How common are these sports scientists around the league? I know the Seahawks have a couple but don't know much about other teams. I like that they are embracing it than shunning it.

 

You know, this sports science stuff is used far more in European soccer, and the stats stuff obviously a lot more in baseball. But from a pure scientific point of view, American football seems like the best new frontier for advancing this kind of sports science. You have a lot of stats, but often insufficient to really rate how good different players are (it's part-way there, but not far enough), and enough specialization of skills (think of all the different player positions) that you have ample opportunity to introduce measures that are likely to be useful for evaluating at least some positions.

 

,I'm very Interested in the obtaining of optimal performance by our players.. I've talked about it for years how this should be a part of the game, these athletes are machines, you have to look at them like that, the body does the work.. In a billion dollar industry where inches and fractions of seconds can impact the win total and the local economy, it behooves any franchise with pro athletes to exhaust the ways to reach that level of body and cardiac performance, trusting them to get there during an offseason is bush league and a Neanderthal way of thinking, the majority do not have a clue how to get there, this is hard to say but a lot don't have the intellectual capacity to understand how.. Years of nutritional abuse, lack of education on the theories,ego stroking and just a lack of enforcement institutuinally has made them think the opposite of these theories.

There needs to be group meetings about this with our players, they need to have someone either go up or down to their level to explain the importance, the years where fat *** lineman or dlineman think I'll just sit down and add a ridiculous amount of caloric intake is the way to go is just stupid..

They all should have programs to follow year round while their in this league, of course there are times where you allow yourself to cheat and have fun, but the commitment has to be much more intense and focused..a lot of these players use training camp that starts in July as a crutch and training ground to get their out of shape asses in shape and I'm tired of it, that is valuable teaching time as it relates to the playbook and how the design of plays should be executed, if you have 3 or four players that are using camp as a way of getting in shape, cut weight, or working on stamina, they are truly being counter intuitive to the integrity of execution and to the rate of offensive efficiency.


These players are employees..their owners and coaches should have it written in their contracts, and they should be held to a much higher standard then they have been..if some players like JJ watt and Cam wake are doing it, then there is no excuse for the rest..none, everyone physique has its own unique potential, it may not all wind up looking the same when finished but that level of optimal performance can be achieved thru nutrition, training, and most important, strict regimented rules.

 

i guess ross wants to know whom to bet on every week. what a waste of money. money wouldve been better spent on hiring two grizzled veteran scouts. one an expert in oline and the other linebackers

 

I'm strictly talking about the sports performance and nutrition part..

 

Football has for years existed as a half-art, half-psuedoscience largely because it's an incestual good ole boys club. I'm glad they're hiring people with a new take on things, there's plenty of room for professionals with some sort of methodology to intervene and improve aspects of the sport.

Much better than hiring a couple old hacks who say "Footbaw" because they've been doing it for decades.

 

There are too many tools to use in today's world to study football (and other sports) to simply not put them to use.

 

Still wouldn't take any analytics assessment over a good scout if it came down to making a call..I understand that scouts need to use them more for streamlining their job

 

i just think the statistical analysis of football is at its embryonic stages and offers nothing because of garbage in garbage out reality. the one area i'm okay with it in football is when studying tendencies of the opponent in terms of playcalling, formation, etc...if its used for other areas i dont think we're at the stage in its development where it will do any good

 

Great! Now all they need is a head coach.

 

I will, but would rather you tell me how you see all this fitting into becoming an edge for teams that use it correctly? Im only interested in how it can make us a better team, not how it predicts scores of games.

 

Why not?

 

That paper by Lock and Nettleton seems to be solid after a cursory read through. Just to summarize what kind of analytics it is for those who don't want to read through it, it's basically an exercise in conditional probability. That is, given x,y and z are true, what is the probability of winning the game (what is the win percentage given certain conditions)?

The "conditions" they look at is in table 1 of their paper (there are 10 of them, such as the current down, current difference in the score, time remaining in the game, etc..). Their analysis allows them to look at combinations of those conditions, and then say what the statistical probability of winning is (just based on historical data). That's really all it does. It's solid, but simple.

What analytics of this sort is useful for is in making game-time decisions, such as when it's statistically best to call a run or pass play, or when to kick vs. go for it on 4th down, etc... It's exactly the sort of thing that, if applied, could help out where Philbin is the worst at. Question is Ross that smart or is that just a theoretical possibility that won't be realized because the head coach ultimately decides what to do?

