QB Hits - Spencer and Ware

[percyhoward]

Dash28, great, great sig pic

That may have been the defensive play of that game--and by extension, of the season.

 

[Manster54]

How many Pressures does Ellis have?
:laugh2:

 

[theogt]

Nice work, again.

(Are number of opponents' pass plays figured into this?)

This is the kind of info the pro's should be giving us, obviously. We could take these ratings, then and look at splits, such as down-and-distance, by what packages were being used, in close games/blowouts, field position, etc.

We'd know a lot more about who's doing what on the field, and there would be less speculation. It would basically be a clearer picture. "High-Def Stats."

Looking at a per pass play basis would be ideal, but entirely too difficult. I could look at it on a per snap basis, however, since PFF includes players' snaps in their stat sheet.

That's interesting work you did with the variances but I just question the weighting of sacks versus hits which are pretty close to equally valued in this system.

Personally I would rank them sacks more important than pressures (since they definitely alter the play) which are more important than hits (in general).

I've played with weighting sacks more important and the rankings change very little.

I do think you get a more valid and reproducible quantum out of this measurement than you do out of the simple sacks number. I wonder if this was run over the last several years if we saw this QB pressuring stat you have created is more stable (less variability) than sacks. One thing that you could do if you haven't already is normalize everything per game played if you so desired so that it was more about rates than seasonal statistics - this would likely make the construct more stable as well.

See above re: per snap basis. That gets even more clear than per game basis.

 

[percyhoward]

That's interesting work you did with the variances but I just question the weighting of sacks versus hits which are pretty close to equally valued in this system.

Personally I would rank them sacks more important than pressures (since they definitely alter the play) which are more important than hits (in general).

That's the 1-2-3 order I was talking about. If theo weighed all 3 equally, then of course that's a flaw. But I say "if," because I don't understand variance or how it applies, and I'd be dumb enough to give 3 points for a sack, 2 for a pressure, 1 for a hit, then add'em all up and divide by number of opponents' pass plays.

 

[percyhoward]

I've played with weighting sacks more important and the rankings change very little.

Gotcha. But if you want to find # of opponents' pass plays, all you have to do is just add attempts plus sacks, don't you?

 

[theogt]

That's the 1-2-3 order I was talking about. If theo weighed all 3 equally, then of course that's a flaw. But I say "if," because I don't understand variance or how it applies, and I'd be dumb enough to give 3 points for a sack, 2 for a pressure, 1 for a hit, then add'em all up and divide by number of opponents' pass plays.

They're not really weight equally or unequally. They're weighted such that the variance is the same for each.

But here are the correlation coefficients for each statistic in relation to the overall formula:

Sacks - 0.913766765
Pressures - 0.912298953
Hits - 0.746856308

In other words, sacks correlate higher with the final result than pressures. And pressures more than hits.

The correlation between sacks and pressures is very high (0.769650447). Thought that was pretty interesting.

 

[percyhoward]

They're not really weight equally or unequally. They're weighted such that the variance is the same for each.

Like Eskimo said, as long as sacks count for more than pressures, and pressure for more than hits. And as far as per-pass-play vs. per-snap, I don't think there would be any worry about which pass plays the particular player was in on, and which ones not.

Since I'm too lazy to do it myself at the moment, I guess it doesn't matter, though.

Suffice to say we've got the best pair of pass rushers in football.

 

[Eskimo]

One way to consider weighting them is how much they change the offenses ability to generate points has changed after that play. This would be a very difficult sort of statistic to generate and validate but I do think this would be the best way to go about it in general.

Baseball is so far ahead of football when it comes to this. I do think a large part of this is that offensive stats are easy to keep track of and they had lots of data going back many years that could be mined. I do agree that someone needs to go about trying to quantify these things better. Personally, I think the league should hire someone for every team to map out every snap of the year using the all the video cameras available to generate more relevant and precise set of data. You could even have that person conferring with a member of the team's coaching staff to help understand player's assignments on a certain play.

