Read defense skill

[thenewchuckd]

Ok, I keep seeing people bring this up and it is something that has started to bother me recently. I see people say that all the QB's read defense skill does is tell him when to avoid double coverage. And while I thought that seemed a weird way to develop the game, I accepted it for fact for a while. However, more recently, I have been noticing QBs with really high read defense throw into double coverage often enough, especially if there seems to be an exploitable match-up there. But maybe I was experiencing small sample effects?

So I decided to test it out. I started a test multi-player league where I controlled two teams: Baltimore and the New York Jets. It just so happened that these two teams were playing each other in week 1 of the season. I gave New York the best QB in the league at read defense (Drew Brees - 92 read defense). I also gave them the best WR in the league (Santonio Holmes - 84/84) and an ok 2nd WR (Braylon Edwards - 60/60).

I was less generous with Baltimore. I put a pretty sucky guy at LCB (35/35) and an ok player at RCB (56/56).

In terms of game plans, New York had a pretty pass heavy one on offense (a modification to one that I use on a regular basis). Baltimore had a neutral defense.

Then I ran two tests - 10 games for each test. In one test, Baltimore double covers the flanker all the time (Holmes was set as the flanker). The sucky LCB was covering the flanker, too. In the other test, Baltimore ran no double coverage (no other change to the defense). Baltimore was never set to blitz more than 1 player. Here are the results:

Test 1
Holmes average per game: 4 catches, 8.5 targets, 80 yards, 41 YAC

Test 2
Holmes average per game: 5.5 catches, 7.5 targets, 97.5 yards, 31 YAC

Again, those averages are over 10 games (when I say average, I used the median because it is less affected by outliers).

Checking the logs, there was indeed no double coverage for test #2. For test #1, Holmes was doubled for about two out of every three passes thrown to him. There was no double coverage if Holmes wasn't doubled (as expected). Now, you will have to explain to me how Holmes was not doubled on every play. For sure, there were some run aggressive defenses, where I understand there can be no double. On the other hand, I saw other situations where no one was doubled but it looked like a double was possible. But that's another issue.

Anyhow, what to conclude from this? For sure, it seems clear that even QBs with near maximum read defense will throw into double coverage. Also, if a WR is the best option on the field, the QB is going to throw to him even if he is doubled.

But does this debunk the theory that read defense = throw away from double coverage? And if so, what does the read defense skill actually do?

Not convinced? Sample size still too small for you? Comment away.

[QuikSand]

To help save people the time I spent trying to figure out what you did, let me abbreviate:


Test 1, 10 trials, heavy double coverage on top WR
Top WR's median results: 4 catches, 8.5 targets, 80 yards, 41 YAC

Test 2, 10 trials, no double coverage on top WR
Top WR's median results: 5.5 catches, 7.5 targets, 97.5 yards, 31 YAC


Please correct me if I have erred in synthesizing here.

[QuikSand]

If I am understanding correctly (and I'm not 100% sure that's so) then a few comments/observations:

-kind of odd that you're presuming to test the effects of the read defense rating, but then you don't make that a variable in your test

-looking at only ten games, and oddly using a median only, leaves me unmoved generally

-target percentage might be valuable here, rather than raw number of targets

-I don't recall how double coverage works exactly -- I don't recall being able to create a defense that "double covers the flanker all the time" -- but we'll have to take your word that SH was getting the doubles in the first trial



I won't rule out that there's either something wrong, or not as billed, with the read defense skill. Not sure what I'd make of that, if it were proven true. Not sure yet that we have much evidence to suggest that it doesn't work more or less as billed.

[thenewchuckd]

QuikSand: I really appreciate your comments. I must admit that I had no idea how to test my theory that read defense was not "avoid double coverage" as so many people seem to believe. But after watching recent games, I was just starting to get a feeling about it. So this is where I started. If there are other suggestions on how to test this, I am really open to doing just about anything.

Let me try to answer your points:

-kind of odd that you're presuming to test the effects of the read defense rating, but then you don't make that a variable in your test

Well, it is a variable in the test, in that the QB in question has the best read defense skill in the league. So if he is the best player at avoiding double coverage, wouldn't you see some difference in the target numbers? How else would you include it as a variable?


-looking at only ten games, and oddly using a median only, leaves me unmoved generally

As I said, I used the median because it works better in the presence of outliers. But I can present whatever numbers you would prefer.


-target percentage might be valuable here, rather than raw number of targets

That's a good point but I did not keep that information. I could set up another test looking at target %. While I'm at it, anything else that you would include?

On the other hand, while that might give a more definitive answer, I still think this test has its merits. This QB who is supposedly fantastic at avoiding double coverage somehow chose to throw into it quite often in this test.


-I don't recall how double coverage works exactly -- I don't recall being able to create a defense that "double covers the flanker all the time" -- but we'll have to take your word that SH was getting the doubles in the first trial

That's what you get when you line up your corners by the side of the formation and double cover the strong side. Although, like I said, it still does not "double cover flanker all the time". For sure, when there is double coverage, the flanker is doubled. The problem is that on a good deal of plays, no one gets doubled.

[gstelmack]

Well, it is a variable in the test, in that the QB in question has the best read defense skill in the league. So if he is the best player at avoiding double coverage, wouldn't you see some difference in the target numbers? How else would you include it as a variable?

Run the same test with the QB who is worst in the league. Although ideally you'd create the roster from scratch setting up identical QBs except for the Read Defense attribute.

[thenewchuckd]

Run the same test with the QB who is worst in the league. Although ideally you'd create the roster from scratch setting up identical QBs except for the Read Defense attribute.

Ok, I can test that out. However, before I run any further tests, I want to see if there is anything else I should include. For example, Quiksand mentioned target %, which seems quite logical.

PS Having absolutely no knowledge of how to create players, I won't be able to do exactly as you suggest.

PPS I still maintain, however, that if you can prove no difference in targets for the high read defense QB, that the point is already proven. You can debate whether I have the wrong data or not enough data, but it is tough to debate the point that a high read defense QB is consistently throwing into double coverage.

[QuikSand]

Well, I guess if I could call the shots on a re-do, I'd be thinking:

For each trial, keep track of:
-targets, catches, yards, TD, and YAC (why not?)
-target percentage (or, if it's easier, total attempts...and maybe total passes targeted to any WR)
-number of instances where flanker was targeted despite being double covered
-number of instances where other WR was targeted despite being double covered
-number of instances where QB threw away from DC to other WR
-number of instances where QB threw away from DC to other non-WR target

I think more trials would help a lot. 10 games is less than one season (obviously) and I know how much stake I put into the results from one full season in any testing that I have done (i.e. nearly zero). I don't think this sort of thing requires a ton of trials, but if we really are talking about small differences in effect, maybe 30 each?

[QuikSand]

One other thought would be to reduce cohesion as a variable as much as possible... so if you're going to run testing with two different QB, make sure they each have the same duration on the team. Ideally, probably best to just trade for two new guys and play it out with each of them for separate trials.

[thenewchuckd]

Ok, when I get a chance I will redo with your suggestions. I will note:

-I did keep track of Edward's targets (the WR2). However, I did not post them up because I already thought it was information overload (and the target situation was about the same, much like Holmes). I did not think about target % or total attempts but it is a very good suggestion.

-I did keep track of the number of doubles. Again, to avoid information overload, I simplified by just saying that Holmes was doubled about two thirds of the time in the 100% double scenario. There was never a case where another receiver was doubled. Along a similar track, in the 0% double case, there was never a case where a receiver was doubled.

-I kept track of the "threw away from double" cases, I just did not post. Again, when the pass was not to Holmes, it was about two thirds of the time.

-A couple of things about the 10 games. First, I wanted to gauge people's reaction first. No point in simming 60 games if someone already knows the answer to the question or knows a better way to test this. Second, even with 10 games, I would still expect to see some difference. And I would not expect to see Holmes get more targets if doubled... And to me, from looking at the logs and box scores, it seemed clear that Holmes was the preferred receiver whether he was doubled or not. He was usually the #1 receiver in terms of targets. I just wish I kept the target % numbers to prove it.

[thenewchuckd]

Now, to answer the question of why I thought the "avoid double" theory would be a weird way to design the game... It is just too gamey. If it were true, it would allow you to do what I described in this scenario and get away with it. That is to say: if you are up against a QB with high read defense, put your worst CB on the other team's best WR. Then double cover him.

If the "avoid double" theory is true, then the high read defense QB should throw away from his best WR often. Even though he is still probably the best option on the field. That's not what I would want.

On the other hand, as I have been experiencing recently and what I hope to show is... The game does not seem to work that way. In fact, I have another slight feeling that even a very solid CB, on his own (without doubling), can make a high read defense QB throw to another WR. But that is the subject of perhaps another test... One thing at a time.