 

What makes the statistical analysis of football (and most other team sports) so difficult is that it's hard to tease apart the contribution of each player to the end result. However, unlike a lot of other team sports (soccer, hockey) where that's also true, football at least in principle shouldn't suffer from "garbage in garbage out", because you actually can measure a huge amount of interesting stuff (most of which hasn't been measured of course).

Precisely because there are fixed plays, common terminology for many formations (and plays), etc.. you have the ability to condition outcomes on all these things if someone actually took the time to compile the stats. I think the only way it gets done is either for academics to get really interested in the stuff (unlikely right now), or for someone to create a site where you take advantage of crowd-sourcing (let users find clips, share them, and have other users measure very specific things). Create an open database that way, and the analytics guys (and teams) will have a field day.

 

exactly. at its current stage stats can be used for overall team analysis like tendencies, play formations, situational game analysis but if we start getting into the moneyball type of individual analysis using stats is a losing game right now and is still surpassed by the old time grizzled vet eyeball test scouts.

the problem i think, is you have to get a football expert to team up with a math geek and then analyze every play for all 22 positions and then come up with formulas where you determine degree of success or failure of the play for each position. let's say the right guard misses a block, the DT rushes through and causes the QB to rush his throw to a WR who rather than fighting for the ball , allows it to come to him and by doing that allows the DB to jump the route and get an interception. now you have three ****ups on one play, what percent of the failure of the play do you assign to each of them? The default answer is probably 1/3 each but a vet scout might skew the percents depending on how the play unraveled and what options he saw possible for each player. the QB could get 50% or 10% offf the blame depending on how fast the rush occured

After you do this, you have to do the same for every play of every game of every season. That is a ****load of work and noone is doing it. and no PFF and football outsiders are not doing it either

 

my favorite was game spread or what las vegas line was used for the game. lol. but you're right, for situational football it would be a good crutch to have for a coach like philbin. of course it would be deadly when playing against a coach like belichik who would purposely go against his trend if he knew you were using stats

 

If it's possible to measure the individual components, and IF you have tons of data, you can let a computer run through all possible weights (so.. all three must be between 0 and 1 and all three have to sum to 1) and determine the optimal weights given the dataset (optimal at predicting some outcome, like wins). The expert is useful in determining which parameters matter. The computer is better at determining the relative weights (assuming you have enough data).

 

Did you notice their study was done early last year and they showed win probability for the Ravens SB? Their graphs would I bet look at lot less convincing for this year's SB (I mean, all the twists and turns at the end of the game, especially what happened at the 1 yard line!!).

 

Yeah, I bet this is where an expert could actually improve things. A statistician knows only to weight various variables by their (absolute) correlations. The statistician does NOT have a good intuition on how the individual stats (that they measure the correlations of) are related. In other words, they're assuming independence among the variables. The expert I bet could be used to suggest cause-and-effect mechanisms that would lead one to no longer think of each variable as independent of another. It's that "correlation vs. causation" thing where experts can really help.

 

and thats fine, I'm all about educating the coaches to understand probability relative to down and distance and situations better, just don't want it to be the driving force or final decision maker when it comes to personnell.

 

exactly. and thats the problem i have right now with the stats guys. they are math geeks not football guys. it could be done right but the man hours and the right correlations arent there yet. i mean if you include coaching decisions thats 24 people you have to evaluate every play. if it takes you one minute per play to evaluate one player, right it down and rewind for the next player that comes out to just under 3 plays per hour that you can evaluate. there are on average about 130 plays per game or so. it will take approximately 45 hours to analyze one game properly, maybe 50. times that by another 15 games and 16 weeks and you're talking 12,000 hours of evaluation or 500 days. this doesnt take into account lunch, going to the bathroom, taking breaks, etc.....that's a helluva commitment

 

Take a cue from the history of sabermetrics and how that's ultimately been applied. Teams tend to use it most for such game-time decisions and for evaluating personnel ONLY when the stats can be shown to increase win probability (the famous example is on-base percentage being much more highly valued with sabermetrics than scouts used to value it). Sabermetrics has not found much influence when they can't show there is some stat the scouts don't already value in selecting personnel (famous example is that pitching prospects are still best evaluated by scouts.. stats really haven't helped much there).

 

I can't find the paper itself on the web. The abstract is easy to understand (in that I can infer what they're doing) until they say they used graph theory etc.. That info on its own tells me nothing about the method they used to determine causal relationships. Do you have a free copy of that article so I can see? Or can you summarize their methodology for that part?

 

Yeah I know.. I think the only viable way is to crowd-source it. There are more than enough people interested in doing a tiny bit of that work.

