What Correlates most with Winning? WARNING: STAT MONKEY THREAD!!!

I hope the Mods won't mind my tongue in cheek jab at myself in the title. I know calling people stat monkeys isn't nice, but am hoping it's ok since I'm ripping on myself.

What I did was took loads of data on every team from

http://www.pro-football-reference.com/

and copied into an excel spreadsheet. This data was all totals for the entire year in several offensive and defensive categories including points for, points against, Passing TDs, Passing yards, INTS, Rush Yards, Rush TDs, Wins and Losses, DEF Passing TDs allowed, DEF INTs, DEF passing yards against, DEF rush yards against, DEF Rush TDs against, Total TDs, DEF Total TDs against.

From this data I also compiled Winning Percentage, Points Ratio, TD/INT ratio, and INT/TD ratio.

I took just about every category and calculated (in Excel) how it correlated to winning percentage. As I mentioned, this is overall data from all 32 teams. It's a good sample size, meaning we can be somewhat confident in the results. Before I go any further, he's some data on correlation between variables:

Hence, when we see a positive number approaching 1 we can assume as the variable I'm comparing to WP (Winning Percentage) increases, so will winning percentage. Vice versa for a number approaching -1. As X increases, WP will decrease.

I should note, that none of these numbers are exactly 1, so none of this should be looked at as an absolute statement. These are very simple comparisons. Irrespective of other categories. That is to say, If I'm comparing total TDs to WP, it doesn't factor in TDs against. If I wanted to get crazy I'd compare multiple variables, do regression analysis, confidence tests, etc. but since I'm not getting paid and since I've plain forgotten how to do a lot of this, it's just a simple correlation test. Here's the data, and I'm going to present it in order of highest correlations to WP to lowest. To keep this from becoming a novel, I'll only examine in detail some of the correlation coefficients > 0.500.

Points Ratio to WP: Correlation Coefficient = 0.890
Total Points Against to WP: Correlation Coefficient = -0.721
Total Points For to WP: Correlation Coefficient = 0.707
Total Offensive TDs to WP: Correlation Coefficient = 0.636

Que John Madden. "The team that scores the most points is probably going to win." No shocker here. Also no surprise that some of the teams that had plenty of points and TDs but posted poor winning percentages had troubles on defense. Note that the points ratio (points for/points against) correlates more to winning than total points. Common sense. This ratio is one of the few that captures both offense and defense and correlates it to WP, so it makes sense that it's correlates highest to WP.

TD/INT ratio to WP: Correlation Coefficient = 0.631

The forgotten QB stat. Especially when discussing and comparing QBs on this board. I included every TD and INT thrown by every player on every team. A strong TD/INT ratio is 2nd only to total offensive TDs in terms of it correlation to winning percentage. Makes sense. Most of us will agree that QB play is crucial. Mistakes such as turnovers are devastating. This category is one of the most individual categories I looked at. That is to say, for almost every team, one person was responsible for the bulk of the TD/INT ratio. So we see here that a stat largley based on the play on one individual correlates quite highly with a team's overall success. And the TD/INT ratio correlates higher, at least based on this data than total yards or total TDs thrown. It is also greater than the negative correlation to total INTs.

Passing TDs to WP: Correlation Coefficient = 0.585

Initially I thought this might have a higher correlation to WP than TD/INT ratio. It's close. I'm not sure if there's enough of a gap to conclude the ratio of TDs to mistakes is more important than total TDs thrown, but it's interesting to note.

Rushing yards against to WP: Correlation Coefficient = -0.560

Stands to reason the more rushing yards a defense gives up the worse a team will do. Of note here is how much more rush yards against correlates negatively with winning percentage than passing yards against (-0.127). I think this is probably due to the fact that poor teams often end up with a lot of passing yards due to playing from behind. Inversely, good teams tend to run more against poor teams when they have a lead.

Total TDs against to WP: Correlation Coefficient = -0.518

Another common sense negative correlation. Interesting that it's less a negative correlation than the positive correlation of total TDs scored to WP. If I had to venture I guess I'd say the fact that teams with great offenses tend to give up a few more TDs due either focusing more on offense than defense, or giving up TDs in garbage time.

Other correlation coefficient's calculated -

Rush TDs against and WP: -0.508
INTs to WP: -0.502
Defensive INT/PassingTD allowed to WP: 0.451
Defensive INTs to WP: 0.406
Total Rush TDs to WP: 0.405
Rush yards to WP: 0.348
Passing TDs against to WP: -0.300
Passing yards to WP: 0.294
Passing yards against to WP: -0.127

Conclusions:

I use the term loosely as it's not really prudent to draw any hard and fast conclusions from correlations that are not exactly close to +1 or -1, and given the fact we understand it's all these things combined that really defines a team's success or failure in any given season. Aside from the Points ratio, which I think any of us will agree is very highly correlated with a teams winning percentage. It's also important to note that these numbers, aside from the total points, do not include any special teams data which we all know is important to any team's success or failure.