[Ben E Lou]

Why does it have to be "often?" What if it's 25% of the time, versus 5% of the time?

[Johninja]

I always read avoids double coverage as he knows the guy is in double coverage but can make the throw anyways. Think there is somthing in the game logs about "he threw away from the double coverage"

[thenewchuckd]

Just for more info on the original test:

Test 1, 10 games, heavy double coverage on top WR

Top WR's totals: 41 catches, 90 targets, 848 yards, 469 YAC, 64 double covered, 26 single covered

2nd WR's totals: 31 catches, 49 targets, 605 yards, 239 YAC, 31 "threw away from double", 17 single coverage

Test 2, 10 games, no double coverage on top WR

Top WR's totals:: 57 catches, 77 targets, 1047 yards, 433 YAC, always single covered

2nd WR's totals: 38 catches, 56 targets, 499 yards, 83 YAC, never "thrown away from double coverage"

That's all the information that I kept, unfortunately. I will test again with more iterations but it may be a week or so. This time I will keep the target % data.

[thenewchuckd]

Why does it have to be "often?" What if it's 25% of the time, versus 5% of the time?

Yeah, that could be one explanation. We shall see. If that's the case, then I must say that I am pretty unimpressed with the read defense skill.

[thenewchuckd]

I always read avoids double coverage as he knows the guy is in double coverage but can make the throw anyways. Think there is somthing in the game logs about "he threw away from the double coverage"

For sure, in the log it is a display thing so you know if the throw was into double coverage. Or, on the other side, if the throw was into single coverage but there was another receiver doubled.

The display point definitely matters because, even from my small array of numbers here, you can see that a double covered receiver will perform worse than if he was single covered. Similarly a receiver will benefit if someone else is double covered. Now whether the penalty/benefit goes far enough is another question, but this performance aspect has already been proven by others.

However, it has been (widely?) accepted in the FOF community that a QB's read defense skill is all about avoiding double coverage. I am trying to show that maybe that's not the case.

[Johninja]

I think there may be somthing else to the read defence bar as well. Kind of like the play diagnosis bar for defenders. My hunch is that it increases the frequency of hidden offensive familiars (the opposite of defence seemed familiar with that play). Will be interested to see what you come up with.

[PiemasterUK]

I guess the easiest way to test this would be to start a game and give Team A two QBs, both with similar ratings except one had very high RD and one had very low RD. Then run the first game of the season 50 or so times, half with one QB starting and half with the other QB starting and look at the differences in the stats. There should be a big difference in how often they throw to double coverage, but you oculd also look at how often they throw to the best receiver, how often they throw to the receiver covered by the best CB etc.

[Ben E Lou]

... There should be a big difference in how often they throw to double coverage....Why? I say again...Why does it have to be "often?" What if it's 25% of the time, versus 5% of the time?To flesh that out, where does it say that a QB who is great at this skill avoids double coverage all the time, or even most of the time. What if it simply means that a crappy one recognizes and avoids it 5% of the time, while a great one does it 15%, or 20%, or 25%. Treating this like an all-or-nothing proposition will likely get you in trouble.

[sidthelid]

A while back I wrote down over a 1000 plays where double coverage was used. I recorded the QB's overall rating and his Read bar. What i found was there was little difference between a QB with an 80 read bar and a QB with a 30 Read bar when it came to throwing into double coverage. I came to two conclusions
a) the read bar does not work or
b) The Brady to Moss scenario, that being that the QB would throw into Double Coverage as he can make the throw. Maybe just because you have double coverage it does not mean it's good coverage!

Just my thoughts :)

[strickzilla]

perhaps (as usual) ben has a point, also johninja and sid may have hit it perhaps its the QB reading EFFECTIVE double teams. in other words if your playing base 4-3 cover 2 and doubling a wr the only "free" player would be a LB with sub par coverage skills so it not too risky

as opposed to in a cover 1, nickle or dime package the doubler would be another DB better coverage skills thus the qb looks elsewhere?

so perhaps an intrestign study would be to test against an all base (3-4/4-3) defense and against nickle/dime and see what happens if what i and others are thinking is true the targets would decrease as you go from base to nickle to dime

[thenewchuckd]

perhaps (as usual) ben has a point, also johninja and sid may have hit it perhaps its the QB reading EFFECTIVE double teams. in other words if your playing base 4-3 cover 2 and doubling a wr the only "free" player would be a LB with sub par coverage skills so it not too risky


Actually, my theory is more that it is about "effective coverage" and has nothing to do with double coverage at all.

When I really started to think... I thought that on many plays, there are two possible receivers. What if the skill is more about reading the best receiver to throw to?

As I mentioned before, one of the things that I noticed was this "throwing a lot into double coverage" issue. But the other thing that I have noticed is that, if a receiver is covered by a great corner, a good read defense QB will tend to look to a more favorable match-up. Still no evidence on the second part of my theory, but the evidence is starting to build on the first part based on 20 more trials (unless you think a 1% difference in target %, double vs no double, is significant).

[QuikSand]

So, if we are thinking there's a full alternative explanation for the rating, maybe we test for that, too... right?

If there's a theory that the RD rating directs the QB to "the best fit" or somesuch, then it seems to me your variable is the quality of WR2, the non-doubled guy. With a quality WR2, the high RD quarterback out to throw away from the DC a lot, with a stiff as the alternative, he should force it to the better guy despite the doubling more often. Basically right?

Rather than shrug our shoulders, let's take chuck at his word, and try to help direct his admirable enthusiasm toward any real insight we might be able to glean from this. Seems like that might help.

[thenewchuckd]

So, if we are thinking there's a full alternative explanation for the rating, maybe we test for that, too... right?

I'm going to continue with my test for the moment, maybe I am getting excited too easily. I am up to a 2.5% target difference at 28 trails... Maybe it will go higher. ;-)

Actually, with more trials it does seem that the target difference is non-negligible for double vs no double. But the difference is still pretty small. I don't think it will be enough to completely disprove my point.

[strickzilla]

touching on your great cb vs wr another possible varible is how much "better" the CB has to be before the QB looks else where. this would explain a lot about my abysmal pass d ive been working under the double the weaker wr with my lower cb and leave my "better" guy "on the island" so if my "better" guy isint much better that would lead to the 32nd ranked pass d i have

[thenewchuckd]

this would explain a lot about my abysmal pass d ive been working under the double the weaker wr with my lower cb and leave my "better" guy "on the island" so if my "better" guy isint much better that would lead to the 32nd ranked pass d i have

I think what you have to realize is that it is a trade off. Essentially what you are doing is equalizing your pass d. You have to try and figure out if that's a good thing or a bad thing. What I have seen from this particular example is that it is essentially a damned if you do, damned if you don't type of scenario. Holmes usually has a fantastic day whether he is doubled or not. He does a bit less well when consistently doubled (certainly in terms of catches and yards, although it is less clear for targets), but Edwards benefits from whatever small bit that Holmes loses.

On the other hand, maybe things would be different with a QB who can't read defenses quite as well. And another scenario would be a receiver like Holmes with only crappy receivers around him, etc. Still too much testing to do...

By the way, you said this:

in other words if your playing base 4-3 cover 2 and doubling a wr the only "free" player would be a LB with sub par coverage skills so it not too risky

My understanding is that, for the vast majority of the cases, one of the safeties will be in double coverage. I think that's what it says in the manual.

[thenewchuckd]

Yes, from the player guide:


Double-coverage means that a second defender (usually the free safety
or a nickel back) is assigned a responsibility to join the cornerback as
soon as the play begins.

Also interesting:


their ability to read the opposing defense, which determines how effective they are avoiding double-coverage to throw to a secondary receiver.


But then, there is also this:



You may double-cover either the split end or the flanker in most cases.

Double-coverage means that a second defender (usually the free safety

or a nickel back) is assigned a responsibility to join the cornerback as

soon as the play begins. This doesn’t mean that double coverage is
impossible without this selection, as the defensive backs manning the
deep zones will join in when a receiver enters their zone







Which seems to imply that double coverage can mean a few different things in the manual, since it is fairly easy to show that when you set all double coverage to 0, that none shows up in the game log (even if you run a lot of 2 deep defenses).

[QuikSand]

A lot of this is drifting toward a fairly uneasy direction of trying to explain FOF in real football terms. Some time ago, this (actually really focusing on the relative importance of different coverage skills by different defenders in different formations) was a subject of a lot of debate and discussion... which ultimately proved very frustrating to a number of members of the FOF community (myself included).

