Back in the day, before marriage and children came along, I used to watch a ton of football. I don’t mean all day Saturday and Sunday – I mean every day of the week, every spare moment. It was also a time when my brain worked much better and with great efficiency, so the combination produced some pretty good results when it came to fantasy writing … Sports Illustrated online, Fantasy League Football printed magazine, the iconic SportsGrumblings.com site… all benefited from my obsession.
The area of my game that really benefited was the IDP (Individual Defensive Player) work. You have to understand something: back in the late 90s, there weren’t that many IDP leagues and there were even fewer writers who covered IDP. Mind you, I was no Gary Davenport (I mean, who is?) – but I did manage to build a minor rep as an IDP “expert.”
Well, I’m older and slower now, and every season it takes me a bit deeper into the preseason to decide if I’m coming back – and even though I’m back for 2018, I simply don’t have the time, energy or testosterone required to spend 70 hours a week watching football.
Here’s where computers and algorithms come in handy. Rather than sit here and watch all the preseason games to find the one LB3 that might become IDP relevant, I sat in front of my computer, brought up some NFL stats and went to work on writing this column.
Sifting through the Mess
Look, sometimes I come up with novel ways to apply statistical theories about things like anomaly detection and try to apply them to NFL players – some work out (see The Curse of 325), some need to be proven out (see Targeted Opportunities), and some just don’t pan out (let’s not talk about the aborted attempt to correlate quarterback performance to the number of throws they make and the sacks they take).
This year, I decided to try and identify IDP bounce-back players in a more scholarly way. Thinking through some statistical control processes that originated at GM in the 1950s to help identify anomalies on the assembly line, I decided to apply a variation to NFL players.
I started by selecting all the IDPs in 2017 whose fantasy output (IDP points) declined compared to 2016, but were still above their positional average (so the players we’re looking at were fantasy relevant). Next, I looked at players whose decline was between one and two standard deviations from the average – meaning significant enough to warrant being considered as “bounce-back candidates,” but still within the “anomaly range” (as opposed to two to three standard deviations, which would indicate the beginning of a systemic decline). Finally, I took the results and eliminate those players whose age would indicate a high likelihood of declining performance in 2018.
The following was the set of players who fit all the criteria listed above:
Remember, the greater the number of standard deviations the more likely that the player’s 2017 performance was not due to chance.
And keep in mind this is one of those ideas I worked out at 2 a.m., so … caveat emptor.