That being said, it's interesting to see how some of these individual categories, standing alone, correlate to winning. If we were to come away with anything from these numbers it would be that our starting QB really needs to have a positive TD/INT ratio. Moreso than passing yards, total passing TDs or total INTs, the ratio of TDs to INTs need to be positive. And of course, we need to score more points than we give up. But we didn't need any stats to prove that. The good news is, I think this is something Henne could conceivably correct. It might be a stretch for him to throw for more yards or a lot more TDs, but more disciplined play and better coaching could get that ratio on the right side of 1. It's also worth noting that even if Henne does throw more TDs and more yards, if the mistakes also increase it might not make as much difference as we hope.

Oh, and someone tell coach FistPump that the correlation coefficient to field goals made to WP is only 0.236.

Edit to add:

Passer rating margin = 0.781

Per CK's post, I crunched the numbers. Just goes to show how important passing well and stopping the pass is in today's NFL.

Other numbers crunched after the fact:

net pass yards/play differential = 0.648
adjusted pass yards/play differential = 0.739

Some special teams Data:
[TABLE]
[TR]
[/TR]
[TR]
[TD]AVG. Kick return yards diff to WP[/TD]
[TD="class: xl63, align: right"]0.272[/TD]
[/TR]
[TR]
[TD]AVG. punt return yards diff to WP[/TD]
[TD="class: xl63, align: right"]-0.049[/TD]
[/TR]
[TR]
[TD]KR TD Diff to WP[/TD]
[TD="class: xl63, align: right"]0.175[/TD]
[/TR]
[TR]
[TD]PR TD DIFF to WP[/TD]
[TD="class: xl63, align: right"]0.113[/TD]
[/TR]
[/TABLE]

 

Hey Padre, did you already do something like this? Can I get the link to the thread? I'd enjoy reading it. Sounds like the only difference is I looked at a lot of non-QB stuff. Didn't want to focus too much on QB, but it always comes back to that...

Also if there's something else someone would like me to run the correlation on let me know. I might already have the data in my spreadsheet or I can add it.

 

Guilty. The good thing is, while I sat here compiling this, it actually looked like I was working. Boss came back and I didn't even minimize...

 

I crunched the numbers on it, and for 2010 it's second only to Points / Points allowed ratio at 0.781. You mentioned it predicting championships -

Guess which team had the highest positive differential? Packers.
And the worst? You guessed it - Panthers.

 

As simple as some of this stuff is and as much as it seems like common sense, I'm worried if our front office and Tony Sparano are aware of some of it. When Tony chooses to compare Henne's numbers to other QBs numbers but not his TD/INT ratio and when Jeff Ireland refuses to invest more resources into the QB position it makes you wonder.

 

The Chargers were an outlier in just about everything. Great point ratio, passer rating differential, TD/INT, etc. Goes to show how important special teams is and how bad theirs was. Having Norv as head coach probably doesn't help either.

 

I crunched the other variable CK mentioned which was pass yards / play differential.

Quite a bit lower than pass rating differential at 0.648 but still higher than most others besides points differential.

But if we're talking about gambling and predicting the results of an individual match up between 2 teams, I'd have to say, I might have to lay some money using pass rating and pass yards / play differential this year.

 

Got a specific piece of it you want me to look at? I could probably look at some kind of average kick return differential. I did Field Goals made special for coach Fist Pump and it was 0.236 I think. So positive, but not a very strong relationship at all.

 

Here's some stats on special teams -

[TABLE]
[TR]
[TD]AVG. Kick return yards diff to WP[/TD]
[TD="class: xl63, width: 46, align: right"]0.272[/TD]
[/TR]
[TR]
[TD]AVG. punt return yards diff to WP[/TD]
[TD="class: xl63, align: right"]-0.049[/TD]
[/TR]
[TR]
[TD]KR TD Diff to WP[/TD]
[TD="class: xl63, align: right"]0.175[/TD]
[/TR]
[TR]
[TD]PR TD DIFF to WP[/TD]
[TD="class: xl63, align: right"]0.113[/TD]
[/TR]
[/TABLE]


Which is almost none. Punt return yardage being basically so close to 0 demonstrates no relationship. I'm betting fair catches and how few really gifted punt returners there are in the league explain this.

I would expect this category to continue to not correlate much with winning at all with the new kick return rule this year.

Still, that doesn't mean special teams doesn't have a profound affect on a team's wins and losses if they're especially good or especially bad on special teams. It's just that all the other teams are so even in these categories that the correlation doesn't come through.

 

I should point out, I (accidentally) used net yards / passing play (factors in sacks) in my analysis. In that regards, the Falcons and Dolphins were dead even in 2010 on both offense and defense. To explain the difference, you need only look as far as the TD/INT ratio of the two players.

Also, pro football has a neat stat called adjusted yards / pass play which factors both sacks and interceptions. Using those numbers -

Miami: 4.9 - 6.1 = -1.2
Falcons: 6.2 - 5.2 = + 1

 

When using adjusted yards per pass play, which again factors sacks and interceptions the correlation jumps to 0.739.

 

The stats I used were all differentials between offense and defense, but I thought I was using just avg. yards / pass when in reality I was using the net (which factors sacks).

To see the exact formula for adjusted yards/pass play go to profootballreference.com and hover over the column in passing stats. I tried to copy it, but it goes away as soon as I take the cursor off it.

Damn it must be great being a professional gambler. But I'm betting retirement could be tricky...

 

Post #31