For those in the industry, be it analysts employed by third parties (PFF, Football Outsiders, etc) or coaches trying to gain an edge, the offseason is really the “regular season”. That is, when we have the time, data and models ready to answer questions that handicappers, coaching staffs and talking heads have been (or should be) asking last fall. With that, I broke down the 5 most pressing questions/topics that FPS will tackle this spring/summer, so let’s jump right in
- Does Pre-snap Motion offer AS MUCH “free value” as we assume?
Using motion (PSM) before an offensive play, has been part of the “low-hanging” fruit the analytics community has been screaming about from day one. Heck, the flagship podcast for respected analytics company, Sharp Football is actually called “Pre-Snap Motion”. It’s “one of those things” I would classify as “stuff kids do in Madden, and adults finally figured it out”. Meaning kids for years have used this tactic to help determine coverage and rush # for years, and its hard to debate its value.
However, not debating the aspect that using this tactic is “free points”, but I wonder if the data that objectively “proved” this, is not fraut with other variables propping it up. One such that jumps to mind is “Time left on Play clock”. This occurred to me during my time as a coach, on a staff with a OC, who was fantastic, but calling plays for the first time (ever, was a DC before this). And me, being the poor coach that had to signal in the play calls to my QB, I saw we almost never had the luxury of using motion, as my QB would get the call with typically 15 seconds left on the play clock. Knowing this, AND knowing that this is NOT a datapoint captured in any of the typical NFL/NCAA databanks (play clock time leaving the huddle and/or play clock time at snap) I wonder if the models that “proved” to us, all other things equal, (that an offensive play with motion has a better expected outcome than one without PSM) may have a different answer if they did/could HOLD FOR Time Left on Play Clock, post huddle.
Beyond the obvious ability to actually execute your PSM, you’d have to think that a QB having 20 seconds, compared to 10 seconds to survey the defense, make checks and audible has a ton of expected value to the play outcome. With that, I have a feeling that although I’m not saying “PSM = bad”, but that another variable, Time Left on Play Clock, is a variable not being tracked among those hardcore “more PSM” folk out there.
- Is it possible that Play-Action sets up the run game (more than the already debunked inverse)?
Admittedly, this is a topic we have broached before in “Does the lack of Play-Action PASSING, lead to a decrease in RUN success”. And the jist was, we know that play action passing is one of the most efficient tactics. We also have seen those, like Football Outsiders present solid evidence that you “don’t need to run to setup the pass”, and therefore play-action passing has “standalone value”.
What we haven’t looked into, is the inverse. That is, could there be a relationship in the other direction: improve your run game by running more (and or better) play-action concepts (ideally tied to the run game strategy). Again, this is an obvious one for me, but with all the objective truths we have learned about the subject, I think this could be a very valuable aspect to tackle.
- Who will be the first coaching staff to bring “analytics” to their practice script/plan/etc?
While I was coaching, I was constantly trying to find a way to gain an edge. Duh, right? But the way I went about it, was a little different than traditional staffs attempted to do this. Basically, in learning every aspect of our program, I was trying to
- Ask questions we never asked
- Provide answers with objective data (vs. “its what we have always done”)
I wanted to understand if there is a relationship between practice reps of play XYZ*, and their respective performance on the field. Kind of a “holy grail” for coaches and staffs that have so little time, and know every rep is so vital to the success on Saturday. You could take it a bit further (to the micro level): is each Individual Drill being run based on a skill that we know (empirically) will improve with said drill? That is, if the RB Coach wants to decrease fumbles, and runs em through the blaster for 3 minutes, do we know we have empirically improved our RB’s ability to hold on to the ball (at least 3 minutes worth of value, or was there another drill that could have used better use of the time)?
*If you’re curious, here were the initial findings, with various relevant pivots along the way.
One of the best examples of this “Thrid down conversions don’t matter”. What I mean by this, and originally brought to the forefront by Sharp Football, is, yes, converting third downs matter, BUT, its really a function of what you do on first and second down. And this is painfully clear when you compare 3rd conversion rate (at any level) and see you can draw (an almost) straight-line correlation between conversion and “yards to go” on 3rd down.
So, if this is the case, why do coaching staffs invest so much time in practice on converting 3rd downs? Here was my original plea if curious.
All of this is to say, with all the “analytics out there”, why has it not crept into what coaches care most about: what is the optimal way for MY TEAM to practice?
