Grassy Knoll/Theory/Hunch/Little Angels and Devils on My Shoulders

[Ben E Lou]

Anyone else think that BPR is now a part of the YAC calculation?

[QuikSand]

It wouldn't shock me, there seems to be some synergy there.

[Ben E Lou]

My specific hunch is something along the lines of...

first x YAC (maybe 10 or so) = heavily governed by GD
does....he....go....all.....the.....way = heavily governed by BPR

[Hammer]

Anyone else think that BPR is now a part of the YAC calculation?


It always has been.

BPR is effectively breakaway speed, besides a WR being able to catch the long pass. A guy with maxed GD, but zero BPR will typically get you a lot of 20 yards after the catch type of receptions, but won't break the big one. A guy with zero GD and maxed BP won't get you many breakouts, but when he does he may go the distance. Obviously both is optimal.

In stat testing GD came out as a larger factor in YAC than BPR, but BPR wasn't that far behind.

[jeffrey]

See this (http://www.themoyes.com/rnfl/forum/phpBB3/viewtopic.php?f=4&t=1152&sid=3b28b3a58ec7b27c17410516b0c8af79).

I assume that's the "stat testing" Hammer is referring to.

By the way, the magnified height effect in the link above was because of a confound. Tight ends were included in the study, if I remember correctly. I think it was minimized when I redid it without TEs, but I don't remember. I also think I redid this for 6.3a, and found the same results, but again...I don't remember...

p.s. That's a temporary link as the RNFL is in the process of changing hosts. I'll try to remember to update when the move is complete. The new domain in the above address would be replicatednfl.com.

[Ben E Lou]

The problem with any testing on YAC prior to 6.3 is that it was pretty much cosmetic. It was calculated *after* the full play result. So, sure, guys with high BPR got bigger plays and then afterward got some YAC. But it was a stat that couldn't be trusted.

Completely different now.

Hammer's comment above does seem to fit the theory I'm throwing out there.

Of course, csv.txt hasn't been updated in years.

Column AU: Getting Downfield - ability to gain yardage after the catch.
Column AX: Big-Play Receiving - ability to catch long and very-long passes.

The Help File in this area is from either FOF2K1 or FOF4.
Receivers are scouted for their ability to catch what's thrown in their direction, their yards per reception, their catch frequency, their ability to catch third-down passes, their courage in catching passes thrown over the middle and their endurance. Tight Ends are rated for run and pass blocking and wide receivers are rated for their ability to return punts and kickoffs as well as their potential for breaking away from the defense for a big play.

[Ben E Lou]

(And jeffrey, I still say the magnified height effect is at least partially due to how JG creates great players... ;))

[jeffrey]

I just did a 6.3a version with the data I used for the QB 6.3a study. I'm not writing up the whole thing, but these are the results for YAC per catch.

height: .085 YAC per catch for each inch.
GD: 0.320 YAC per catch for each 10 points.
BPR: 0.204 YAC per catch for each 10 points.

Very similar to the previous result.

And interestingly popularity is 0.0425 YAC per catch for each 10 points, but I strongly suspect the causation on this one is reversed... It also has a p-value of 0.0013, so it isn't a very reliable result anyway.

This time, I only did it for WRs, so you're right that the height survives without TEs. My feeling is that this is causal, since Jim says height helps receivers. I don't know anything about how players are generated, though, so if you have any insights as you hint, I'm interested. I don't remember your thoughts on that from our last conversation about this.

[Ben E Lou]

Oh, I agree that it's causal. I'm just saying that I don't think that "height cause better performance" tells the entire story. My insight is a simple one--that at several positions (I think WR included) taller and heavier players show up more frequently in the high-ratings guys. This seems to be particularly true among the heavily masked. Among the guys who are unexpected creepers (i.e. the combines didn't indicate it), there seems to be an disproportionally high population of players who are much taller than the norm for their position. So, I'd add to "height causes better performances" something along the lines of "and good players are more likely to be tall than bad ones."

[jeffrey]

Good to know. It's worth noting, though, that the 0.08 above is probably largely independent from what your saying. I believe you that better players tend also to be taller, but that should be accounted for mostly by the inclusion of GD and BPR in the regression. That is to say, when you include one of two highly-correlated regressors (in this case, the regressors are abilities/attributes) in a regression, adding the other usually doesn't add much in terms of explanatory power.

Also, I should point out that this analysis just indicates that, in a linear approximation, the average YAC per catch over a given season depends on GD, BPR, and height. It says nothing about the actual mechanism. I don't know what Hammer is basing his interpretation of this on, but these data neither affirm nor deny it.

[Ben E Lou]

He's probably basing it on the same thing I am: lots of Solevision.

One other thing: I noticed some interest coefficient changes when I looked at only the population of players with a significant difference between BPR and GD. (like more than 30 points)

[jeffrey]

coefficients of what?

[Ben E Lou]

correlation

[jeffrey]

oh, you mean with height? I'm confused.

[jeffrey]

Oh wait, I see. I think you're saying there is an interaction between GD and BPR. I hadn't bothered to do that analysis. Probably worthwhile.