My take-away from that discussion, which I think echoes here, is that it's important to remember this is a computer game. It's not an assembly of little men playing a game. As football fans, we really want to know "who is the free safety covering on this play?" but we might have to settle for the answer being the equivalent of "that's not really a question that has any meaning in this computer game."

Recognizing the limits of our inquiry is difficult, but valuable, I think. The game does actually tell us certain things about double coverage -- specifically when a pass is into, or away from, a double. I think it's a fair assumption that this is a meaningful element of the game's mechanics, and other inquiries have backed that up (all things equal, it seems that double coverage makes passes to that target less effective). So, from that In conclude that it's that information that is the best foothold for study into double coverages, and not anything that gets to far afield from what we think might be happening based on our knowledge of real football.

[dubb93]

I think it is important to note that even if the effect is small, that it is really not small.

In the NFL a QB completion % of over 62% puts you in very good Manning/Brady/Brees/Rodgers/Flacco/Ryan company while a completion percentage in just the high 50's, like say 58% puts you in the Ortan/McNabb/Cassell/Fitzpatrick/Collins territory of being replaced and/or benched mid-season. To be fair the elite of the elite are around 65% but even at 62% you are talking pro bowl level while 58% is backup level.

So if we assume that say this one attribute is increasing the QB completion percentage by 2-5% that is in itself significant. Something around 10-25 % would be expecting something Madden like.

[QuikSand]

So if we assume that say this one attribute is increasing the QB completion percentage by 2-5% that is in itself significant. Something around 10-25 % would be expecting something Madden like.

But this is missing a step in the math. We're talking about how frequently does the good RD quarterback switch away from double coverage? If that number is something like 10% of the time, you then need to multiply that 10% by the differential in completion percentage (between DC and non-DC) to get the proper net result of his decisions.

[thenewchuckd]

Yes, no final numbers yet. But I am sort of in QuikSand's camp here. If I am to read from the results so far that the best QB at read defense is only going to gain a 10% difference in the number of times the top receiver is targetted, I say that the read defense skill is just about useless. In other words, that means that if you double Holmes all the time, his target % would be around 30% while if you don't double him would be around 33% (again, I don't have the final numbers but I am sort of basing this example on reality).

That says nothing about completion %, which probably stays about the same either way. In fact, if you double cover, I could make the argument that the overall completion % could be better no matter what, since even a fantastic game for a receiver still means the majority of the passes are going to someone else.

But if the main goal of read defense is to get balls away from the doubled WR, a 10% difference for the top player in the game is a pretty poor job by my estimation.

[dubb93]

But this is missing a step in the math. We're talking about how frequently does the good RD quarterback switch away from double coverage? If that number is something like 10% of the time, you then need to multiply that 10% by the differential in completion percentage (between DC and non-DC) to get the proper net result of his decisions.

I wasn't trying for my math to be absolute. Maybe I didn't explain what I was saying well enough.

What I meant was that if the QB was avoiding the DC about 10% of the time(not a real number but pulled completely out of my backside) then I would guess that completion % would increase by about 2-5%(he isn't going to complete every pass away from the DC nor is every pass that was going to the DC going to be incomplete so a 10% difference throwing away from the DC can't = a 10% increase in completion %) based on the numbers in post #13 and the awful job done of throwing to WR 1(again I didn't do any math here, just quickly looking at the numbers I was given). And if that is the case then I think it is significant. I wasn't really actually doing any math at all. :D

[thenewchuckd]

I should also comment that Ben's 20-25% avoid double could be about right. You have to consider a couple of things:

1) I can only get Holmes doubled on about 70-75% of the balls thrown to him. I got the number up a bit by taking out all run aggressive defenses in the game plan. That's in addition to never blitzing more than 1 player (like I had before). But 70-75% is still about my max. I'm not sure if anyone can explain that one?

2) There are some plays where there is no secondary receiver.

Now, I am not sure if points #1 and #2 are related? But certainly, it adds something extra to the equation.

[thenewchuckd]

And if that is the case then I think it is significant. I wasn't really actually doing any math at all. :D

I think you still missed the point. As I said in my post, if he is avoiding DC 10% of the time, it means Holmes' target % goes from 33% to maybe about 30% (give or take). So 3% of those balls are going to someone else.

Let's say, to make things interesting, that the completion % to Holmes in DC is 50% while away from double it is 70%. That means on 3% of the plays, you are doing about 20% better...

Multiply those two numbers, and your overall increase in completion % is only about 0.6%. And hopefully you will agree that my example was already pretty out there.

[Ben E Lou]

1) I can only get Holmes doubled on about 70-75% of the balls thrown to him. I got the number up a bit by taking out all run aggressive defenses in the game plan. That's in addition to never blitzing more than 1 player (like I had before). But 70-75% is still about my max. I'm not sure if anyone can explain that one?Perhaps this has changed, but I don't think it has. In the past, there was *never* double coverage reported when you're in run or run aggressive defense. Only pass and pass aggressive allow for doubling.

[sidthelid]

My take-away from that discussion, which I think echoes here, is that it's important to remember this is a computer game. It's not an assembly of little men playing a game. As football fans, we really want to know "who is the free safety covering on this play?" but we might have to settle for the answer being the equivalent of "that's not really a question that has any meaning in this computer game."


You can tell which 11 players are on the field on most downs. If that matters is another story as you rightly said.

[thenewchuckd]

Perhaps this has changed, but I don't think it has. In the past, there was *never* double coverage reported when you're in run or run aggressive defense. Only pass and pass aggressive allow for doubling.

Oh yeah, that's definitely it. I thought it was just aggressive run but a look at the logs show that the no double plays are all when the defense expects the run.

[thenewchuckd]

The results with 48 games (3 seasons) of data. The only difference is I switched QBs to Payton Manning. His read defense skill is the same (92) but he is a better overall QB. Receivers are the same. CB #1 still sucks, CB #2 is about average.

Test 1, heavy double coverage on top WR

Top WR's totals: 232 catches, 454 targets, 3720 yards, 1426 YAC, 29 TDs, 314 thrown into double, 138 thrown into single, 1466 pass plays, 31.0 target percentage

Test 2, no double coverage on top WR

Top WR's totals: 286 catches, 453 targets, 4177 yards, 1454 YAC, 26 TDs, always single covered, 1376 pass plays, 32.9 target percentage

Any comments? I will test a crappy read defense QB next... It may take a bit longer, though.

[gstelmack]

You can tell which 11 players are on the field on most downs. If that matters is another story as you rightly said.

Did someone else crack the Solevision format? If so have them get in touch with me. If not, well, that's the only way I'm aware of to know which 11 players are on the field. The game logs and Solevision playback tell you at most a couple of players involved in the play.

[Ben E Lou]

Did someone else crack the Solevision format? If so have them get in touch with me. If not, well, that's the only way I'm aware of to know which 11 players are on the field. The game logs and Solevision playback tell you at most a couple of players involved in the play.Well, there's another way: you could pause Solevision, open the box score, print it, then run another play, pause again, print, and compare the run/pass play counts. Pretty sure that's what this guy did.

[QuikSand]

The results with 48 games (3 seasons) of data. The only difference is I switched QBs to Payton Manning. His read defense skill is the same (92) but he is a better overall QB. Receivers are the same. CB #1 still sucks, CB #2 is about average.

Test 1, heavy double coverage on top WR

Top WR's totals: 232 catches, 454 targets, 3720 yards, 1426 YAC, 29 TDs, 314 thrown into double, 138 thrown into single, 1466 pass plays, 31.0 target percentage

Test 2, no double coverage on top WR

Top WR's totals: 286 catches, 453 targets, 4177 yards, 1454 YAC, 26 TDs, always single covered, 1376 pass plays, 32.9 target percentage

Any comments? I will test a crappy read defense QB next... It may take a bit longer, though.

I think the more data you assemble, the stronger the case becomes that the Read Defense, taken at face value to be the tendency of the QB to throw away from a double covered receiver, must not be a huge factor. I'm not sure precisely what I would have expected, but a drop from 32.9 to 31.0 at the outlier extremes of double coverage is a narrower range that I'd have guessed.

I also confess that I already don't pay particular attention to the RD rating in any serious setting anyway, whether in acquiring mixed-skill QB for my team, or in game-by-game evaluation of opponents.... so I doubt this changes my approach at all, even if there's a new conventional wisdom arising from this effort.

[Ben E Lou]

I also confess that I already don't pay particular attention to the RD rating in any serious setting anyway, whether in acquiring mixed-skill QB for my team, or in game-by-game evaluation of opponents.... so I doubt this changes my approach at all, even if there's a new conventional wisdom arising from this effort.Bingo...