- Will there be a team that deploys a 0 PR strategy?
I’ve always had a soft spot in my heart for Special Teams, as it fits so neatly into the “what are we not paying attention to, but can easily gain up 1-2 percentage points in win probability” bucket . With that, and I do not have the objective data to back this up (yet), it feels like the amont of muffed punts has massively spiked in recent years. Combine that with limiting ST returns overall, and there’s a part of me that wonders given how impactful a muffed punt is, and how rare its becoming to see an explosive punt return, will teams (given the right situation), simply stop sending a PR back, and instead use an extra punt blocker (heck, or even swatter).
I heard about this originally when getting a demo of the edjsports software years back, and they made an excellent pitch in how valuable this strategy is for the high school level, where fielding a punt is much more difficult, and doubly so, the ensuing additional “chaos” of having an extra punt blocker could reap on the operation of the long snap/punt is almost a no-brainer for about 80% of the high schools out there.
- What’s next
It feels, rightfully or not, that we have (mostly) moved past the stage of “acceptance” that analytics is going to affect the game. And most reading this can spit out a few sound bytes they’ve heard the talking heads referring to:
- Going for it most on 4th down
- Going for 2 after a TD when down 8
But what’s next? I think there are lots of places for analytics to go (including some of the items above), but I also think there’s ALOT of opportunity in the following:
That is, the VERY OLD mathematical model set that takes into account your opponent’s options/liklihood of deployment vs. the traditional one-dimensional predictive look at play outcome. Lots of words there, but here is what I mean. Most coaching staffs have some sort of grid/matrix when scouting an opponent, that will show something (much more sophisticated than this) like “3rd down and short, opponent blitzes 40% of the time”, etc. It’s a bit one sided, meaning, what if a staff instead looks at something like this:
|3rd and Short, FZ abc, score diff. XYZ, etc||Def Blitz (40%)||Def NO Blitz (60%)|
|We run the ball||.5||1.5|
|We pass the ball||3.5||1|
*Number in box = Average yards gained for offense given the circumstance (but could/should be win probability with more sophisticated models)
In this overly simplified view, combining the expected outcomes:
- Run: (.4 x .5 + .6 x 1.5) = 1.1
- Pass (.4 x 3.5 + .6 x 1) = 2
We see, mathmetically its in our interest to pass the ball, given this game situation, the opponent tendencies, AND our own historical performance (likely vs. similar teams/schems). This is a much stronger predictor as its taking defensive scheme likelihood AND historical offensive performance together, and spitting out the optimal choice.
- “Deriavative” Prediction models/grids
Almost as a distant cousin of the above thought, most “scouting reports”/grids are static and “singular level”. Meaning they are looking at the game situation within a vacuum. Think of it this way, if you get handed a scouting report that shows a defense deploys Man coverage 50% of the time in THIS game situation, that’s helpful sure…BUT what if you just had an explosive pass play right before this. Just a guess, but the DC may be less inclined to deploy MAN after they gave up an explosive pass. Hence, it would be much more beneficial to that OC to know other SIGNIFICANT contexts, before presenting him with a sheet that shows what a defense is likely to do.
Note, other situations that come to mind:
- What a defense does after a 1st and 10 incompletion
- What a defense does after an offensive loss of yardage
- What a defense does after an exposive run, in a heavy box
There are so many of these, and AI will eventually help us get to 5-10 of these situations that are most valuable, but with simple math and intuition about the game, some staff will come up with a great advantage in advanced scouting.
- Open space/out of structure coaching
So much of coaching, since the dawn of time ENDS in the execution of a play based on lines, circles and squares being nicely blocked, etc. BUT, be it a scrambling QB or a RB in the open field, in a game that’s about creating space, a large portion of the game occurs out of structure. And whether its drills and practice time devoted to this, or zooming out and consider strategy overall, there’s very little time/resources put into this aspect of the game. Yes, we have had the “scramble drill” forever with QBs and WRs, but what about this: coaching a ball carrier in the open field, using GPS chips to see the field in front of you, PAST THE LINE OF SCRIMMAGE.
There’s alot more questions than these, and I even had 3-5 bonuses setup here, but I think these are the top 5 Analytics Questions for the 2023 Offseason, and at least for FPS, ones we are going to try and get answers to.