Thread: Fumbling and RB position

Originally Posted by underrated29

Buff, I love you, but what a load of crap!


It Basically does not. It basically does nothing. You are not dumb enough to say 4 fumbles 100% completely wipes out 10+ tds. Basically wipes out game winning drives and points. Basically wipes out amazing 1 handed catch on 4th down- to get the first so we can go win the game.




All of you arguments are Basically Crap. Correction, all of your arguments are crap.


Basically Gordon is the best thing to happen to all of you since your mother giving you birth. Now go build him a shrine!!

My mom giving birth to me might be the worst thing that ever happened to me since that means now I have to be alive to read your posts!

Originally Posted by BroncoWave

My mom giving birth to me might be the worst thing that ever happened to me since that means now I have to be alive to read your posts!

Ok I spent too much energy looking at this issue with too little to show for it. All the same, I'll share what I think is useful.

First off, I grabbed a bunch of data from Pro Football Reference related to the 2020 season--all of it team data. Team offense, defense, passing, etc.



I am severely hampered in exploring the most meaningful 'cruxes of the issue' because I don't have great data. For example, I'd really like to focus on stats from the first 3 quarters of the game, since so much nonsense happens in the 4th quarter once the outcome is all but decided.

Oh well.

The first thing I wanted to check is the relationship between fumbles and wins.

Well. It's not so easy. The thing is, fumbles, as a team statistic, is not that indicative of RB fumbles. Most of a team's fumbles are from the QB position. Many more are from wide receivers. For example, the Broncos has 23 fumbles on the season: the 4 from MG3 that have been beat to death, 8 from Lock, 2 from KJ, 2 from Jeudy, 2 from Diontae Spencer (!!), and 3 more from others.

Many of those fumbles are not ball-security, but are the result of a sack.

So let's shift our focus instead to the idea that "offensive turnovers of any variety are very damaging to championship hopes."

I mean, they are. Duh. No one needs this frickin thesis for that. But still, here's some charts...

This chart is a scatterplot of a bunch of coordinates. The X-axis is TEAM WINS (and will be for all of these that follow). The Y-axis is % of plays that ended in a fumble or interception.

Note: many of these fumbles were recovered.

The strength of the simple linear relationship between these two variables is related to the "R squared" or R² number...in this case, a meager 0.11. That is saying 'yeah, there's a positive relationship, but it's tenuous at best'.

But how about this--what if we just consider the fumbles that were actually lost? What if we bring pure luck into the fold and see how much that determines if a team wins a game?



Well...the R² value goes up to .28, which is a significant bump.

It's kind of like saying..."Offensive fumbles and interceptions explain 11% of winning football games. But fumbles LOST and interceptions explain 28% of winning football games."

That's kind of wild when you think about it--that the arbitrary results from fumble recoveries determine so much of a team's fortune.

Some teams were lucky, and they recovered a LOT of their fumbles--more than 2 out of 3. Others were less fortunate. How many games do you think this spread accounts for?

Well, in any event, there's a lot more to winning and losing games than just offensive turnovers.

There are also...DEFENSIVE turnovers. Yes...what percent of the time did the defense end an opponent's drive with a turnover? We would expect the contribution of this aspect of a team's performance to be similarly important.

It turns out, it was slightly more important (at least the numbers work that way).



The Y-axis is the percentage of drives that ended in a defensive turnover.

Oh, btw, as an aside, the Broncos led the league in 2020 with the WORST offensive turnover rate--17% of drives ended in a turnover. Blech.

Ok, back to our regular broadcast.

What we see here is that 39% of NFL season wins are explained by how often a team's defense creates a turnover. (the R² value)

If we combine offensive and defensive turnovers, the relationship to overall winning becomes even stronger.



The .55 R² is a decent indication that there is a strong relationship.

Again, this is a correlation, not a directional causation. Winning games also generates turnovers if you think about it.

The Y-axis in this chart is the percentage of defensive drives that end in a turnover minus the percentage of offensive drives that end in a turnover. It is probably about the same chart you'd see if we just used 'turnover differential' rather than as a percentage of drives ending that way (if not exactly the same).

