Wow, if that's true that's really useful info. So, any inside info on how Tannenbaum and Hickey are supposed to mesh?
Knowing all too well our track record and organizational aptitude when it comes to hiring the right people, I immediately assumed we hired the mad scientist from the opening credits of Robot Chicken.
No idea. There's no information on it. The Dolphins say it's a great marriage. They said that about Philbin and Ireland too. But I think we can all use our own heads and our own perspectives to figure it out. You're Dennis Hickey and you just got hired to be the General Manager. You had what looks like a pretty good draft. A year later you get knee-capped by having a former GM placed over top of you ostensibly to take over everything except running the scouting department. Except he's going to be watching film of the players you're evaluating, and he's going to be in the draft room with you, and he's giving you the impression all major decisions in free agency and probably the draft run by him first before they're executed. Stuff that you told everyone including the media that you were going to personally spearhead, Tannenbaum is doing them all. You answer to him now, whatever they have to say publicly. Are you happy?
Love it. I hope they can find a coaching/personnel staff that they can work well with and properly implement their ideas. Given his experience in the league I'm hoping Tannenbaum is well suited to be able to communicate this stuff to more traditional-minded football people.
I think the context you're using is giving too much credit to Dennis Hickey. I think you could just as likely say that the guy should be a head scout or director of scouting somewhere, but he's been given the benefit of an inflated title in Miami.
Perhaps but we've got no real evidence the Dolphins knew this going into it, and I do have a pretty solid chunk of evidence that the Dolphins really did put Hickey in charge of setting up an analytics department...until they took it out of his hands.
Why does being at an early stage of development preclude further investment? If anything, I think the fact that analytics in the NFL are so bad is a big reason why it warrants more investment.
Isolating the individual contribution of players really is not that difficult in theory. There is nothing inherently preventing analysis to the same degree as you see in baseball or basketball. The only thing inhibiting more informative statistical analysis is the NFL itself. This is a league that is notoriously opaque and resistant to sharing information with the consumer.
You don't need to go through all this. "Stats guys" accurately pick winning and losing corporations without any real insight into the business. Football is not nearly as complicated as businesses.
Well.. unless you do controlled experiments, isolating the individual contribution of players is even difficult in theory. It really goes back to that correlation doesn't equal causation argument. That structural equation modeling Tannephins referenced earlier is an attempt at doing so just from the aggregate data, but there are just so many possible sets of causal relations that you need tons of data to really show one set of hypothesized causal relations is very likely the right one. And while I agree that the NFL could help out more in sharing information, a lot of the stuff that would help is just not measured by the NFL either. For example, who has stats on the estimated distances between any two players at any time during a play? The NFL doesn't have that but it's something you could crowd-source effectively (and would be very useful data for better statistical analysis).
All the data is there. The controlled experiments happen all the time. Just off the top of my head using exogenous exits would be an easy way to quantify the contributions of each position. Once you can establish the value of each position, evaluating each player becomes very easy. You don't need to crowd source it. The technology is already there. Its widely used in basketball already. But you don't even really need that. The existing data is enough to make big breakthroughs.
Not getting what you want or having to take a demotion does not automatically equal strife in the workplace either.
Hell, even assuming that he wasn't aware that the Dolphins were bringing in a VP, I'm not sure Dennis Hickey is really upset. I'd venture to guess that if posters on this forum were offered exactly what has happened to Hickey, 99% would jump at it. Its not like Dennis Hickey had a realistic shot at becoming a GM elsewhere.
Almost none of the controlled experiments have been done. For example, what is the probability an average starting NFL QB throws a completion when the WR and DB are exactly 2 feet apart? etc.. I do agree that for just estimating distances you could use technology, so maybe that's not the best example for crowdsourcing, but you can't yet automate measuring the degree to which a receiver is thrown open. Point is, there's a lot of info out there that the NFL (likely) doesn't have itself but could be crowdsourced.
So let me get this straight. You get hired on to be a team's General Manager, and you have no idea the Dolphins intend to kneecap you one year down the road. You execute a good Draft and an up-and-down free agency where you come off convinced you suffered a little bad injury luck (Albert & Moreno, particularly). Then a year later they hire a former General Manager in the prime years of his career to take a position above you, overseeing you, and in charge of everyone except the coaching staff. Supposedly you have autonomy on the roster but he's watching film on all of your players, he's holding the purse strings, and making it clear no major decision happens without his say-so. He's also primarily the person in charge of doing what you spent months telling the press was your baby, your project that you were going to execute. And you're not upset by that development? Come on. I think you're thinking several steps ahead, on the basis of "well at least I'm not fired". But yesterday you were in charge of the Dolphins and today you're not. Yeah, I think most people would be upset.
So what though? Fine, he's upset. That doesn't mean he won't fall in line and do his job to the best of his abilities. It doesn't mean he won't get over it. It doesn't mean he hasn't already gotten over it.