 

and thats baseball.. If I'm a coach in football and there are metrics that show me better probability when it comes to a in game decision I'm all ears.

If the metrics wanna tell me that their player A is better than my player A then Ill scout both and make the call...pretty cool system.

 

Well, that definitely means I need to see the paper. Let's say they looked at potential causal relationships among the 4 variables for QB rating. That's 2^6 = 64 possible directed graphs (how do I get that? Each vertex in the graph has a directed edge to every other vertex in the graph.. that's 6 edges, but each edge could be directed in one of two ways, so it's 2^6). Now, other than having 64 different graphs, you need to do parameter estimation for each directed edge. At this point my intuition tells me there simply isn't enough data to really determine which graph is in some sense statistically significant in terms of being "most likely to be the correct one". That's an astronomical amount of data you need.

I'd need to see what assumptions they made to reduce the search space. The devil is in the details there.. are the assumptions justifiable or not?

The rest of the abstract regarding breaks in the historical record that correspond to rule changes is nice to see. It's something I've always assumed but never knew someone actually did a statistical analysis of.

 

Yeah, the sabermetrics guys initially wanted to select players ONLY based on stats. No team (after some tussle and experimentation) subscribes to that today (including Oakland).

It's just important for people to remember that people who have tons of experience sometimes still infer the wrong things. In basketball, the famous case is that most players and coaches think you get "hot" or "cold" when shooting. Statistically speaking it turns out that's a hard theory to defend. Human perception is biased, and when everyone seems to have similar biases it's hard to correct those biases without a more objective analysis.

 

but then you run into the problem PFF has in that are they getting good evaluations from their volunteers? i mean you're not getting Bill Cowher to evaluate, you're getting some guy who has an unknown level of knowledge about football

 

This would be different than pff. With pff, each volunteer acts as an "expert". Here, there are designated (verifiable) experts that tell us what measurements to make. It's up to the volunteers to tag each video clip with what they think is the correct measurement. It could be as simple as estimating the distance between two players (we'd probably use hash marks to estimate that), or whether a ball was "catchable", etc.. The crowd would only make those types of decisions. And in practice, such crowdsourcing tends to work relatively well. I can see there potentially being a bias when you rate your own team, but that should be drowned out as noise with enough people.

 

I'm certainly not opposed to trying something new to see if it can improve the performance of the players. Because what they have been doing for the past several years certainly hasn't given us the results we demand from our team.

Of course no matter what this new approach might offer, the game of football hasn't changed at all. It has always been about lining up and beating the man in front of you. Stats and probabilities are all great, but they won't make one bit of difference if your team can't block and tackle better than the opposing team.

The fact is the game of football really isn't as complicated as some people want to make us believe it is.

 

I don't agree. Football is really complicated. IMO, the most complicated of the major sports.

It is also a game of inches. If a sport scientist can help gave a few extra yards a game or stretch a drive or some little difference, than I think it is worth it.

 

The story behind the story to me with respect to these hires is that Dennis Hickey was hired on as General Manager making all kinds of promises about how he was going to personally spearhead the founding of an analytics department, and how he was also going to implement new technologies for use in evaluation.

And his version of all that was essentially some touch screen televisions and iPads.

Fact of the matter is he knew absolutely nothing about any of it. The Buccaneers under the management of Dominik/Hickey used to do statistics/analytics projects by simply telling their player/media relations people to put something together, presumably because they knew how to operate computers and Microsoft Excel. When Mark Dominik tried to get serious about implementing analytics, engaging in a hiring process, I know for a fact that Hickey resisted. The iPad initiative with the Buccaneers was known around town as Raheem Morris' project, not Dennis Hickey's. I've spoken to people that know Hickey and they say it's a wonder he can work a smartphone.

What Dennis Hickey is great at though is selling himself, and saying all the right things. It could very well be that when Hickey tried to sell the uppers on the idea that he was experienced with and would implement new and exciting analytics/technology-based initiatives, they simply never believed him...and knew they were going to have to supplement that effort all along. But I am not so sure, personally. I know for a fact that in those first couple of months after the Draft when Hickey was supposed to set up the analytics department he'd promised in all the newspapers, that applications and resumes were being funneled directly to him as everyone in the organization knew that the whole initiative was "his" baby.

A few months later they hired Mike Tannenbaum as a consultant to specifically take over this area of the franchise.

So somewhere in-between I think Dennis Hickey completely dropped the ball and they either discovered that he was a fraud with respect to the analytics/technology initiatives, or were reminded that they never truly believed him when he tried to sell himself that way to begin with.