Team Pass Defense % Help

[Dave Hansen]

Through 7 games my team has a 65.2% Pass Defense % rating as shown on the team summary page. On the team summary page I have 16 players listed with defensive info...all 16 have a PD% rating of 73 or better.

Starters:

*LCB-73.1
*RCB-81.7
*SS-88.9
*FS-82.3

I need help understanding how the team PD% is so low compared to all the individual player ratings.

6.3b

TXS

[QuikSand]

Well, just for starters... I'm not sure anyone *needs* to understand this. Even if you look at it,all you really *need* is to know that a higher number is better, and that it's some sort of aggregated statistic combining catches allowed, passes defensed, and interceptions.


Okay, with that in mind, here's my quick sketch how it works:

-For every player, the rating starts at an odd place (82.5) when there are no stats in effect

-After that, catches allowed (on a per-pass-play basis, as is everything here) are a demerit, driving the number downward

-Pass defenses and interceptions have a positive effect, driving the number upward

(The exact formula is not documented, nor had it been completely cracked... the FOFL has taken a stab at it, but it was deliberately complicated by the game designer for some reason)

From that, you generate a value for the player. Look around your league, and you'll see what values seem to represent a quality performance. For starting defensive backs, a value in the high 70s or better is generally a sign that either the player was playing well, or the defense as a whole was clicking well (hard to separate, for reasons beyond this post). Not many guys manage to get above the 82.5, but most who do are probably either part-timers or high quality players... it's not that common.


Now... basically the same calculation is used, in the aggregate, for the entire defense's rating. The thing is, the weighting of the components is basically skewed in the negative... so even if you have 7 relevant guys (LB and DB) contributing stats, and even if they were all pretty good (like a rating of 80, for instance) -- the way they got to 80 (from 82.5) was by only contributing a small net negative from the base. This is why the aggregate number doesn't just look like an average, it looks like a cumulative effect of everyone together, but starting from the same arbitrary base of 82.5. So, a whole team of guys rated 80 might yield a teamwide number of something like 66 (I don't know exactly) -- because each guy is contributing some minor downward toward a larger total.

Sorry if that's a bit opaque, but that's basically how it works. Hope it is helpful. I honestly do advise just skipping the detail and sticking with "higher number is better" if you look at this stat at all. I actually think it has some value as an overall indicator (something like the QB passer rating formula) but surely has its limits, especially given some of the play-resolution mechanics that underlie the whole thing.

[Dave Hansen]

TXS QS...

Now that I know that there is nothing peculiar about my league/stats...that all others see and experience the same variation...then higher the better works for me.

[bulletsponge]

seriosly 73.1 for a starting cb is horrible. id put in another corner there even if he isnt rated as highly and give him a few games to see how he does

[MIJB#19]

seriosly 73.1 for a starting cb is horrible. id put in another corner there even if he isnt rated as highly and give him a few games to see how he doesAt the same time, 7 games is just a little bit short of reliable data. We're also seeing a SS here at around 89. These two might even be related, in a sense that the CB is credited with catches allowed that over a full season the SS and other players would be credited with.

[Koprnkc]

If I remember correctly, wasn't 80% PD considered a shut down corner or safety?

[QuikSand]

If I remember correctly, wasn't 80% PD considered a shut down corner or safety?

I think it's pretty tough to make a judgment as simple as that, but that's not a bad rule of thumb. On a good team, a high quality starter ought to be able to post 80 or so over the long haul, I'd say that's about right. Down to 75 or lower and either it was a "bad year" or the guy is just not performing.

Trouble is, there are a lot of variables here, some of which are fairly understandable (on good teams who have the lead a lot, they expect the trailing opponent to be passing and benefit from that expectation and generate better numbers defending the pass) and some of which are not (this game has some totally weird ways of "assigning stats to individual positions, and some of them are awfully close to being "bugs" rather than just curious design decisions). So, I try not to get too terribly invested in the minute details of the player-by-player stats and focus on whether the bars make sense and the teamwide outcomes are what I want.

[MalcPow]

Fwiw, Team Pass D % is on a different scale and is only meant to compare teams to other teams. Basically 40 is the bad end and 80 is the good end, with potential extreme outliers on either side. The number itself is not an average of individual ratings or a universal representation of the individual formula.

I just take the approach of 40-80 being the team range and 70-90 being the individual range. You may see teams/guys outside of that, but they're pretty far from the norm.

[Ben E Lou]

...or just use QB Rating against, which is a much easier-to-understand metric, and usually falls in roughly the same order as Team PDPct.

[QuikSand]

Fwiw, Team Pass D % is on a different scale and is only meant to compare teams to other teams. Basically 40 is the bad end and 80 is the good end, with potential extreme outliers on either side. The number itself is not an average of individual ratings or a universal representation of the individual formula.

I don't often challenge you, sir, but I think I would here. Are you confident that it's not just an aggregated composite of all the relevant events by the whole team? My explanation above may have been a bit muddled, but I'd need to be talked down from believing that's basically how it works.

[Ben E Lou]

That's very easy to check. Gimme a few minutes...