Once a more rich form of stats were available to the general public we would develop a football version of sabermetrics like that which exists in baseball. It would definitely add to the complexity with which the fans could analyze the game. Having the fans think about football more is good for the game so I think it is in the league's best interest in the end to do this.

Theo, your idea about doing it per snap is fine but as mentioned it should ideally be done per pass play. PFF does mention snaps, but I don't think they break it down into pass versus run plays for each individual player. I do think going per game would be an okay starting point until a richer set of data is available.

 

[theogt]

Like Eskimo said, as long as sacks count for more than pressures, and pressure for more than hits. And as far as per-pass-play vs. per-snap, I don't think there would be any worry about which pass plays the particular player was in on, and which ones not.

Since I'm too lazy to do it myself at the moment, I guess it doesn't matter, though.

Suffice to say we've got the best pair of pass rushers in football.

Combine that with one of the greatest defensive minds ever in terms of getting pressure, and it's pretty nice.

Here's on a per snap basis.

Name		Snaps	QB Sk	QB Ht	QB Pr	Tks	Pres.  Form.
DeMarcus Ware	927	12	16	54	34	125.13
T. Banta-Cain	444	7	7	15	27	104.73
Elvis Dumervil	829	17	7	29	22	103.14
Anthony Spencer	947	5	26	21	54	90.29
James Harrison	899	10	12	31	47	88.99
Aaron Kampman	519	4	12	11	21	86.71
Lamarr Woodley	897	10	11	27	33	82.50
Clay Matthews	813	10	9	24	28	82.41
Brad Jones	322	4	3	9	15	77.64
Matt Roth	240	3	2	7	11	77.08
Manny Lawson	781	6	9	19	36	66.58
Robert Ayers	411	0	5	17	17	65.69
Clark Haggans	771	6	7	21	44	64.85
Parys Haralson	857	5	7	26	25	61.26
Jason Taylor	777	8	4	19	23	60.49
Bertrand Berry	463	4	0	18	2	60.48
Joey Porter	690	8	5	11	23	59.42
K. Wimbley	885	7	7	21	44	59.32
Shaun Phillips	836	7	8	16	43	59.21
Larry English	482	2	6	10	15	56.02
Shawne Merriman	614	4	4	13	24	50.49
David Bowens	416	1	6	2	18	39.66
Chike Okeafor	623	3	2	13	30	39.33
Jason Trusnik	267	2	2	1	17	37.45
Derrick Harvey	337	0	3	6	18	35.61
Clint Ingram	304	1	2	4	21	34.54
Mike Vrabel	779	2	2	13	38	28.24
Mario Haggan	564	1	3	7	36	27.48
Brady Poppinga	248	1	0	2	7	18.15

 

[theogt]

Theo, your idea about doing it per snap is fine but as mentioned it should ideally be done per pass play. PFF does mention snaps, but I don't think they break it down into pass versus run plays for each individual player. I do think going per game would be an okay starting point until a richer set of data is available.

You're more than welcome to add up the pass plays of each. I'll be sure to feed it into an excel sheet.

The great thing about PFF's website is you can copy and paste into Excel. Beyond copying and pasting, my curiosity is outweighed by my laziness.

 

[theogt]

The per play pressure statistic, combining 3-4 OLBs and 4-3 DEs.