[thenewchuckd]

It is sort of interesting that you guys think that read defense is a non-essential skill for a QB. I could only find one other thread on QB skills. It is an old one but many people indicated it was one of the top 3 skills that they look for in a QB.

Anyhow, I am going to keep testing. I'm just curious - would you suggest going to an all pass defense so that I can be sure that the flanker is always double covered? That seems a bit hokey since I don't think anyone would ever use a defense like that (the one I used here already has a pass bias). But it would make my life easier since I wouldn't have to comb through game logs, just box scores.

[gstelmack]

A quick search on "read defense" turns up:

QBs - Front Office Football Central (http://www.operationsports.com/fofc/showthread.php?t=54567&highlight=read+defense)

Note the dates, this info is 5 years and many patches ago, and may largely be based on past-FOFk24 experience.

[thenewchuckd]

Yep, that's the exact thread I was talking about where many put read defense in their top 3 for QB skills. But as you and I both said, it is an old thread.

But now, Ben and Quik are saying that it does not even register for them when evaluating QBs (at least, that's what I think they're saying).

[QuikSand]

Yup, that's what I'm saying.

[Ben E Lou]

...and that's what I'm saying, too.

[Firefly]

Actually... it would appear read defense is a very valuable bar NOT to have, because it inversely correlates to a higher avoid int%

[Steel]

Actually... it would appear read defense is a very valuable bar NOT to have, because it inversely correlates to a higher avoid int%

And there you have it.

[wade moore]

Actually... it would appear read defense is a very valuable bar NOT to have, because it inversely correlates to a higher avoid int%

And there you have it.

Hate to be "that guy" here, but...

Do we have any data to back this definitive statement up?

[Rizon]

Help File

Quarterbacks are scouted for their ability to complete screen, short, medium, long and deep passes, as well as their ability to complete third-down passes, their ability to hit a receiver in stride (accuracy) and their ability to take advantage of defensive miscues (timing). They are also rated for how often they choose to run when a pass is called, their ability to sense an opponent's pass rush, their ability to run the two-minute offense and their ability to read the opposing defense, which determines how effective they are avoiding double-coverage to throw to a secondary receiver.

It's similar to the "check down" skill a QB has ... in real life (the difference between a Drew Brees passing up 2 covered targets to throw to an uncovered 3rd VS Brett Favre throwing to his first target that is quadruple covered).

How that correlates to FOF and if it's important I don't know. I'd guess you have to factor in that every pass play only has one or two options, or just one check down.

* and this is checking off double-covered targets not single-covered vs wide open

*dola: just read the thread and I guess I'm preaching to the choir.

[Rizon]

So is the agreement that RD means Avoid Throwing Into Double Coverage OR Throwing Into Double Coverage Effective %*, but we aren't sure if a 80 RD rating is that much of a factor vs a 30 RD rating? (as said before)

* if that's the case, the help file is misleading.

[thenewchuckd]

So is the agreement that RD means Avoid Throwing Into Double Coverage OR Throwing Into Double Coverage Effective %*, but we aren't sure if a 80 RD rating is that much of a factor vs a 30 RD rating? (as said before)


I'm still simming (when I have the time). The only thing that I can tell you (almost) definitively is that a high read defense QB will not always throw away from double coverage. In fact, in this example (and another one that I did with the same number of iterations), the difference in target % for double vs not double, while non-zero, is pretty unimpressive.

Regarding the effective double coverage point (and also the low read defense QB), I am running a few scenarios and I hope to say something definitive eventually. It will just take a bit more time.

[thenewchuckd]

Ok, some people might be waiting for data and while I'm not done, I do have some interesting results to share.

First off - I started a completely new test. I wanted to mess around with a few different variables and use an example that was not as hokey. So I started a new league. I still used a Buffalo and New York game for my test sim... With the WR in question being Holmes and the high read defense QB is again Manning (97 read defense). Holmes is not as good this time - "only" 72/72 but still one of the best WRs in the league. On the other side, I put Vincent Jackson, not as good at 62/62 but still solid.

For the first iteration I wanted to use two CBs that were good and about equal strength. 60/60 guys - not fantastic but ok. The LCB was always on the flanker.

For the second iteration I put the best CB in the league (Champ Bailey - 82/82) next to one of the 60/60 guys. Bailey was always covering Holmes.

I did double / not double scenarios. In the double scenario, Holmes is always doubled (the defense is never expecting the run). Similarly, in the not double scenario, no one is ever doubled.

[thenewchuckd]

Results with the high read defense QB, 48 games simmed for each test:

Test 1, top WR always doubled, ok corners

Top WR's totals: 231 catches, 482 targets, 3005 yards, 757 YAC, 20 TDs,
1759 pass plays, 27.4 target percentage

Test 2, no double coverage on any WR, ok corners

Top WR's totals: 280 catches, 508 targets, 4510 yards, 1496 YAC, 25 TDs, always single covered, 1754 pass plays, 29.0 target percentage

Test 3, top WR always doubled, fantastic top corner

Top WR's totals: 199 catches, 431 targets, 2886 yards, 844 YAC, 19 TDs,
1652 pass plays, 26.1 target percentage

Test 4, no double coverage on any WR, fantastic top corner

Top WR's totals: 268 catches, 461 targets, 3592 yards, 1007 YAC, 27 TDs, always single covered, 1743 pass plays, 26.4 target percentage

[thenewchuckd]

Ok, the next set of results is for a low read defense QB (Jay Cutler = 25 read defense). His overall skills are also considerably lower (overall current rating = 39). Maybe that is an issue but I digress.

Results with the low read defense QB, 48 games simmed for each test:

Test 5, top WR always doubled, ok corners

Top WR's totals: 187 catches, 477 targets, 2446 yards, 723 YAC, 15 TDs,
1769 pass plays, 27.0 target percentage

Test 6, no double coverage on any WR, ok corners

Top WR's totals: 243 catches, 412 targets, 3307 yards, 1010 YAC, 21 TDs, always single covered, 1624 pass plays, 25.4 target percentage

I don't know how everyone else feels, but I feel like this is trending towards my original line of thought. That is to say, read defense is not about avoiding double coverage. I think it is more along the lines of finding the best receiver to throw to (or not being able to find him).

The impact is subtle, but I definitely think it is significant and a very valuable skill for a QB to have. However, I want to run test 7 and 8 (with the stud CB and the poor read defense QB) before finalizing the conclusions.

I think, if nothing else, this really shows the impact an elite WR has on a QB. Holmes is even making a crap QB look pretty solid, although he is happier with Manning throwing to him.

[aston217]

WRs aside, I wonder if the difference between a high RD QB and a low RB QB is simply the number of incidences of "____ threw away from the DC", ceteris paribus.

[thenewchuckd]

WRs aside, I wonder if the difference between a high RD QB and a low RB QB is simply the number of incidences of "____ threw away from the DC", ceteris paribus.

For the latest experiments, because of the way I set it up, the % of thrown away is 100 minus the percentage thrown into double. So in other words, based on my tests, that's not it.

Actually, once I post the last two experiments, I think it will solidify my theory even more.

[thenewchuckd]

Results with the low read defense QB, 48 games simmed for each test:

Test 7, top WR always doubled, fantastic top corner

Top WR's totals: 195 catches, 487 targets, 2600 yards, 804 YAC, 14 TDs,
1715 pass plays, 28.4 target percentage

Test 8, no double coverage on any WR, fantastic top corner

Top WR's totals: 228 catches, 449 targets, 3020 yards, 883 YAC, 19 TDs, always single covered, 1664 pass plays, 27.0 target percentage

[thenewchuckd]

So are you seeing the trend? The results are not over the top but it seems clear that if the coverage is better, the high read defense QB will throw to that spot slightly less. However... and this is the shocker... the better the coverage, the more likely it is for the low read defense QB to throw in that direction.

I bet you didn't see that one coming?

But like I said, the results are not spectacular. I am guessing route running is still the major factor as to where the QB throws. I will back pedal a bit since I'm still not sure this study shows that read defense is a significant skill to have.

On the other hand, maybe this is proof that read defense is more than just about double coverage? Any doubters?

[QuikSand]

I think there's a simpler explanation for these results than the complicated "it's about finding the best target." These results are unexpected, but I think there's a proto-logic that could still work here.

We all assumed that the RD skill was essentially on a zero-to-100 scale... that throwing away from doubles was a sort of binary decision, either the QB passed the RD test and he did so, or he failed the test and he didn't do so.

Here, it seems like it works on a different scale, something (keeping the terms comparable) like -50 to +50. The low-RD quarterback won't just fail to avoid the double -- he will actually misread the double and throw into it more.

