Dowdle is a High Variance Rusher

Compared to Warren, Dowdle is a high variance rusher. Almost 50% more variation in rush attempts and rush yards than Warren;




(Coeff. of variation = measure of the spread of data expressed as a %)

 

Are you looking at his variation carry to carry?

Or the variation in yards game to game? <-- I think this is what you mean.

Variance by itself isn't bad in a RB. Since the game breaking runs that you want from a RB are uncommon, high variance in itself might be a good thing. As long as the success rate is acceptable / high. We could imagine a back who always gets at least 50 yards a game, but every second game gets 100 yards. He'd have high variance, but would be a pretty good back (~1250 yards).

The thing I worry about Dowdle, is that his success rate last year wasn't good 47.9%. But it was good in 24 @ 53.6%.

This is something that Warren has been good at...basically always at 50% or more on the season...technically only 49.7% in 23. He was 54.5% last year and 50.0% the year before.

Dowdle's success rate last year is about where Najee was in years 1-3 (47.6%, 46%, 48.2%). Before he cratered in year 4 to 43.7%.

Aside: I think variation per carry is probably more important. Because per carry is a better "rate" than per game (since number of carries per game will change). I wish there was a site we could go to that published the distribution of carries that all RBs have. I think this would really help with the confusion that some people have re: what YPC is and why succ% matters. I also think it would be really interesting to overlay distributions of two guys (e.g. Warren and Dowdle) to really see how long and fat the positive tail is (what YPC tells us) and how many low yardage carries they get (what succ% kind of tells us...although succ% is better that this because it's relative to the objective of the play).

 

Lots of variable variations in a rush

 

Invariably

 

Maybe people actually understand and still think you overrate a statistic that isn't as much in the back's control as yards per carry. Oh wait. That's exactly what it is.

 

I am not sure I understand the OP so this may not be applicable

You have two backs one named Billie and one named Zerome

Billie averages 4.3 y/a and gets a big part of that from big plays combined with many from small gains

Zerome averages 3.9 y/a and gets most of that from runs very close to his average without many long gains or losses

Which one do you hand the ball to on the first play of the game

Which one do you give the ball to on the 3 yard line down by 4 with the clock running out

 

Measuring/comparing one standard deviation is statistically irrelevant over a few games or a single season because of offensive and defensive game plans weather personnel etc

Over a 10 year career it is more relevant

 

for tea too two for tee fore

forelzzzzzzz

 

How much time is left? And do we need a first down to run out the clock (or at least leave them with little time)?

And how good is our QB at tackling? I'm very, very glad that's not how the fictitious Zerome ended his brilliant career.

Billie did have a lower success rate (44.9%) than Zerome (50.0%). So I think your "combined with smaller runs" comment is probably true.

Again, having actual images of the distributions would help us see what the expected result might look like here.

 

I still haven’t seen anyone actually explain the OP accurately

 

Here's what I think (but I didn't make the chart).

High level:
I think this chart mostly tells us that Dowdle had a much smaller role (read: far fewer carries per game) in 2023 than in 2024 and 2025. And that Warren's role (again...carries per game) in our offense was more consistent across these three years. I don't think it's all that meaningful. Two big issues I think I see: (1) sample size across these three years doesn't make sense because the roles are quite different each year (especially for Dowdle), this is made worse by (2) looking at per game averages and not per carry averages (which would be better but still not great by themselves).

More detail:
I think the variance number doesn't mean a whole lot because it looks at yards per game and not yards per carry. And even yards per carry doesn't tell us all that much without knowing the objective of the carry (i.e., 1.5 yards is good on 4th and 1 but bad on 1st and 10). The idea of breaking it down into per game results is interesting because it tell us something about consistency per game. But I think succ% is going to tell us a lot more about consistency (of staying on schedule) than per game stats.

Another big issue I think I have with this (looking at it briefly) is that I think looking across the seasons from 2023 - 2025 artificially skews the variance for Dowdle high because his role was so small in 2023 (per carry stats smooth this out more...but it means that YPC could have really been juiced by one big carry...which he didn't really have in 2023 so YPC here is probably OK).

Again, guess is that the variance in per game stats for Dowdle is mostly driven by the fact that in 2023 he had 89 carries spread over 16 games. Then he had 235 and 236 carries spread over 16 and 17 games in 2024 and 25, respectively. His per game metrics would almost certainly be much tighter if the sample was 2024 and 2025. Seems like that sample for him would make way more sense since his role (and production) appears to be similar in those two years.

I think there's probably a similar issue with Warren's role changing from 23-25, but he doesn't have the huge "light" year that Dowdle has (carries: 149, 120, 211).

I didn't check the calculations though...so it's possible that they only mean the 2024 and 2025 seasons? I don't love the per game stats either way, but using these two seasons (especially for Dowdle) would make more sense IMO because Dowdle got such limited usage in 2023.

I think it would be much more useful to look at their season (a) Y/A and (b) succ% to get a sense of their (a) ability to generate big runs and (b) ability to stay on schedule and how consistent that is. This is what you did with your completely made up RBs in the post above...just you described succ% in words (the orange parts..but the brackets are my interpretation of how the succ% would look...and how it does look for backs with very similar first names in that year)

IMO, this is the best way to compare RBs if all we have are these "boiled down" single number stats. You can kind of back of the envelope check this by looking at the annual long to see if it fits the conclusion from (a) and (b). But that's pretty crude without knowing the frequency the back gets runs around that long.

