*Predicting Tackles, Sacks, and TFLs

[Gamaliel]

Hey all you cool cats and kittens, Joe Exotic here to take you on a wild ride in the world of (basic) statistical analysis of some football stats and the NSFL. First off, who am I? I am Gamaliel, my player is Joe Exotic – DT. I recently created and am eligible? for drafting in the DSFL. I found out about the NSFL from the PBE and am fairly active in that league. Having just joined the NSFL, I know literally nothing about the teams, players, or the league in general. So what is the best way to learn about a league like this? Well the best way to take a deep dive into the stats from the games. I want to emphasis again that I haven’t read any past articles about this and might be rehashing some already tired arguments, or I might be completely wrong because again, I know nothing about this league.

I started this analysis with the question. What in this league predicts the success for a defensive lineman? My working “definition” for success in this will be counting stats, those are nice and quantifiable and easily tracked from the Index. And what do defensive lineman do? They tackle people. So I’m going to start this with the question. What attribute predicts the amount of tackles for the NSFL Defensive Lineman?
For starters, I pulled in information from the Index for this year’s individual defensive stats. Sidenote – If anybody has a convenient way of pulling historical data over a number of seasons I would love to get my hands on that because currently it’s a bit of small sample size zone. Then I joined this data to TPE tracker data which shows TPE totals and where the TPE has been applied to, very handy for looking at the impacts of individual attributes. I then cleaned this data up, removing any free agents and any players who haven’t been seen in 2020. I ended up with 32 players that accumulated stats in this most recent season of the NSFL. Let me know if I missed anyone in this. I’m pretty happy with how clean the data was. Also, I am using a program called Alteryx for all of this data cleansing and analysis. You see this data in the picture below:


Alright, time to run this data through a linear regression. First off, how does current TPE impact tackles?

We can see a fairly decent relationship (for “real” data) with about 22% of tackles explained by TPE totals. Let’s look at some attributes that can also effect tackles, namely, Tackle, Speed, and Strength
You would think that Tackle would predict tackles right?

Not quite, an even weaker relationship than just total TPE. What about Speed?

Oof, no need to put speed in based on this linear regression. Time for Strength.

Again, not a great relationship. Time to try a multiple regression with all 4 of these variables.

Hey! This regression is decent, but not great. This thought led me to the thought that maybe it’s not really a great idea to be looking just at tackles for the defensive linemen because most tackles don’t happen from a linemen. They are more likely to have a tackle for a loss or a sack, and most team’s percentage of tackles are going to be coming from either their secondary or linebackers who are covering the receivers and running backs. So to help my investigation out, I took the team total defensive stats and compared them to the DE and DT summed totals that I was already looking and found which percentages of Tackles, Tackles for losses, and Sacks came from the defensive linemen for each team and took the average of that. Check out the results below:

So based on these results, on average only 12% of a team’s tackles are coming from their D Lineman, but almost 40% of sacks and tackles for losses came from the DLine! It seems that those two stats would be better at taking a look at. For brevity’s sake, I am just going to highlight a few more of these that illustrate some of the better R-squared values. First off, Sacks:

There wasn’t anything that was a great predictor of sacks individually, but all together they made a decent estimate. I wonder if this was because of the role of team strategies, Quarterback elusiveness, and just plain luck. Any thoughts from anyone on that? Next up, Tackle for a loss.

It turned out that the best predictor for tackles for a loss was speed. Which logically makes sense. Without speed how can you get behind the offensive linemen and tackle behind the line of scrimmage? Again, I think there is a lot of small sample size baked into these results, but it is at least interesting to see which factors can play into these stats. Alright all, roast me on why my method is flawed!

[iStegosauruz]

I love this.

So I haven't run regressions on any of my work like you did here, but I have looked at how the various attributes effect stat production for each position. I've published those numbers for CBs and I should publish the number for DL soon.

in general, however, what you find in the sim is that speed is far and away the biggest predictor of almost anything. The sim puts a heavy influence in your ability to get to the play. Once you get to the play its a combination of any number of secondary attributes - some of which are logical depending on the stat you're looking at and some of which aren't as logical. We typically see strength being the second most impactful for a lot of positions, and my theory is that its very similar to speed - its somewhere between a primary and secondary attribute. Speed gets you to the play and then strength has bearing in almost every outcome of the play at that point.

Then you see things like tackling, handles, and intelligence all factor in depending on the type of play/stat you're looking at it. Inherently the tackling value does nothing in the sim if you can't get to the guy or have the strength to bring him down. Its just how the system determines things.

[Hallmonitor_20]

Wow, even in media I don't make the DE list

[Gamaliel]

I love this.

So I haven't run regressions on any of my work like you did here, but I have looked at how the various attributes effect stat production for each position. I've published those numbers for CBs and I should publish the number for DL soon.

in general, however, what you find in the sim is that speed is far and away the biggest predictor of almost anything. The sim puts a heavy influence in your ability to get to the play. Once you get to the play its a combination of any number of secondary attributes - some of which are logical depending on the stat you're looking at and some of which aren't as logical. We typically see strength being the second most impactful for a lot of positions, and my theory is that its very similar to speed - its somewhere between a primary and secondary attribute. Speed gets you to the play and then strength has bearing in almost every outcome of the play at that point.

Then you see things like tackling, handles, and intelligence all factor in depending on the type of play/stat you're looking at it. Inherently the tackling value does nothing in the sim if you can't get to the guy or have the strength to bring him down. Its just how the system determines things.

That makes sense, there is probably a lot of comparisons between player A speed against player B speed and the higher speed wins and the loser won't be around by the time the play finishes developing. Great to know!

Wow, even in media I don't make the DE list

Ya, I saw your player on the TPE tracker as a DE, but couldn't find your stats on the index for some reason?
Found it, your name is listed backwards on the index, compared to the TPE tracker.

[timeconsumer]

So I used to do this regression stuff as well, except one of the things I did was just have the sim generate its own teams and players and I'd sim 20 seasons worth of data and extrapolate off of that. It got me some useful info here and there, of course speed mattered a lot as we all know, but I caution you on something about reading too much into what attribute affects this stat or this stat.

The position you play on defense matters more than almost anything else. A RDT will never get as many TFLs as an RDE in the nickel 35 everyone runs. A SLB will almost always lead the league in sacks in Nickel 35. A Nickel MLB will almost always lead in tackles.

I don't fully understand why this happens, only that it does. Speed and strength still matter because you need to beat your friends to the ball to get the stat. Strength helps you get off that block and/or take him down on the first try. Speed just helps you get there. Other stats matter too but yeah, position is really important.