[Ben E Lou]

Either the formula we've been using (that virtually *always* gets it the same as in-game for individual players) is wrong, or MalcPow is right. The results are different.

[QuikSand]

Sorry if this question is insulting, but how many pass plays are you counting in the composite? My working theory had always been that the counting stats add, but that the number of plays does not, and that it conceptually worked out that way just fine, and that also explained why the range for a team is lower than the range for a player (since it's an inherently imbalanced +/- calculation, basically covered up by the arbitrary starting point).

[Ben E Lou]

Sorry if this question is insulting, but how many pass plays are you counting in the composite? My working theory had always been that the counting stats add, but that the number of plays does not, and that it conceptually worked out that way just fine, and that also explained why the range for a team is lower than the range for a player (since it's an inherently imbalanced +/- calculation, basically covered up by the arbitrary starting point).
I tried it both ways: sum(passplays)/11, and just sum(passplays). Neither one worked out to what's reported in-game.

EDIT: Sum(passplays) puts everyone right between 80 and 82, basically. Sum(passplays)/11 gives a lower range than is reported in-game. For example, last year in the WOOF, the teams ranged from 42.8 to 76.1. Sum(passplays)/11 makes the range 34.2 to 69.8.

[Ben E Lou]

Oh yeah: it also orders the teams slightly differently.

[QuikSand]

Hmm, curious now. Looking at my BAL team in the CCFL, I see these raw stats:


144 pass plays
18 PD
76 catches allowed
0 interceptions
in-game PD% 50.4

I don't have the "best guess formula" handy, but does that match up?

[QuikSand]

I tried it both ways: sum(passplays)/11, and just sum(passplays). Neither one worked out to what's reported in-game..


I don't speak whatever that is, but I'll fake it. i don't think either of those is what you want to use there, I think you want the *actual* pass plays the team faced. On most teams this ought to be a shade higher than the maximum value for a full time player... but not an average of the 11, nor a sum.

[Ben E Lou]

82.5 + (505*sum(ints) + 240*sum(passesdefended) -135 * sum(passescaught))/sum(passplays) as pdpct

41.25

[Ben E Lou]

On most teams this ought to be a shade higher than the maximum value for a full time player... but not an average of the 11, nor a sum.???

"Uh, Quik."

You can only have 11 guys on the field per play. Sum(passplays)/11 matches up to number of plays faced reported in-game 100% of the time.

[QuikSand]

yy, got it, sorry (i thought you might have just been using the top 11 guys or something). i just pulled it from the team stats rather than player stats, but ends up the same place... my bad

[QuikSand]

Hmm, how 'bout that. Not that the real end product is really materially different, though, I guess.

Have to move this from "incomprehensible formula I think I understand but still find pointless" to "incomprehensible formula I can't understand and definitely find pointless."

Glad to clarify things a bit there. Sorry for my doubts.

[Ben E Lou]

Have to move this from "incomprehensible formula I think I understand but still find pointless" to "incomprehensible formula I can't understand and definitely find pointless."+1.

Which is why some dummy said earlier in this thread:


...or just use QB Rating against, which is a much easier-to-understand metric, and usually falls in roughly the same order as Team PDPct.;)

[QuikSand]

Well, I can't help but *want* to know how things work. And I'm not necessarily sold that running and scraping up the passer rating (which I know, and know its limitations) makes a better metric than the team PD% (which is conveniently calculated for me, presented in the team summary, along with a ranking). I'm still not sold that the correct view isn't from some other dummy above, who posted:

Even if you look at it,all you really *need* is to know that a higher number is better, and that it's some sort of aggregated statistic combining catches allowed, passes defensed, and interceptions.

[Ben E Lou]

Hehehe. I guess that's a fair point. Passer rating against isn't in the most convenient place in FOF, but it *is* there. I guess what I'm saying is that if I'm trying to determine not just where I stand, but how good I am, I'd rather use a metric that I know than one that's fairly arcane.

[QuikSand]

Some time ago, back in the patching process, if we were having this discussion about PR% (arguably the closest relative in the FOF game to this PD% thing), we would have found that it, too, didn't work the way one might have expected. It turns out there was a bug in the formula -- IIRC, sacks were being counted only 1/10 as much as they should have been.

Anyway... I'm still holding out hope that the simplest explanation is still in play here -- that this manufactured composite is in fact meant to be calculated the same way at both levels, and that there's some sort of implementation error, rather than just a wholly different approach.

Perhaps at some point, I could do some tinkering with a one-game team summary, and try to do some working backward. Assuming this is a linear thing (though the imprecision introduced in FOF 2007 always bugs me about that) it ought to be crackable, especially if there's just a typo of some sort.

[MalcPow]

I'd be interested to hear how any research turns out. It may be two or three years now since I've looked at this at all. (Right around when PR% was fixed. And I think the issue there was a double counting of plays or something.) If I remember right, support responded in an email that the Team % is meant simply to compare teams to other teams and not necessarily relate to what you see among individual players.

That said, I'm not sure that explanation rules out some sort of calculation error with something that is meant to aggregate team stats. I read it as the formula was scaled differently. There might just be some sort of funky calculation screw up going on instead.