Maybe comparing (a) and (b) on a game-by-game basis would be useful, but at some point the samples will get too small to mean anything (because the noise will increase as the sample gets smaller). And it would be a pain in the ass without writing out some kind of script to scrape the data from a place like PFR IMO.

Here's what that comparison looks like...can we make tables on this site? Excuse the format here. This should be transposed, would have been easier to read...

View attachment 18885


What I think the chart (in the OP) says...again I think mostly tells us Dowdle's usage was low in 2023, much different then 24/25. So his role (carries / game) changed more from 2023 - 2025 than Warren's did.

4 columns in the graph:

• Number of rush attempts per game. This seems to be averaged out from 2023 - 2025 (if I'm reading it right...I didn't check the math). I don't think this is a particularly useful sample since Warren's role has changed a lot in that time frame. So looking at his per game stats probably isn't all that helpful. I don't know a ton about Dowdle, but it looks like he had a huge different in workload in 23 vs 24/25
• Coefficient of Variance. This is a stats thing. You take the standard deviation of the sample it's probably and divide by the mean of the sample. Again, I think the variance here is driven by Dowdle having much fewer carries in 23
• Yards / game. I think this stat is more about usage than performance. Because I think the thing that's going to drive yards / game is carries / game. They have similar yards / game numbers. Warren has 80 fewer carries in this sample. I don't have it in my table, but number of games is similar Warren (17 + 15 + 16 = 33) Dowdle (16+16+17 = 34). I wonder if this is why this sample was picked? But if it was just 2024 and 2025 (which I think would make more sense for this graph...but I still wouldn't love the /game stats), Dowdle only played 2 more games.
• Variance on yards per game: Again, I think this tells us that Dowdle's variance is higher because he got 361 yards in 16 games in 2023. I didn't look to see how those yards were distributed. But his average yards / game in 2023 is much less than in 2024 and 2025 (where he played basically the same number of games, but had 700 more yards in both seasons).
• Aside: For basically all backs (even Najee who I think was a good all around back who didn't have a long enough tail here to deserve a big 2nd contract), the distribution should look like the "positively skewed distribution" in this figure. Guys with a longer tail to the right will have higher YPCs...but don't tell Scribe . Not too many guys will have a long tail to the left (since the mode for every half-decent back with a reasonably number of carries is going to be somewhere between 3-4, there's only so low you can go). So just having high variance doesn't tell you if the back is good or bad. If Merrick Lenry had a 50+ yard run every 4th game he'd had a high variance. But he might also have a high succ% (consistently keeping ahead of the chains). Anyone with big (and semi-frequent) longs will have high variance. Because the "mode" for all backs will be pretty similar.
• With the two single number stats we have: YPC is going to tell us something about how long / thick that tail on the right is and succ% is going to tell us something about how high the "mode" is (somewhere close to...but below the player's YPC).
• But if the mode (the highest frequency point in the distribution) is still high, that's good. In your post above, this would be a guy who consistently has a high average AND breaks big runs as consistently as it's possible to do in the NFL (Merrick).

 

When Rico is in the zone he's fun to watch. The dude will absolutely pour it on a defense when he's got them on their heels. He was a shrewd addition.

 

Looking forward to seeing this.

It was fun when Najee would run guys over.


We did still get this though...wonder what he'll look like when he's "only" 285-290?

 

I'm just excited that we're passed confusing the meanings of the words "objective" and "subjective"!

(FWIW, the smiley means I'm saying it with affection...talking **** with friends).

 

I felt Jaylen Warren was a bit underused by the Steelers last year. He’s always been productive and has a higher percentage of explosive runs (10+ yards). As a pass-catcher and pass-blocker, he’s very good. This year I expect to see a more 3-WR sets, which should open things up and give Warren even more opportunities in space. Mike M's offense is going to be tough to match up with.

 

Three seasons is a good dataset.

 

A lot to chew on here. I'll wait for a rainy day to look at per carry info.

But maybe here's a simpler way to compare their rushing values to the offense.

What % of rushes result in a first down? After all this is a key to keeping the offense on the field.

 

I like Warren too. I wouldn't have been opposed to him getting more usage last year.

My guess is that (1) they were already increasing his role a lot and they didn't want to go crazy and burn him out and (2) I think we were pleasantly surprised with Gainwell.

I wanted to see more of Warren this season, but I think the Dowdle signing means they'll shoot for a little less work for both of them than they had last season.

Hopefully, the D is better at getting off the field this year (and the O is better at staying on it) so we get more offensive snaps (near the bottom of the league last season).

 

Warren and Dowdle both have top 10 YAC so yeah we should utilize both more often. I have a feeling there will be close to equal carriers for both RBs not named Johnson.

 

You mean how success rate is subjective, and yards per carry is objective?

 

Really that question has so many variables you haven't mentioned to make the correct, or IMO correct answer. Is everyone healthy, what D is the opposition in?
What down is it??
How much time is actually left???
What's the weather like??
Am I feeling lucky??
LoL. Lots of stuff to consider. LoL