[jeffrey]

Just did a quick-and-dirty calculation. I sorted my data into 4 categories: lowGD/lowBP, lowGD/highBP, highGD/lowBP, highGD/highBP. Low and High are <50 and >50, respectively. The chart below shows the YAC per catch for these groups.

http://i656.photobucket.com/albums/uu287/nideic/lowGD-highGD.png

The two bars on the left are low GD for low and high BPR, respectively. The bars on the right are high GD for low and high BPR respectively. If there were an interaction we would expect the difference between the bars on the right to be very different from the bars on the left.

The difference between the bars to the right is a bit larger that that between the bars to left, but not much. A Guinness t-test indicates the difference isn't significant enough not to be accounted for by random variation (p = 0.1413). This could be due to a small sample size, though. I just don't have enough data. This was from 1291 receiver-seasons.

[QuikSand]

A couple of thoughts (in English) on this subject:

-It's possible that the current "calculation" of YAC for a given play is indeed more complex than before, but not all that much more... which could open up a lot of weird doors for things like "YAC per catch" if the generator remains fairly simple (like how do we fairly talk about two receptions when they are something like Yds=10,YAC=5 v. Yds=50,YAC=10). The notion that the YAC calculation built into a longer-by-design play might yield higher numbers than that built into a shorter-by-design pass play could have its own effect here, which would complicate a look at the effect of the BPR rating (which presumably alters the likelihood of a given receiver making longer-by-design receptions).

-Another variable here is the type of rating involved - BPR is basically a static rating, and GD evolves/creeps. Any analysis based on those ratings could have an inherent bias due to that difference alone.

[jeffrey]

Excellent points, both.

I'm inclined to think that the GD/BPR interaction works as you describe. That is to say, my guess is that the gain after the catch is a function of the length of the catch, rather than the gain after the catch being a function of the BPR (although, in this case, it would be indirectly a function of the BPR, since the length of the catch is likely a function of the BPR).

The second point almost certainly has an effect. It's just a matter of how great of an effect it has. I would assume that the effect is small since masking is gone after the first 3 to 5 seasons and because only some small percentage of players are masked (less than 1/4? I don't know), but the effect should still be felt to some extent.

[MRL17]

Just a thought, and I'm probably wrong, but could BPR's effect on YAC be explained away by the idea that longer passes have inherently higher amounts of YAC?

[MIJB#19]

Just a thought, and I'm probably wrong, but could BPR's effect on YAC be explained away by the idea that longer passes have inherently higher amounts of YAC?As in a deep pass has a built in higher chance to have YAC than a screen pass?

[MRL17]

Yes. I was thinking maybe short/screen passes have an average of 4 YAC, and medium/long/deep passes have an average of 6 YAC. Something like that

[Hammer]

Yes. I was thinking maybe short/screen passes have an average of 4 YAC, and medium/long/deep passes have an average of 6 YAC. Something like that

You will find that the high BPR guys will produce the long touchdowns much more often, regardless of the distance of the route. The GD is the initial factor after a catch, and then the BPR comes in after a certain distance it appears. Thats how I see it.

[QuikSand]

What it seems we're missing here is a clearer definition of the real underlying question.

What *I* want to know, especially for my team that uses short passing a lot (http://www.fof-ihof.com/teampages/summary.php?t=29), is basically the answer to this: "If my gameplan calls for an X-yard pass, what things increase the chance that upon completion, the play will go for more than those X yards, and if it does, what things will increase the amount of extra yardage it goes for?"

Right now, the simplest answer to that question -- just looking at the documents provided by the developer -- is that the two relevant ratings are QB Accuracy and WR Getting Downfield. To me, the operative question here is: should we add WR Big Play receiving to that list?

To the extent this conversation isn't completely soiled by now, maybe that helps with those who are considering this discussion.

[Ben E Lou]

Yes and yes.

[jeffrey]

The answer to your latter question, "should we add WR Big Play receiving to that list," is decidedly "yes" based on collected data and analysis.

As far as I can tell, what I've seen in this thread on the actual mechanism (i.e. the model of what roles these factors play, rather than just the identification of the factors) is idle speculation. I don't know to what extent these models are right or wrong, just that I'm not convinced by nebulous impressions.

[Ben E Lou]

Plenty of things are discovered via careful observation. This is one of them. A regression analysis isn't the end-all, be-all. *shurg*

[jeffrey]

Plenty of things are discovered via careful observation. This is one of them. A regression analysis isn't the end-all, be-all. *shurg*

I agree, but statistical analysis is precisely "careful observation" and is particularly suited for a game based on probability and statistics. Much of what was presented in this thread was not "careful observation" but rather anecdotal hypothesizing.

Regression analysis is certainly never the end-all, be-all. It's a very crude first-start, and a very limited model at that. It is useful, though, because it produces information that one can believe with some measure of confidence. My point in the above post was that anecdotal knowledge gets one into trouble. I think when Quicksand posed the question, he was asking for something more than impressions; for something we can be a bit more confident in.

I don't think you are guilty of it, but this thread is full of anecdotes and speculation being passed off as fact. I enjoy discussing my guesses about how this game works as much as the next guy. If I'm going to make assertions instead, I should probably have justification if I want it to be taken seriously.

I'm sure that is why you started the thread as you did. You had a (very good) intuition that BPR was an integral part of yards-after-the-catch and you presented it as such (hence the title of the thread). You didn't try to pass off your intuition as certitude.