I think that might turn out to be the simplest explanation of the data you're finding here, chuck. It's not as complex as a QB making decisions about what's best... it's much simpler than that. When there's a double coverage, the bad RD guy throw into it a little bit more, and the good RD guy throws away from it a little bit more. So far, that's the cleanest way I see to explain this.

I'm still pretty convinced that this is marginal enough to more or less ignore. It does, potentially, open the door to some weird outcomes, though. What if your best WR is really leaps and bounds better than all other targets? Maybe in that case your ideal QB would actually be a guy with a zero RD rating. Odd.


Thanks for the labor, chuck.

[thenewchuckd]

I think there's a simpler explanation for these results than the complicated "it's about finding the best target." These results are unexpected, but I think there's a proto-logic that could still work here.

I disagree. That's why I included the single coverage examples. Let me put all of the single coverage results in one spot for comparison.

First, very high read defense QB:

Test 2, no double coverage on any WR, ok corners

Top WR's totals: 280 catches, 508 targets, 4510 yards, 1496 YAC, 25 TDs, always single covered, 1754 pass plays, 29.0 target percentage

Test 4, no double coverage on any WR, fantastic top corner

Top WR's totals: 268 catches, 461 targets, 3592 yards, 1007 YAC, 27 TDs, always single covered, 1743 pass plays, 26.4 target percentage

Then, lowish read defense QB:

Test 6, no double coverage on any WR, ok corners

Top WR's totals: 243 catches, 412 targets, 3307 yards, 1010 YAC, 21 TDs, always single covered, 1624 pass plays, 25.4 target percentage

Test 8, no double coverage on any WR, fantastic top corner

Top WR's totals: 228 catches, 449 targets, 3020 yards, 883 YAC, 19 TDs, always single covered, 1664 pass plays, 27.0 target percentage

-----------

As you can see, the absolute best target result comes with the high read defense QB, ok corner (where you would expect it to come). The low read defense QB is 3.6% off of this target number - not a big difference but far from negligible.

In fact, if you put a fantastic corner on Holmes (still no double coverage), the low read defense QB actually targets him more (1.6% more).

I'm really having trouble seeing how the double coverage theory still holds with this evidence.

[Firefly]

That's interesting!

[Jughead Spock]

Ok, the next set of results is for a low read defense QB (Jay Cutler = 25
read defense). His overall skills are also considerably lower (overall current
rating = 39). Maybe that is an issue but I digress.



That kinda throws out the results for that test, IMO. Too many variables. I'm trying to find trends & threads, but can't yet.

[thenewchuckd]

That kinda throws out the results for that test, IMO. Too many variables. I'm trying to find trends & threads, but can't yet.

Maybe there is an argument that you can't compare the low read defense and the high read defense QBs. But the double vs no double, standout corner vs ok corner results stand regardless. It is the same QB for goodness sakes.

I'm also not 100% sold on your argument. It would mean that other skills besides read defense impact who the QB targets. If true, it would be a fantastic discovery on its own.

In any event, I can't do everything. It is up to others to do their own tests to figure it out for themselves. Alternatively, if someone can teach me how to create players, I will make two QBs with the only difference being their read defense skill. Then I will run more tests. I'm at a loss as to how to proceed otherwise.

[Ben E Lou]

Over the 22 regular seasons WOOF has played, right at 70% of double coverages have been avoided. (This number appears to be unaffected by cohesion changes, the big-stats era, or 6.3/6.4. In 22 seasons of WOOF history, the higest number has been 70.8%, and the lowest 68.5%. Actual overall is 69.66%.)
Single-season best for a QB with 300 attempts or better: 82.6%. Worst? 53.7%
Three of the worst give season in WOOF history (58.3%, 55.8%, 55.7%) were all turned in by the same QB. If any WOOF member reading this remembers anything about Luke Gaulin (http://www.younglifenorthdekalb.com/woof/ben/playercard.php?playerid=346), or wants to reload an old file, I'd be very curious to know what his bars were like.
Brett Shaw (http://www.younglifenorthdekalb.com/woof/ben/playercard.php?playerid=9850), with a 58.6% number and a low read defense number, turned in the 8th-worst performance in avoiding double coverage in league history this past year. And oh yeah, many people thought him to be the league's Most Valuable Player.
Among the top five guys in the league in percentage of double coverages avoided, the *best* QB Rating was 75.1.So, yeah, it appears that avoiding double coverage doesn't necessarily translate into better QB performance.

[Ben E Lou]

Heh. One more interesting data point. When looking at entire careers, the range is fairly small. 142 QBs in league history have faced double coverage 200 times or more. Among those, the range of avoidance is 63.5 to 75.4. Also, just by eyeballing, the QBs at the top *and* bottom of the list are relative unknowns. Nearly all of the top QBs by DogPoints in league history have avoidance percentages quite near the average. The one who does not spent a good bit of time throwing to only *one* good wide receiver, so I can't help but wonder if he just threw it to Will Barnes anyway. WOOF owners will recognize these names...

DeVito--71.1%
Beckles--70.6%
Phillips--69.7%
Spence--66.6%
Williamson: 68.3%
Passut--69.2%
Ackerman--70.4%

[gstelmack]

What are the completion percentages when throwing into double coverage, avoiding double coverage, and with no double coverage?

[Ben E Lou]

What are the completion percentages when throwing into double coverage, avoiding double coverage, and with no double coverage?If only there were a way to tell that... :devil:

[thenewchuckd]

First: thanks for all the data.

The one who does not spent a good bit of time throwing to only *one* good wide receiver, so I can't help but wonder if he just threw it to Will Barnes anyway.

That kind of fits with my theory that read defense is about finding the best option (maybe out of the two routes on a given play) and not about avoiding double coverage.

[gstelmack]

If only there were a way to tell that... :devil:

Oh crap, that's right.

[gstelmack]

Okay, WOOF 8 years of data, 2020-2027, 6.3/6.4 era:

TOTAL PASSING
Drop Backs = 193259
Attempts = 174792 (90.4%)
Completions = 100635 (57.6%)
Sacks = 10024 (5.2%)
Scrambles = 8443 (4.4%)

NORMAL PASSING
Drop Backs = 109002
Attempts = 98346 (90.2%)
Completions = 59400 (60.4%)
Sacks = 5876 (5.4%)
Scrambles = 4780 (4.4%)

INTO DOUBLE COVERAGE
Drop Backs = 26024
Attempts = 23513 (90.4%)
Completions = 9446 (40.2%)
Sacks = 1380 (5.3%)
Scrambles = 1131 (4.3%)

AVOID DOUBLE COVERAGE
Drop Backs = 58233
Attempts = 52933 (90.9%)
Completions = 31789 (60.1%)
Sacks = 2768 (4.8%)
Scrambles = 2532 (4.3%)

[gstelmack]

The two big notable numbers I see above are:

- Double coverage when the QB tries to throw into it drastically reduces completion percentage. If the QB does recognize it, there is minimal impact on completion percentage.

- If the QB recognizes and avoids the double coverage, he also tends to get sacked less often when he does it.

I may rerun these numbers tomorrow or over the weekend and include yards, TDs, and INTs. Will be interesting to see if avoiding the double coverage leads to shorter passes, and does double coverage affect INT rates.

[Yoda]

I am a bit surprized that the completion percentage doesn't go up when throwing away from DC..... changes my thinking a bit....

Is there a way to tell if there is double coverage on run plays? And if so, how that affects the rushing numbers?

[gstelmack]

Is there a way to tell if there is double coverage on run plays? And if so, how that affects the rushing numbers?

Not that I'm aware of. It may be buried in there, not sure. There are a host of fields in the Solevision for the different play types that haven't been cracked yet.

[gstelmack]

I am a bit surprized that the completion percentage doesn't go up when throwing away from DC..... changes my thinking a bit....

With just completion numbers there does not seem to be any reason NOT to double-cover on every single play. However I'm working on additional stats to show how yards per attempt / completion stacks up, and maybe throw in INTs as well. Still not sure if a QB who avoids double coverage has to throw short or long. If you give up less yardage because he dumps off, then yeah no reason not to double-cover every single play. If you give up more because he hits single coverage streaking downfield, there's your negative you have to weigh.

