A Defense Stats Question - and statistical whining at then end!

Hey remember me? I've had my head in the dissertation sand, so that I can be doing the less weekend-time-consuming data collection during the football season. I hope I'm not asking something that is obvious to anyone who doesn't get their ATQ fix by asking Nico what's going on on ATQ

My question: Is there a site that lists Points per minutes the D is on the field, or perhaps Points per plays for the D? I was wondering after reading the Aliotti interview.

Also when will people start listing the Standard Deviation for the averages we see? Would that be so hard? It would be ridiculously informative.

over 2 years ago bradLL99 11 comments 0 recs |

Story-email Email Printer Print

Comments

I'm an idiot

Somewhere my inner Jared reminded me to check cfbstats,com, and sure enough they had Points/Play.

So, I was trying to think what was a relatively easy way to evaluate Dfenses across teams. Based on the Aliotti interview I realized that obviously, points/play will still be biased to D’s that have to face a lot of plays because they will have to platoon more. Now the best measure would take into account the offenses a team faced, but I’m doing this as an exercise of procrastination not work

So I thought I would use plays to predict points (# of plays faced predicts 25% of the variance btw, not to shabby for a simple linear regression) and rank teams by their residuals. This number is the difference between how many points they allowed and how many points that the regression predicted should be allowed. (Predicted points = 140.41 + 5.17*#plays).

So a good D should have a positive score (meaning they gave up less points than the average team would have given their number of plays). A bad D would have a negative score, (meaning they gave up more points than the average team would have given their number of plays). The mean is, of course, 0 and the SD is 562.67

So for example the top 5 last season were:
USC 1384.31
TCU 1131.19
Tennessee 991.54
Boston College 983.93
Boise State 918.10

And the bottom 5 were:
Washington State -1046.76
UTEP -1080.078
Iowa State -1091.80
Idaho -1103.42
North Texas -1456.46

Oregon was 27th with 417.63. Right in front of Florida State and Right behind ASU.

Like I said before, this ranking is still not taking into account strength of schedule or anything, but I still find it kind of interesting

by bradLL99 on Aug 6, 2009 9:30 PM PDT reply actions

Oh

These signs are reversed, and these are season stats.

by bradLL99 on Aug 6, 2009 10:05 PM PDT up reply actions

First, I think there is a pretty good system in place. Jared and I both preach this stuff.

I can’t find my spreadsheet, but Jared has one for every team, and it’s all adjusted for strength of schedule.

S&P+ = the greatest metric in football.

Second, Welcome back!

"It’s spelled S-H-U-F-E-L-T-M-E-U-P"

by JShufelt on Aug 7, 2009 7:17 AM PDT up reply actions

Very cool!

Far better that what I can do.

Does S&P work for D? I didn’t quite get that. Also I don’t like metrics that use rating systems. I hate that people tweak numbers so that their results fit their expectations. I know it works, but the methodologist in me does not care for it.

The first thing I had to do with Success Rate was adjust it for the college game. Yards are a bit easier to come by in college, so I made the following adjustments:

PPP is a good measure but not perfect. I assumes that playing D is equally difficult for 1st play as it is for the last 50th play (given the situations are equal) but Ds that are on the field longer usually have to use more depth which is a major disadvantage for most college teams (save the reloading teams like USC, LSU, and Florida).

Rather than looking at PPP you need to look at points while statistically controlling for plays. I think residuals is the way to go for that, but I really have thought the stats theory through.

by bradLL99 on Aug 7, 2009 9:21 AM PDT up reply actions

I’m actually drawing a bit of a blank with how PPP is calculated for defense…

Success Rate, I think is pretty dead on.

When I get time, I’ll refresh my memory. If you get real board, you can go through a lot of his stuff on Football Outsiders. He writes all the stuff on the Varsity Numbers section, which is for college ball.

"It’s spelled S-H-U-F-E-L-T-M-E-U-P"

by JShufelt on Aug 7, 2009 9:51 AM PDT up reply actions

Success rate still doesn’t account for the problem of evaluating a defense while controlling for plays. Wouldn’t you expect tired Defenses would perform worse the less tired? So when evaluating Aliotti and our D this seams like something that needs to be control for.

Btw, PPP is Points allowed Per Play (points allowed/plays the D is on the field) – it’s half of S&P – football outsiders version of OPS. Success rate is the other half.

by bradLL99 on Aug 7, 2009 10:48 AM PDT up reply actions

Brad, is tired defense not balanced by also tired offense? Both have played the same number of plays. Maybe you’re considering “tired” to have a larger impact on the defense, which could be true. How much is the spread of the score in the fourth quarter a factor? Back-ups in, prevent defense, yards come easier, maybe the defense gives up a couple meaningless scores when up by 30?

I’m surely not the math wiz some of these other guys are, but I’ve learned to appreciate the truth revealed by the numbers. I enjoy where you’re going with this.

I'm really tired of it not being football season.

by JConant on Aug 7, 2009 11:22 AM PDT up reply actions

I have always found it funny that the defense gets so tired but never a mention of the offense. Did you know that opposing offenses had to run more plays against us than any team besides Mizzu, phew I need a G just thinking about that.
Seriously, I like where this thread is going, I cant contribute, but its a good read and you guys are getting somewhere.

Sleeping under an avalanche with Cartman, wake me Sept 3.

by trumpetduck on Aug 7, 2009 12:09 PM PDT up reply actions

Perhaps, but the “tired” variable I think is incalculable. There are some metrics that could correlate, such as 3rd down conversion rates (Which IS factored with Success Rate), and plays per game.

I think it’s also evidential that Oregon’s offense, despite having low TOP, also tires out defenses very quickly, because they are unable to substitute effectively, and generally don’t have 20-30 seconds of rest between plays.

Oregon plays 30% faster than most college teams (I can show these numbers), but the Oregon opponents average the same tempo as most other teams have.

I compared Oregon (Lowest TOP), LSU (Middle of the Pack – 60th in TOP), and TCU (Highest TOP – Best defense).
The thing that really strikes out to me is TCU’s 3rd down conversion differential between them and their opponent, and interception margin. +7 (Fumble recovery is more luck)

TCU: 51.61%
Opponent: 28.74%
Differential: +22.87
Interception Margin: +7

LSU: 38.25%
Opponent: 31.15%
Differential: +7.1
Interception Margin: -10

Oregon: 39.56%
Opponent: 39.38%
Differential: +0.2
Interception Margin: +4

My opinion on these stats are this:
I think the amount of time on the field isn’t as much of an impact on the players as most people do. TOP shows, in an average tempo offense, the results of turnovers (By interceptions and 3rd Downs). TCU Dominated the time of possession, and had a large interception Margin.

S&P+
No.1 TCU: 171.25 (19 points higher than No 2. UCS)
No. 27 LSU: 114.30 (57 points lower than TCU, 10 points higher than Oregon)
No. 53 Oregon: 104.38 (10 points lower than LSU)

The statistical whining really comes from people looking at Oregon being ranked 114th in the country in passing yards allowed, and they remember “Oh so many 3rd down conversions given up” – when they also have one of the largest sample sizes to remember from. (226 3rd downs) These are terrible metrics, and people that use them are being stupid. It isn’t like were Texas A&M, and gave up 3rd downs 55% of the time. Oregon’s defense was more average than bad, and more average than good.

It’s was good enough to keep us in any game against anyone, but not good enough to rely solely on it against any better than average offenses.

"It’s spelled S-H-U-F-E-L-T-M-E-U-P"

by JShufelt on Aug 7, 2009 12:03 PM PDT up reply actions