Name		Snaps	QB Sk	QB Ht	QB Pr	Tks	Pres.  Form.
DeMarcus Ware	927	12	16	54	34	125.13
T. Banta-Cain	444	7	7	15	27	104.73
Elvis Dumervil	829	17	7	29	22	103.14
Anthony Spencer	947	5	26	21	54	90.29
James Harrison	899	10	12	31	47	88.99
Aaron Kampman	519	4	12	11	21	86.71
Lamarr Woodley	897	10	11	27	33	82.50
Clay Matthews	813	10	9	24	28	82.41
Brad Jones	322	4	3	9	15	77.64
Matt Roth	240	3	2	7	11	77.08
Tamba Hali	954	8	11	30	34	75.47
Manny Lawson	781	6	9	19	36	66.58
Robert Ayers	411	0	5	17	17	65.69
Clark Haggans	771	6	7	21	44	64.85
Parys Haralson	857	5	7	26	25	61.26
Jason Taylor	777	8	4	19	23	60.49
Bertrand Berry	463	4	0	18	2	60.48
Joey Porter	690	8	5	11	23	59.42
K. Wimbley	885	7	7	21	44	59.32
Shaun Phillips	836	7	8	16	43	59.21

 

[Eskimo]

You're more than welcome to add up the pass plays of each. I'll be sure to feed it into an excel sheet.

The great thing about PFF's website is you can copy and paste into Excel. Beyond copying and pasting, my curiosity is outweighed by my laziness.

The problem that I have with going only per pass is that you gve a huge advantage to the nickel pass rushers over the everydown player. Both may play a fairly similar amount of total pass plays but the everydown player will have played twice as many snaps. I still think going by per game played is a fairer starting point then per snap presently.

 

[theogt]

Oddly enough, according to this formula, Ware had a better year this year than last.

 

[RS12]

Oddly enough, according to this formula, Ware had a better year this year than last.

That is odd, D Ware seemed like the second coming of Deacon Jones last year with 20 sacks. Doesn't look as dominant this year with the injuries.

 

[Eskimo]

Oddly enough, according to this formula, Ware had a better year this year than last.

Is the difference due to the ridiculous number of QB pressures he has had this year? He continues to be the best defensive player in the league for the 3rd straight year.

I always envied some of the older guys telling stories of a couple of all-time greats like Bob Lilly and Roger Staubach when I had never seen any of them play at all. Some day I can tell my kids about Demarcus Ware and what a force he was (and fine human being to boot as far as anyone can tell). At some point I imagine the league will make all the season's videos available on DVD.

 

[theogt]

That is odd, D Ware seemed like the second coming of Deacon Jones last year with 20 sacks. Doesn't look as dominant this year with the injuries.

It's the 54 pressures.

It's an absurdly high number. For comparison, Ware and James Harrison did not have 54 pressure COMBINED in 2008.

Sort of makes you question its accuracy.

 

[Eskimo]

It's the 54 pressures.

It's an absurdly high number. For comparison, Ware and James Harrison did not have 54 pressure COMBINED in 2008.

Sort of makes you question its accuracy.

Has anyone checked the log for some sort of a typo?

I checked and it is not due to an obvious single typo or improper addition. The amazing thing is how dominant he has been down the stretch here. He has 35 pressures in the last 6 weeks including a monster performance 9 against Oakland and 6 against the Giants. He contributed 5 this last week against Washington. His worst performance in this stretch was 3 pressures which if extrapolated over the whole year would give him 48 pressures which would still put him far ahead of everyone else.

It'd be interesting if someone with a PVR and lots of extra free time wants to try and score QB pressures next week or from the tape of this last week against Washington.

It beats me if the methodology or the scorer changed in the middle of the season.

 

[JBell]

Interesting that Orakpo doesn't appear on either list.

He used 3-4 OLB's and 4-3 DE's.

Orakpo is a 4-3 OLB.

 

[Eskimo]

He used 3-4 OLB's and 4-3 DE's.

Orakpo is a 4-3 OLB.

I think Washington is one of those hybrid teams that is sometimes a 4-3 and sometimes a 3-4. I don't watch any of their games except against us but I think that's the way they do it.

 

[goliadmike]

I think Washington is one of those hybrid teams that is sometimes a 4-3 and sometimes a 3-4. I don't watch any of their games except against us but I think that's the way they do it.

New Orleans uses some 3-4 alignments as well.