[gstelmack]

Once again, same 8 seasons of WOOF, 2020-2027, all on 6.3 or higher:

TOTAL PASSING
Drop Backs = 193259
Attempts = 174792 (90.4%)
Completions = 100635 (57.6%)
Thrown Yards = 1110547 (6.35 YPA, 11.04 YPC)
Yards After Completion = 329827 (1.89 YPA, 3.28 YPC)
Total Yards = 1440374 (8.24 YPA, 14.31 YPC)
Sacks = 10024 (5.2%), yards lost = 63119(6.30 avg)
Scrambles = 8443 (4.4%), yards = 44766(5.30 avg)

NORMAL PASSING
Drop Backs = 109002
Attempts = 98346 (90.2%)
Completions = 59400 (60.4%)
Thrown Yards = 656124 (6.67 YPA, 11.05 YPC)
Yards After Completion = 197745 (2.01 YPA, 3.33 YPC)
Total Yards = 853869 (8.68 YPA, 14.37 YPC)
Sacks = 5876 (5.4%), yards lost = 37060(6.31 avg)
Scrambles = 4780 (4.4%), yards = 25136(5.26 avg)

ALL DOUBLE COVERAGE
Drop Backs = 84257
Attempts = 76446 (90.7%)
Completions = 41235 (53.9%)
Thrown Yards = 454423 (5.94 YPA, 11.02 YPC)
Yards After Completion = 132082 (1.73 YPA, 3.20 YPC)
Total Yards = 586505 (7.67 YPA, 14.22 YPC)
Sacks = 4148 (4.9%), yards lost = 26059(6.28 avg)
Scrambles = 3663 (4.3%), yards = 19630(5.36 avg)

INTO DOUBLE COVERAGE
Drop Backs = 26024
Attempts = 23513 (90.4%)
Completions = 9446 (40.2%)
Thrown Yards = 114358 (4.86 YPA, 12.11 YPC)
Yards After Completion = 32527 (1.38 YPA, 3.44 YPC)
Total Yards = 146885 (6.25 YPA, 15.55 YPC)
Sacks = 1380 (5.3%), yards lost = 8652(6.27 avg)
Scrambles = 1131 (4.3%), yards = 6065(5.36 avg)

AVOID DOUBLE COVERAGE
Drop Backs = 58233
Attempts = 52933 (90.9%)
Completions = 31789 (60.1%)
Thrown Yards = 340065 (6.42 YPA, 10.70 YPC)
Yards After Completion = 99555 (1.88 YPA, 3.13 YPC)
Total Yards = 439620 (8.31 YPA, 13.83 YPC)
Sacks = 2768 (4.8%), yards lost = 17407(6.29 avg)
Scrambles = 2532 (4.3%), yards = 13565(5.36 avg)

COMPLETIONS
Total = 57.6%
Normal = 60.4%
All DC = 53.9%
Into DC = 40.2%
Avoid DC = 60.1%

YARDS
Total = 8.24 YPA, 14.31 YPC
Normal = 8.68 YPA, 14.37 YPC
All DC = 7.67 YPA, 14.22 YPC
Into DC = 6.25 YPA, 15.55 YPC
Avoid DC = 8.31 YPA, 13.83 YPC

So yards do go down when a QB avoids double coverage, but they go up if he manages to complete one into double coverage. But overall completion percentage goes down, yards per attempt goes down, and yards per completion goes down. I may try an "always double cover" gameplan in WOOF since everything is better except the occasional longer completion.

And yeah, the final GamePlanAnalyzer utility will have all this in a nice neat table, I'm just not that far along yet.

[Yoda]

That is how I have most of my current defensive gameplans set up, to DC one side always at 100%. Depending on my CBs. Looking at this, I might to it regardless.

[Ben E Lou]

Thrown Yards = 1110547 (6.35 YPA, 11.04 YPC)
Yards After Completion = 329827 (1.89 YPA, 3.28 YPC)
Total Yards = 1440374 (8.24 YPA, 14.31 YPC)

This can't be right.

[Ben E Lou]

Thrown Yards = 1110547 (6.35 YPA, 11.04 YPC)
This looks like it would be correct for overall.

[isaccoubaldi]

But if I set in my defensive gameplan the "Double-Team Top WR" and "Double-Team Second WR" options with 50-50 or 100-100 is the same or change something?

[Steel]

But if I set in my defensive gameplan the "Double-Team Top WR" and "Double-Team Second WR" options with 50-50 or 100-100 is the same or change something?

The game won't allow you to manually take away both outside WRs by doubling them. So, basically, the 2nd WR is being doubled a % of the time when the 1st WR isn't.

[isaccoubaldi]

The game won't allow you to manually take away both outside WRs by doubling them. So, basically, the 2nd WR is being doubled a % of the time when the 1st WR isn't.

So if I understand well a good solution when the opponent have 2 good WR is 50-100. In this way 50% I'll cover the WR1 and when the WR1 is not covered I'll cover 100% WR2. Right?

[Steel]

That is correct.

[gstelmack]

This looks like it would be correct for overall.

Okay, I'll check and recalc.

[gstelmack]

Post above fixed.

[Ben E Lou]

Post above fixed.You sure about that? I don't see an edit...

[gstelmack]

We really need to update the board software to something that lets us post tables and doesn't drop edits when completely changing a post:

TOTAL PASSING
Drop Backs = 193259
Attempts = 174792 (90.4%)
Completions = 100635 (57.6%)
Thrown Yards = 780720 (4.47 YPA, 7.76 YPC)
Yards After Completion = 329827 (1.89 YPA, 3.28 YPC)
Total Yards = 1110547 (6.35 YPA, 11.04 YPC)
Sacks = 10024 (5.2%), yards lost = 63119 (6.30 avg)
Scrambles = 8443 (4.4%), yards = 44766 (5.30 avg)

NORMAL PASSING
Drop Backs = 109002
Attempts = 98346 (90.2%)
Completions = 59400 (60.4%)
Thrown Yards = 458379 (4.66 YPA, 7.72 YPC)
Yards After Completion = 197745 (2.01 YPA, 3.33 YPC)
Total Yards = 656124 (6.67 YPA, 11.05 YPC)
Sacks = 5876 (5.4%), yards lost = 37060 (6.31 avg)
Scrambles = 4780 (4.4%), yards = 25136 (5.26 avg)

ALL DOUBLE COVERAGE
Drop Backs = 84257
Attempts = 76446 (90.7%)
Completions = 41235 (53.9%)
Thrown Yards = 322341 (4.22 YPA, 7.82 YPC)
Yards After Completion = 132082 (1.73 YPA, 3.20 YPC)
Total Yards = 454423 (5.94 YPA, 11.02 YPC)
Sacks = 4148 (4.9%), yards lost = 26059 (6.28 avg)
Scrambles = 3663 (4.3%), yards = 19630 (5.36 avg)

INTO DOUBLE COVERAGE
Drop Backs = 26024
Attempts = 23513 (90.4%)
Completions = 9446 (40.2%)
Thrown Yards = 81831 (3.48 YPA, 8.66 YPC)
Yards After Completion = 32527 (1.38 YPA, 3.44 YPC)
Total Yards = 114358 (4.86 YPA, 12.11 YPC)
Sacks = 1380 (5.3%), yards lost = 8652 (6.27 avg)
Scrambles = 1131 (4.3%), yards = 6065 (5.36 avg)

AVOID DOUBLE COVERAGE
Drop Backs = 58233
Attempts = 52933 (90.9%)
Completions = 31789 (60.1%)
Thrown Yards = 240510 (4.54 YPA, 7.57 YPC)
Yards After Completion = 99555 (1.88 YPA, 3.13 YPC)
Total Yards = 340065 (6.42 YPA, 10.70 YPC)
Sacks = 2768 (4.8%), yards lost = 17407 (6.29 avg)
Scrambles = 2532 (4.3%), yards = 13565 (5.36 avg)

And the key stats:

TOTAL PASSING
Completions = 100635 (57.6%)
Total Yards = 1110547 (6.35 YPA, 11.04 YPC)

NORMAL PASSING
Completions = 59400 (60.4%)
Total Yards = 656124 (6.67 YPA, 11.05 YPC)

ALL DOUBLE COVERAGE
Completions = 41235 (53.9%)
Total Yards = 454423 (5.94 YPA, 11.02 YPC)

INTO DOUBLE COVERAGE
Completions = 9446 (40.2%)
Total Yards = 114358 (4.86 YPA, 12.11 YPC)

AVOID DOUBLE COVERAGE
Completions = 31789 (60.1%)
Total Yards = 340065 (6.42 YPA, 10.70 YPC)

[Ben E Lou]

Those numbers are colored by the fact that doubling only exists in pass and pass aggressive defense. Ultimately, the key stats would be comparisons of pass aggressive total, pass aggressive normal, pass aggressive all double, pass aggressive into double, pass aggressive avoid double, and the same for pass. The "Normal Passing" numbers include passing against run and run aggressive, while the all/into/avoid numbers do not.

[Yoda]

That's a very good point.

[gstelmack]

Those numbers are colored by the fact that doubling only exists in pass and pass aggressive defense. Ultimately, the key stats would be comparisons of pass aggressive total, pass aggressive normal, pass aggressive all double, pass aggressive into double, pass aggressive avoid double, and the same for pass. The "Normal Passing" numbers include passing against run and run aggressive, while the all/into/avoid numbers do not.