Ok, but there's more than just turnovers. There are also...PENALTIES...

Ok, there's a million other things, and trying to gnat's ass how it all works is complicated af.

So let's stick with big picture stuff.

How about offensive execution? Super important. How about we look at NY/A (net yards per passing attempt--basically yards gained per pass play including the negative effect of sacks) vs wins?



31% of wins explained by passing efficiency...that squares with common sense.

I should note, these R² values will not add up to 100% because there are inter-relationships between these variables--they are not strictly independent of each other.

You can run down a bunch of these variables to see what matters, but there's always some little detail that makes it hard to put in a clean, tidy package.

I'll hit the highlights:

I created a model that uses only these variables to explain/predict 2020 W/L totals and also point differential. What was the most important? In order, it was:

Defensive TO creation
NY/A
Luck (fumbles that bounce in a team's favor)
Y/A (rushing yards per attempt)
Offensive TO
Penalties
Strength of Schedule

That's what I used. I could have used other stuff, but it's a pretty arbitrary division no matter what you choose--and there will be a lot of overlap if you aren't careful. Some stuff (special teams, QB pressures, coach mismanagement, etc.) was completely omitted. For example, at one point I used field position and points per drive--but they made some of the parameters completely useless because they were SOOO outcome-determinative. That makes sense--points per drive is almost the combination of all of these factors--Def TO, Off TO, NY/A, Y/A, Field Position. Yeah, points per drive was too OP.

I could run this on 2019 and see how stable the results are. I'd expect some fluctuation. But so what? Is there a good takeaway?

I think so. Turnovers and luck are hugely outcome-determinative. And while winning games (through execution and strong NY/A and Y/A) kind of creates extra turnovers once the game goes into the 4th quarter, it's still such a strong relationship it deserves crucial attention.

There's a lot to the game, and all the details add up.

For my next dig, I'd like to look if any teams have been built based on the offensive turnover-minimization idea, and how they have fared. But that's enough for now.

Damn you, UR.

Wow you are diving deep into this my man. Impressive

I am way lazier than Hawg - and would love to be able to quantify this even further, but I think it really comes down to this:

Over the past five regular seasons, teams that have won the turnover battle have won their games 76.8 percent of the time and covered the spread 76 percent.

https://www.sportsline.com/nfl/news/...rcent-of-time/

We know the turnover margin matters. We know there are a lot of variables that go into turnovers -- but for a RB it's pretty simple - don't ******* fumble. Ever.

Originally Posted by Buff

I am way lazier than Hawg - and would love to be able to quantify this even further, but I think it really comes down to this:



https://www.sportsline.com/nfl/news/...rcent-of-time/

We know the turnover margin matters. We know there are a lot of variables that go into turnovers -- but for a RB it's pretty simple - don't ******* fumble. Ever.

Yes, and without looking it up- how many RBs have not fumbled ever- RBs with at least 50 carries. You already know 1, Phil. Name 4 more. How many of those (if you can name off without looking up) are in the top 15 of RBs either of the last 2 years.

And the entire premise is the same for all offensive skill positions. WR, TE, RB- and IMO moreso QB. QBs are the ones that should not turn the ball over ever!

Hawg, Great work sir!

It gave you something to do. It gave the rest of us a pointer to watch you get all nerdy!

Originally Posted by Hawgdriver

Ok, but there's more than just turnovers. There are also...PENALTIES...

Ok, there's a million other things, and trying to gnat's ass how it all works is complicated af.

So let's stick with big picture stuff.

How about offensive execution? Super important. How about we look at NY/A (net yards per passing attempt--basically yards gained per pass play including the negative effect of sacks) vs wins?



31% of wins explained by passing efficiency...that squares with common sense.

I should note, these R² values will not add up to 100% because there are inter-relationships between these variables--they are not strictly independent of each other.

Well done my Bronco A-type brother!

Originally Posted by Softskull

Well done my Bronco A-type brother!

Thx.

I feel like I could hit it out of the park with better data. Maybe start a consulting side hustle. Maybe I will...

I love this ish.