Here's an example: http://www.footballperspective.com/which-receivers-elevate-their-quarterbacks-the-most/
Yeah, that is a simple measure of how much a particular player (at some position) contributes to some overall stat, but it doesn't answer questions like what % of the credit for a particular play should go to players x,y and z. This is the problem that adamprez2003 and I were talking about. A simple example: on a given play, let's say the defensive line decreased the time available to the QB to throw (relative to average time available) by a certain amount, and the DB covering the WR kept the separation between the two less than the average separation by a certain amount. The play was either successful (catch) or unsuccessful. In either case, how much of the credit/blame goes to say the defensive line vs the DB (or QB or WR)? And of course, I'm leaving out all kinds of other variables because I just want to make the case with a toy example. If you really wanted to answer this question with some rigor, you'd first measure the completion % of an average QB to an average WR covered by an average DB as a function of time to throw, and then you'd measure the completion % of an average QB to an average WR covered by an average DB with X amount of time to throw as a function of distance between the WR and DB. With that information, you could determine the relative contributions of the defensive line and DB. The defensive line decreased the probability of completion by reducing the time to throw, and the DB reduced the probability of completion GIVEN what the defensive line did. Those probabilities are the relative weights, etc.. There are so many other variables of course. All this is conditioned on the routes that are run, the type of DB coverage, the available throwing lanes (not just time available), etc.. By the time you get what you think are a small set of relevant parameters to measure (determined by an "expert"), you are talking about an astronomical amount of data you'd need if you were just using aggregate data from game film. This is where the value of controlled experiments comes in: you selectively test certain portions of the space of possibilities and infer the rest. Of course, when you are talking about only a few parameters, you could probably crowdsource it. Again, it's unlikely here that the NFL would have all the data.
Although adamprez2003 seems to believe one can "see" all these things below, via the routine watching of games on TV, and therefore any attempt at getting at it all statistically pales in comparison in its validity because "statistics are just numbers."
Two ways I can see analytics being immediately applied. One way is to simply rank FA and draft prospect based upon every relevant measurement you can make. Receptions, yards, yards after catch, yards after contact, drops, fumbles, etc. Then you weight them against the competition they faced; Div I Power 5 opponents would receive full credit, Div I non-P5 would receive 3/4 credit, Div II opponents would receive half credit, and Div III opponents would receive quarter credit. In a given position class of 100 (can be any size, with the top points adjusted for the number in the class), top performers in a single category would receive 100 points and bottom performers would receive 1 point. Then you take all the metrics from the Combine and rank those as well. The beauty of this is that, with the help of scouts, you can filter results for what precisely you want in a player. Need a slot receiver? Then you're going to value YAC, very low drops, triangle drill time, and 10 yard splits. I'm using WR as an example, but it's applicable to any position. There are also certain predictive measurements you can use. Note that none of this is a replacement for scout intuition and investigation, but it is an important supplement/check. Another is to determine plays using down-and-distance, yard line, formation, and personnel grouping. It's more straight forward and can be used in real time if set up right.
You can also use statistics (i.e., analytics) to determine which of those sorts of measurements are important and predictive of success in the NFL. See here, for example: http://www.degruyter.com/view/j/jqas.2014.10.issue-4/jqas-2013-0134/jqas-2013-0134.xml?format=INT
Interesting paper. The abstract mentions the predictive tools I was talking about. I had no idea, however, that there was a model for human behavior (boy was it spot on with Hernandez).
I think that's a crappy article (not attacking you Tannephins, just the authors of the paper). Why? They claim this: "With both modeling approaches, we find that the measures that are most predictive of NFL draft order are not necessarily the most predictive measures of NFL career success." OK, first off, their statistical methodologies are sound. So why is it crappy? Because the measurements they're using to try to predict success (for TE) in the NFL are things like: height, BMI, 40 yard dash, bench press, etc.. I get it that scouts look at those things, but NO scout (in their right mind) would think JUST those physical measurements (without e.g. looking at tape) predict NFL success. So what's the point of the article?? This is the problem with science sometimes.. so many do fancy analyses but the results are often just useless. Anyway, all this is a separate point from whether you can do what DearbornDolfan is suggesting. Yes, you can and should. And it's separate from Tannephins pointing out there are statistical methodologies (obvious ones) you can use to do that. Yes there are, and that paper shows an example of that. But this paper is just s t u p i d!
Actually some of the most predictive variables in their study had to do with college play (receptions, yards) and college pedigree (BCS teams -- i.e., level of competition), stuff one should see on film, as well. Also, these sorts of studies can certainly be used in concert with film study. There's no reason why just one or the other should have to be used.
Yeah they use some other variables you're right, but what they did isn't a post on a message board, it's an article in a scientific journal. I mean seriously you'd expect them to go the extra length to do something that hits you in the face. People spend years sometimes putting together a good research article. This just smacks of something they did over the summer.
A relatively new journal, though, in a new field. Some of the reference lists in that journal have only a handful of studies in them. This is all just now getting going. The journal's rejection rate is probably comparatively low, in the effort to get things moving.
