Analytics: a question a child may ask, but not a childish question.
In this modern day world, data swirls all around us. Where the data is supposed to capture our spending habits, political proclivities, or whether we will like to watch ‘The Hottie and the Nottie’, the reality is we live in a world where data is everywhere and all the big money nabobs want to understand them.
The problem is no one really knows what to do with all these data.
Sure, there are buzz words like ‘data mining’ that tickle every CEO’s interest like a hooker at a yacht party. But no one really knows what the data mining is actually doing….including the stats guys.
So let’s get back to analytics….what are they?
Webster defines ‘analy’ as ‘of or similar to the anus’ and ‘tics’ as ‘involuntary spasms’, so maybe analytics are rapidly firing sphincters. I am sure some believe that.
But from context, I don’t think that is what is meant by analytics. In my humble opinion, I think we can safely and broadly, define analytics as a way of using data for some purpose.
In sports, analytics are used to evaluate and improve player performance (eg; how to run faster and stuff), assess player quality (eg; WAR in baseball), and even to help with decisions and strategy in a game. This last one is the main thing I want to talk about.
Football is being inundated with these type of decision analytics, and frankly, no one really knows where they come from or what to do with these things. Half the coaches say ‘be damned with these new fangled numbers…I’m not letting any egghead tell me what to do’ to ‘I made a laminated chart based on analytics to tell me the best decision possible in any given situation’.
Let me tell you why both are wrong.
First off, the old ball coach approach of ‘I know more than you because…I just do’ doesn’t cut it anymore. Doesn’t cut it in business, doesn’t cut it in medicine, and if you want to win, doesn’t cut it in sports. Football is a multi-Billion dollar industry and has more complexities than a Mandelbrot set, so using data in a smart way to dissect overall strategy and in-game decisions is only smart. Letting one person have unchecked power and carte blanche over an entire franchise is just plain stupid. Just ask the Cowboys and Lincoln Project members about that.
But on the other hand, blindly following ‘analytics’ and letting your little options sheet determine all high leverage decisions is pretty dumb as well. The main reason why is simply put, those head coaches do NOT understand why their options sheet say what they say. In fact, I bet only the few people in their analytics division know what determined what options to do. And I will bet you a shiny new nickel that those same analytics folks are very well aware of the limitations of their results.
Case in point:
Last Thursday the Chargers went for the touchdown two times on 4th down at the Chiefs 5 yard line instead of booting a field goal. Now, I cannot exactly say what their little sheet says, but apparently it said ‘no fg’ that day. Why?
Again, I cannot say I know for certain, but I am 95.3% sure they are toying with the ideas of expected value, or expected points for this example. I bet the logic goes something like this:
Expected points = Probability of success * points scored
Expected points(kicking a fg) = Prob(making fg) * 3 pts
The Prob(making fg) is around .98……so
Expected points (kicking fg) = .98 * 3 = 2.94
Expected points (going for the td) = Prob(play working from the 5 yard line) * (6 pts + 1 pt * .94) (the prob of making the extra point)
The Prob(playing working from 5 yard line ) = …………… no one knows. But in order for this option to be worthwhile, the Expected points (going for td) > 2.94
So….. 2.94 < Prob(play working from the 5 yard line) * 6.94 which leads to…….
Prob(play working from the 5 yard line) > 0.424
In other words, do the Chargers have a > ~43% chance of getting those 5 yards?
Now no one actually knows the answer to this problem. It is called a latent probability. But this is where the coach has to accurately access his real chances of his playcall. In this particular scenario, the Chargers ran 3 plays from around 5-8 yards with no success. Why would they think they have a 43% chance of getting those 5 yards if the first 3 plays didn’t?
To give you a better idea of things, if the Chargers thought their real probability of getting those 45 yards was…say 50%, failing 3 times in a row is 12.5%. One may say that their measure of 50% chance may be wrong.
Now I don’t want to get into further details, such as field position after a failed 4th down or whatever, that would take waaaay too long to flesh out and I simply am not getting paid enough to do that. My point is that these ‘analytic’ heavy coaches really should think more like what I just explained and not just make decisions based on a sheet that derived probabilities based on who knows what.
Side note: I bet they base these probabilities on historical data that calculates probability of the past success/fail at various down and distances. This is a worthy blue print but CANNOT be the end all in a decision. Your team has an offense that varies from the historic mean, and you are playing a defense that probably does as well. Calculate your probability based on your perceived team’s talents and the opponent. And stop with that ‘I am just confident in my guys bullshit. Be realistic, don’t hide behind that shitty coachspeak and excusing bad decisions on a sheet that was possibly compiled using data that are not germane to your situation!!
As all things, the best answer to gray-area issues is somewhere in the middle. Some analytics are very useful and should be modified correctly in the correct circumstances. But in order to do this, the coach has to first understand the derivation of these to SOME degree before they can claim to be some smart in inventive coach.
Brandon Staley……kick the damn field goal in those cases. You are not as smart as you think and your offense is not as good as you think, or else you would have scored on one of your first three downs!