about more team sports, like hockey, basketball, or football, the over reliance on analytics could end up holding you back. I mean, if you look at pure analytics numbers, the Carolina Hurricanes have been the best team in hockey the last 2-3 years. They haven't made the playoffs in the last 9, so analytics can only get you so far.
For football and hockey I don't think the advanced analytics are anywhere close yet. Basketball is starting to get there, but not quite. Baseball has the fewest moving parts, so it was the easiest. Basketball is getting there with player rest and shot selection. But Hockey and Football have a lot of ground to cover. Football will get there because the money is there and it only takes one team. I am not as familiar with sure hockey, but if the numbers says the best team has failed to make the playoffs for 9 years then you have the wrong numbers.
I think analytics would be best in advancing concepts of the game, rather than in-game adjustments. Determine what types of offenses/defenses, offensive sets, play-calling, and situations thereof are yield the best success. Build upon it and expand/exploit successes. Be inventive on the information yielded.
Bud is one of the best at this using just film and team personnel. I would imagine good analytics, he would further boost his ability to adapt year after year.
This may include some bias, as my job is entirely data-focused, but I see a huge benefit in incorporating available data if it means even a slight advantage. As has been said here and in the article, football will never be baseball in terms of all-encompassing data points. However, not utilizing something that may give even the smallest of competitive advantages in a sport where small advantages can make all the difference is poor management.
I think an important thing to remember, and one commentators/analysts often misconstrue, is that just because a particular play call or a decision doesn't work does not mean that it was the wrong call. Choosing to go for it on 4th down rather than punting or opting for a two point conversion over an extra point are often criticized when they don't work out. If all available data is pointing to a higher win (or score) probability by doing so than the other option, then a coach should absolutely do it. It may not work, but your team would be in a worse position if you went against the advanced data - it is a probability after all.