Stat's that are the most meaningful.

[Get it shawtaay]

For defenses, I propose points-per-length-adjusted-drive-minus-clock-stoppage. Look at two games - one where the defense gave up 23 points and one where the defense gave up 20 points.

Length adjusted drive = length of drive/80
PPD-ADJ = points scored on drive / length adjusted drive
Drives where the first play is a kneel are excluded because the offense is not trying to score.

Meets the eye test far better than PPG.

 

[Joker]

that's one helluva first post, shorty...


how about failed love lives...you got any formulas for THAT?...I'd like to find out WTF went wrong and why I'm single with only a cat to show for my five + decades......

 

[Get it shawtaay]

that's one helluva first post, shorty...


how about failed love lives...you got any formulas for THAT?...I'd like to find out WTF went wrong and why I'm single with only a cat to show for my five + decades......

I do not have a formula for the game of football. I can merely attempt to improve upon the existing attempts to model its outcomes.

Similarly, I cannot solve your girl problems. But if attempts to fix that problem are as rudimentary as points per game and yards per game, there is surely room for improvement.

 

[Joker]

so...you're saying I'm a few yards short of a first down in the game of life?

that's cold, man...

 

[jason423]

I track point efficiency and it usually has a pretty high correlation with records. The numbers are basically a measure of how much teams score above their opponents average allowed and by how much a teams holds the opponents down from their norms. This way it takes into account SOS. FWIW the Patriots have the top scoring efficiency in the league at 41.37% and a defensive score of -0.59%. At that pace the predicted wins would be 11.8 with a stdev of 1.23. Overall they rank 6th right now. The rest of the AFCE is awful ranking 23, 25, and 29 (prior to the Mia/Buff game Thursday). I used to track the home and away splits which is probably more accurate since many teams have a big variance but I got lazy and stopped doing that.

Ive been fooling around more with drive efficiency this year (actual points per starting field position vs expected points per starting field position) which I think may be a pretty strong indicator of real offensive and defensive success.

 

[PATSNUTme]

that's one helluva first post, shorty...


how about failed love lives...you got any formulas for THAT?...I'd like to find out WTF went wrong and why I'm single with only a cat to show for my five + decades......

It's probably the give aways- take aways.

 

[Joker]

yeah...dyam straight...whatever I didn't give away, she took away....

 

[JJDChE]

For defenses, I propose points-per-length-adjusted-drive-minus-clock-stoppage. Look at two games - one where the defense gave up 23 points and one where the defense gave up 20 points.

Length adjusted drive = length of drive/80
PPD-ADJ = points scored on drive / length adjusted drive
Drives where the first play is a kneel are excluded because the offense is not trying to score.

Meets the eye test far better than PPG.

You should've picked two games where your model actually did a good job of predicting the outcome. How are you going to convince people that you have a good model when you post an example that predicts 9 points and the actual score was 23?

 

[JJDChE]

that's one helluva first post, shorty...


how about failed love lives...you got any formulas for THAT?...I'd like to find out WTF went wrong and why I'm single with only a cat to show for my five + decades......

Lots of penalties for Illegal Touching, eh?

 

[Get it shawtaay]

You should've picked two games where your model actually did a good job of predicting the outcome. How are you going to convince people that you have a good model when you post an example that predicts 9 points and the actual score was 23?

The model does not produce a score that is immediately comparable to a real game. The numbers can only be compared among other adjusted scores.

To compare the PPD-ADJ scores to real game scores, they would need to be scaled. To do this I would need to collect a ton of data, and correlate the PPD-ADJ scale to similarly "impressive" game score values. Something which I have not done.

 

[BradyBranch39]

there's a ton of confirmation bias going around in this thread.

 

[Get it shawtaay]

there's a ton of confirmation bias going around in this thread.

Care to enlighten us?

 

[supafly]

You should've picked two games where your model actually did a good job of predicting the outcome. How are you going to convince people that you have a good model when you post an example that predicts 9 points and the actual score was 23?

First of all, if you notice the graph you'll see that the Cardinals scored 20 points in total in a game that we lost 20-18 (not 23 as you stated).

Secondly, one of the drives was for 2 yards coming off of a turnover, so in reality the defense gave up 13 where the graph predicted 9.25, so it's pretty close.

In the PIT game, the final score was 25-17 but 2 points came on a safety, so the defense allowed 23 points where the graph predicted 20.2875.

So what we can take from this "system" is that is can get us within a FG or so, which many others probably can too.

I'm not knocking the system itself, I actually think that it showed results that are relatively close to what it predicted. I am saying that I don't necessarily think that he did a bad job in the 2 games that he showed though as it was pretty close in both.

 

[Get it shawtaay]

First of all, if you notice the graph you'll see that the Cardinals scored 20 points in total in a game that we lost 20-18 (not 23 as you stated).

Secondly, one of the drives was for 2 yards coming off of a turnover, so in reality the defense gave up 13 where the graph predicted 9.25, so it's pretty close.

In the PIT game, the final score was 25-17 but 2 points came on a safety, so the defense allowed 23 points where the graph predicted 20.2875.

So what we can take from this "system" is that is can get us within a FG or so, which many others probably can too.

I'm not knocking the system itself, I actually think that it showed results that are relatively close to what it predicted. I am saying that I don't necessarily think that he did a bad job in the 2 games that he showed though as it was pretty close in both.

I appreciate your interest in the stat, but again it is only useful for comparing a units performance across games at this point. The PPD-ADJ values cannot be compared to non-adjusted point totals without being normalized somehow.

What you can glean from it is that the defense was universes better in the Cardinals game than in the Steelers game, even though the final scores (23 vs. 20 points allowed) don't tell that story.

When I have some more time I will refine it and try to reach some broader conclusions.

 

[supafly]

I appreciate your interest in the stat, but again it is only useful for comparing a units performance across games at this point. The PPD-ADJ values cannot be compared to non-adjusted point totals without being normalized somehow.

What you can glean from it is that the defense was universes better in the Cardinals game than in the Steelers game, even though the final scores (23 vs. 20 points allowed) don't tell that story.

When I have some more time I will refine it and try to reach some broader conclusions.

I appreciate the time, thank you.

One thing we can all agree on is that we need to cut down on the 3rd down conversions, and improve overall in the pass defense itself. That is certain. Other than that the run defense is just fine rather good many times, and the turnovers and red zone efficiency are both very good.

If they could just shave off a couple of pts off the total PPG and improve in the pass defense, we could have a very decent shot at making a deep run into the playoffs based off of the defense alone...of course that's a big "if."