Okay, that breakdown is planned for the full utility, so we'll have to wait until then. With any luck not too horribly long.

[gstelmack]

Note that this is only one season, and I will be adding plenty of rows soon. What I really want to know is:

A) Do I have all the columns we want here, or are there any others you'd like to see?
B) Any graphic designers have any suggestions on improving the readability? I might do different colors for the sum columns, percent columns, and average columns, but I'm open to more expert suggestions here.

<table border="1" summary="This table shows some passing statistics"><caption>Passing Stats</caption><tbody><tr><th><th>DrpBk<th>Att<th>Pct<th>Cmp<th>Pct<th>Yds<th>YAC<th>YPA<th>YPC<th>Sack<th>Pct<th>Yds<th>Avg<th>Scrmb<th>Pct<th>Yds<th>Avg<th>INT<th>Pct<th>TD<th>Pct <tr><td class="highlight">All<td class="normal" align="right">23,898<td class="normal" align="right">21,501<td class="normal" align="right">90.0%<td class="normal" align="right">12,304<td class="normal" align="right">57.2%<td class="normal" align="right">138,622<td class="normal" align="right">40,586<td class="normal" align="right">6.45<td class="normal" align="right">11.27<td class="normal" align="right">1,262<td class="normal" align="right">5.3%<td class="normal" align="right">7,984<td class="normal" align="right">6.33<td class="normal" align="right">1,135<td class="normal" align="right">4.7%<td class="normal" align="right">5,931<td class="normal" align="right">5.23<td class="normal" align="right">589<td class="normal" align="right">2.7%<td class="normal" align="right">731<td class="normal" align="right">3.4% <tr><th colSpan="22">Pass Versus Defensive Play Call <tr><td class="highlight">RunAggressive<td class="normal" align="right">1,229<td class="normal" align="right">1,115<td class="normal" align="right">90.7%<td class="normal" align="right">747<td class="normal" align="right">67.0%<td class="normal" align="right">7,557<td class="normal" align="right">2,215<td class="normal" align="right">6.78<td class="normal" align="right">10.12<td class="normal" align="right">54<td class="normal" align="right">4.4%<td class="normal" align="right">324<td class="normal" align="right">6.00<td class="normal" align="right">60<td class="normal" align="right">4.9%<td class="normal" align="right">264<td class="normal" align="right">4.40<td class="normal" align="right">18<td class="normal" align="right">1.6%<td class="normal" align="right">84<td class="normal" align="right">7.5% <tr><td class="highlight">RunNormal<td class="normal" align="right">5,994<td class="normal" align="right">5,398<td class="normal" align="right">90.1%<td class="normal" align="right">3,445<td class="normal" align="right">63.8%<td class="normal" align="right">40,050<td class="normal" align="right">13,110<td class="normal" align="right">7.42<td class="normal" align="right">11.63<td class="normal" align="right">312<td class="normal" align="right">5.2%<td class="normal" align="right">2,028<td class="normal" align="right">6.50<td class="normal" align="right">284<td class="normal" align="right">4.7%<td class="normal" align="right">1,517<td class="normal" align="right">5.34<td class="normal" align="right">126<td class="normal" align="right">2.3%<td class="normal" align="right">245<td class="normal" align="right">4.5% <tr><td class="highlight">PassNormal<td class="normal" align="right">8,455<td class="normal" align="right">7,607<td class="normal" align="right">90.0%<td class="normal" align="right">4,404<td class="normal" align="right">57.9%<td class="normal" align="right">47,976<td class="normal" align="right">14,247<td class="normal" align="right">6.31<td class="normal" align="right">10.89<td class="normal" align="right">451<td class="normal" align="right">5.3%<td class="normal" align="right">2,833<td class="normal" align="right">6.28<td class="normal" align="right">397<td class="normal" align="right">4.7%<td class="normal" align="right">2,052<td class="normal" align="right">5.17<td class="normal" align="right">219<td class="normal" align="right">2.9%<td class="normal" align="right">252<td class="normal" align="right">3.3% <tr><td class="highlight">PassAggressive<td class="normal" align="right">8,220<td class="normal" align="right">7,381<td class="normal" align="right">89.8%<td class="normal" align="right">3,708<td class="normal" align="right">50.2%<td class="normal" align="right">43,039<td class="normal" align="right">11,014<td class="normal" align="right">5.83<td class="normal" align="right">11.61<td class="normal" align="right">445<td class="normal" align="right">5.4%<td class="normal" align="right">2,799<td class="normal" align="right">6.29<td class="normal" align="right">394<td class="normal" align="right">4.8%<td class="normal" align="right">2,098<td class="normal" align="right">5.32<td class="normal" align="right">226<td class="normal" align="right">3.1%<td class="normal" align="right">150<td class="normal" align="right">2.0% <tr><th colSpan="22">Pass Versus Double Coverage <tr><td class="highlight">Normal<td class="normal" align="right">14,233<td class="normal" align="right">12,802<td class="normal" align="right">89.9%<td class="normal" align="right">7,653<td class="normal" align="right">59.8%<td class="normal" align="right">86,139<td class="normal" align="right">25,607<td class="normal" align="right">6.73<td class="normal" align="right">11.26<td class="normal" align="right">745<td class="normal" align="right">5.2%<td class="normal" align="right">4,733<td class="normal" align="right">6.35<td class="normal" align="right">686<td class="normal" align="right">4.8%<td class="normal" align="right">3,601<td class="normal" align="right">5.25<td class="normal" align="right">341<td class="normal" align="right">2.7%<td class="normal" align="right">495<td class="normal" align="right">3.9% <tr><td class="highlight">Doubled<td class="normal" align="right">9,665<td class="normal" align="right">8,699<td class="normal" align="right">90.0%<td class="normal" align="right">4,651<td class="normal" align="right">53.5%<td class="normal" align="right">52,483<td class="normal" align="right">14,979<td class="normal" align="right">6.03<td class="normal" align="right">11.28<td class="normal" align="right">517<td class="normal" align="right">5.3%<td class="normal" align="right">3,251<td class="normal" align="right">6.29<td class="normal" align="right">449<td class="normal" align="right">4.6%<td class="normal" align="right">2,330<td class="normal" align="right">5.19<td class="normal" align="right">248<td class="normal" align="right">2.9%<td class="normal" align="right">236<td class="normal" align="right">2.7% <tr><td class="highlight">Into<td class="normal" align="right">2,974<td class="normal" align="right">2,663<td class="normal" align="right">89.5%<td class="normal" align="right">1,080<td class="normal" align="right">40.6%<td class="normal" align="right">13,302<td class="normal" align="right">3,636<td class="normal" align="right">5.00<td class="normal" align="right">12.32<td class="normal" align="right">188<td class="normal" align="right">6.3%<td class="normal" align="right">1,172<td class="normal" align="right">6.23<td class="normal" align="right">123<td class="normal" align="right">4.1%<td class="normal" align="right">623<td class="normal" align="right">5.07<td class="normal" align="right">90<td class="normal" align="right">3.4%<td class="normal" align="right">60<td class="normal" align="right">2.3% <tr><td class="highlight">Avoid<td class="normal" align="right">6,691<td class="normal" align="right">6,036<td class="normal" align="right">90.2%<td class="normal" align="right">3,571<td class="normal" align="right">59.2%<td class="normal" align="right">39,181<td class="normal" align="right">11,343<td class="normal" align="right">6.49<td class="normal" align="right">10.97<td class="normal" align="right">329<td class="normal" align="right">4.9%<td class="normal" align="right">2,079<td class="normal" align="right">6.32<td class="normal" align="right">326<td class="normal" align="right">4.9%<td class="normal" align="right">1,707<td class="normal" align="right">5.24<td class="normal" align="right">158<td class="normal" align="right">2.6%<td class="normal" align="right">176<td class="normal" align="right">2.9% <tr><th colSpan="22">Pass Versus Double Coverage and Defensive Playcall <tr><td class="highlight">Avoid & PassNormal<td class="normal" align="right">3,368<td class="normal" align="right">3,049<td class="normal" align="right">90.5%<td class="normal" align="right">1,937<td class="normal" align="right">63.5%<td class="normal" align="right">20,217<td class="normal" align="right">6,368<td class="normal" align="right">6.63<td class="normal" align="right">10.44<td class="normal" align="right">169<td class="normal" align="right">5.0%<td class="normal" align="right">1,070<td class="normal" align="right">6.33<td class="normal" align="right">150<td class="normal" align="right">4.5%<td class="normal" align="right">829<td class="normal" align="right">5.53<td class="normal" align="right">72<td class="normal" align="right">2.4%<td class="normal" align="right">112<td class="normal" align="right">3.7% <tr><td class="highlight">Avoid & PassAggressive<td class="normal" align="right">3,323<td class="normal" align="right">2,987<td class="normal" align="right">89.9%<td class="normal" align="right">1,634<td class="normal" align="right">54.7%<td class="normal" align="right">18,964<td class="normal" align="right">4,975<td class="normal" align="right">6.35<td class="normal" align="right">11.61<td class="normal" align="right">160<td class="normal" align="right">4.8%<td class="normal" align="right">1,009<td class="normal" align="right">6.31<td class="normal" align="right">176<td class="normal" align="right">5.3%<td class="normal" align="right">878<td class="normal" align="right">4.99<td class="normal" align="right">86<td class="normal" align="right">2.9%<td class="normal" align="right">64<td class="normal" align="right">2.1% <tr><td class="highlight">Into & PassNormal<td class="normal" align="right">1,483<td class="normal" align="right">1,325<td class="normal" align="right">89.3%<td class="normal" align="right">586<td class="normal" align="right">44.2%<td class="normal" align="right">7,061<td class="normal" align="right">1,999<td class="normal" align="right">5.33<td class="normal" align="right">12.05<td class="normal" align="right">93<td class="normal" align="right">6.3%<td class="normal" align="right">584<td class="normal" align="right">6.28<td class="normal" align="right">65<td class="normal" align="right">4.4%<td class="normal" align="right">296<td class="normal" align="right">4.55<td class="normal" align="right">52<td class="normal" align="right">3.9%<td class="normal" align="right">39<td class="normal" align="right">2.9% <tr><td class="highlight">Into & PassAggressive<td class="normal" align="right">1,491<td class="normal" align="right">1,338<td class="normal" align="right">89.7%<td class="normal" align="right">494<td class="normal" align="right">36.9%<td class="normal" align="right">6,241<td class="normal" align="right">1,637<td class="normal" align="right">4.66<td class="normal" align="right">12.63<td class="normal" align="right">95<td class="normal" align="right">6.4%<td class="normal" align="right">588<td class="normal" align="right">6.19<td class="normal" align="right">58<td class="normal" align="right">3.9%<td class="normal" align="right">327<td class="normal" align="right">5.64<td class="normal" align="right">38<td class="normal" align="right">2.8%<td class="normal" align="right">21<td class="normal" align="right">1.6% </td></tr></tbody></table>

[gstelmack]

Gah, never mind format suggestions, the forum stipped all kinds of stuff. The header rows should be bold, the left-most column should be yellow background. I hate pasting tables here. Plus I'm grumpy. I'll post a web link soon with a real table with more data that we can discuss formatting on, or feel free to copy that off some place and play with it until it looks cool, then send me an HTML sample of it.

[Ben E Lou]

I think the columns are good to go.

[thenewchuckd]

This is interesting stuff and continues to show that double coverage can indeed be effective.

I'm just trying to figure out how it answers the question about a QB's read defense skill?

[gstelmack]

This is interesting stuff and continues to show that double coverage can indeed be effective.

I'm just trying to figure out how it answers the question about a QB's read defense skill?

It doesn't, but once the utility is out the tests can be rerun and pull far more data out.

[Ben E Lou]

It doesn't, but once the utility is out the tests can be rerun and pull far more data out.

In anticipation of that, I've set up the "neutral" league I've mentioned to you. All defenses 25-25-25-25 in all situations, all formations equally used in all situations, etc. I'm through 14 seasons of all 6.4 data in it so far, and if my daughter stays in my lap watching Sesame Street on the other half of the screen much longer, I may get through 25 or so before the evening is over. ;)

[Jughead Spock]

I'm really starting to believe that RD has a significant amount to do with red zone effectiveness. Just anecdotal, though.

[gstelmack]

In anticipation of that, I've set up the "neutral" league I've mentioned to you. All defenses 25-25-25-25 in all situations, all formations equally used in all situations, etc. I'm through 14 seasons of all 6.4 data in it so far, and if my daughter stays in my lap watching Sesame Street on the other half of the screen much longer, I may get through 25 or so before the evening is over. ;)

Hoping to release tonight. I haven't done defensive playcalling tendencies / results yet (strictly the "what did the defense call by down and distance and how successful was the offense against those calls" bit may be missing, lots of other defensive stuff there), but I'm also not quite sure how I want to handle them internally, so I may push everything else out and play with that later.

[Ben E Lou]

I'd say push it out. I don't know what can be learned by seeing others' defensive playcalling tendencies. If there is something there, then we can just catch it later.

(Not that my opinion is colored by the fact that I have time the next couple of nights and some during the day to get some info out to the community about it or anything.... ;))

[Ben E Lou]

(Oh, and I got 40 seasons done. Plenty of data.)

[Ben E Lou]

Those numbers are colored by the fact that doubling only exists in pass and pass aggressive defense. Ultimately, the key stats would be comparisons of pass aggressive total, pass aggressive normal, pass aggressive all double, pass aggressive into double, pass aggressive avoid double, and the same for pass. The "Normal Passing" numbers include passing against run and run aggressive, while the all/into/avoid numbers do not.

That's a very good point.Heh. Ok. I *KNEW* we had discussed this before. I'm guessing you just didn't quite get what I was saying here.

[Yoda]

Wow... completely forgot about this... the sins of being in too many leagues.

That or growing old....

Something I definitely should have remembered.

[scorp]

I was under the impression Read defense also had an effect on completion percentage and interceptions. If a QB does not read defense well and throws into double coverage more he should complete less passes and get picked more often.

This Guy I had for a season in RNFL seems to back up the idea:

http://i96.photobucket.com/albums/l178/scorpionking271/ellisbars.png

His career stats:

http://i96.photobucket.com/albums/l178/scorpionking271/ellisstats.png


granted this guy is an extreme his RD=0 but his bars are pretty good other wise. He only has 2 seasons with more TDs than INTs and one of those is 9 vs 7.

I cut him after 1 season figuring the Read Defense was killing him, but this thread indicates that shouldn't be the case. Am I missing something about him???

[Firefly]

Going by his numbers, he does suck. Why, I've no idea.

[thenewchuckd]

I am not sure you can prove much with this QB. He is pretty average across the board. But...


I cut him after 1 season figuring the Read Defense was killing him, but this thread indicates that shouldn't be the case.

I definitely never said this. In fact, I would actually say the opposite. I am still of the opinion that read defense is one of the more important skills for a QB to have. It will not necessarily limit the number of attempts that a QB throws into double coverage (very much, anyways).

But it certainly makes a QB more effective. Some semi-smart people, not quite understanding the implications of my study above, come to the conclusion that a read defense QB will throw more effectively into double coverage. I would go even further and say that a read defense QB will just be more effective - period.

[scorp]

K thanks Chuck, one thing I haven't seen yet is if read defense has anything to do with the QB and his read progression. does a low read defense guy lock onto WR1.
Looking at plays there is a primary WR for pass plays maybe read defense has an effect on how quickly or well he moves on to the next guy since option 1 is covered.

[johnnyshaka]

I was under the impression Read defense also had an effect on completion percentage and interceptions. If a QB does not read defense well and throws into double coverage more he should complete less passes and get picked more often.

This Guy I had for a season in RNFL seems to back up the idea:

http://i96.photobucket.com/albums/l178/scorpionking271/ellisbars.png

His career stats:

http://i96.photobucket.com/albums/l178/scorpionking271/ellisstats.png


granted this guy is an extreme his RD=0 but his bars are pretty good other wise. He only has 2 seasons with more TDs than INTs and one of those is 9 vs 7.

I cut him after 1 season figuring the Read Defense was killing him, but this thread indicates that shouldn't be the case. Am I missing something about him???

What was this guy's intelligence and SOL score at the combine? His hidden Avoid INT bar could be pretty low and contributing to the number of picks.

[scorp]

OK I chopped it off when I cropped him Intelligence is 37 I do not know his SOL, He was drafted long before I joined the league, only have the stat files for those years.
When I signed him I was going by a 37 Int being fairly safe even is his sol wasn't greatest.

[johnnyshaka]

My guess is if he was drafted in the 6th round that his SOL was pretty low making his Avoid INT bar low as well. We'll never know.

Regardless, the guy looks like an emergency starter/backup type of guy but if he were plugged into an offense with two great receivers he'd be serviceable. But it looks like he's been stuck in either an ultra-conservative short passing offense with garbage for receivers or he's reading his progressions backwards...looking short (and taking that option almost every time) to long...which could be a function of the read defense bar, I suppose.