Help - Search - Members - Calendar
Full Version: Completion % and Win Correlation
Eagles Forum > Philadelphia Eagles Message Board > Philadelphia Eagles or Football Related Discussion
mcnabbulous
Since there is and has been so much talk on this board about the correlation between a ridiculously arbitrary running number and it's correlation with winning games, I decided to spend some time looking into a stat I think is far more relevant to a team's success. That number being completion percentage.

I decided to look at Donovan's long career and Nick's short career, respectively. For both guys, I eliminated their rookie year, which obviously makes Nick's sample size incredibly small.

Here's how it played out:

Donovan (Total 88-36 .709)
Less than 50%
7-14 .333
50-55%
12-7 .632
55.01-59.99%
19-9 .680
60-65%
18-9 .667
65.01% +
32-5 .865

Playoffs
Less than 50%
0-2 .000
50-55%
1-2 .333
55.01-59.99%
1-2 .333
60-65%
3-1 .750
65.01% +
4-0 1.000

Foles (Total 14-4 .777)
Less than 50%
0-2 .000
50-55%
1-0 1.000
55.01-59.99%
1-1 .500
60-65%
5-1 .833
65.01% +
7-0 1.000

Strangely enough, Nick has only lost one game in his career when completing over 65% of his passes. Sadly, it was our playoff loss to the Saints. My personal opinion on that is that he played far too conservatively in that game.

Not sure if the dataset is useful or not, but it's pretty clear that if we can find a guy that can consistently complete over 60% of his throws on a game to game basis, we will win a lot of games.

I think the stat that most surprised me was the significantly high number of games in Donovan's career where he went over 65%.

Hope others find it interesting or informative.
HOUSEoPAIN
QUOTE
Since there is and has been so much talk on this board about the correlation between a ridiculously arbitrary running number and it's correlation with winning games


Again, it isn't an arbitrary number. When your All-Pro RB gets 20 carries in a game, that implies an establishment of the running game. The backups get a few runs, the QB has a called sneak or two, and that works out in most games to at minimum, a relatively balanced attack. In fact, I would argue that pass completion % and rushing attempts (or better yet, rushing attempts as a % of total play calls) correlate positively, as the defense doesn't know what to play for. As has been demonstrated many times on this board, the defense usually knew exactly what to play for when Reid was here. I've used the number 20 from our featured backs (not a total of running plays) for simplicity, I didn't pull the number out of my ass. I'm relatively proficient and experienced in statistical analysis, or at least I can claim so based on the fact that several companies have hired me for that reason.

QUOTE
I think the stat that most surprised me was the significantly high number of games in Donovan's career where he went over 65%.


Who woulda thunk? You mean the 6 time Pro Bowler with just under HoF numbers could actually complete passes, and didn't suck?

Your data set is useful, interesting, and relevant, but it doesn't tell the whole story. For instance, the 'Winston Justice game,' our 16-3 loss to NY in 2007. He completed less than 50% of his passes (15/31), but he was sacked 12 times. While sacks obviously don't directly factor into pass completion %, he was under heavy fire all game, and he was also throwing to Reggie Brown and Kevin Curtis. Buckhalter carried 17 times for 103 yards, 6.1 yards per carry. What would've happened to McNabb's completion % if Buckhalter was given more carries, and he wasn't getting hit every other play? Hurries and hits are in-game stats shown on tv and are certainly subjective, and aren't officially recorded to my knowledge, so that's impossible to determine other than the logical assumption that he was constantly given a very short time to make passes. When looking at overall rushing attempts as a % of play calls, McNabb also ran 4 times for 4 yards that game. Obviously, you can presume those were all pass calls where he ended up running for his life. And this is just one example from one game.

The most intellectually honest way of looking at the entire picture with the data we have access to IMO is to look at completion %, the amount of running/passing plays called as a % of total plays called, turnovers, points scored by defense and special teams, and the amount of drops from receivers, and how those 5 factors correlate to winning percentage. That still wouldn't be absolutely perfect because it's impossible to determine the exact number of called QB runs as opposed to scrambles, as well as the 'running out the clock' factor, but it's the best way I can think of. Perhaps only using data from the first 3 quarters? There's plenty of other factors you can use too, such as penalties, and starting field position. If you call 3 pass plays and get 3 incompletions, but start from your opponent's 20, you're still getting points.

Maybe one day when I have several hours to kill I'll run that. In the mean time, feel free to try it yourself if you want.
mcnabbulous
QUOTE (HOUSEoPAIN @ Nov 22 2014, 07:05 PM) *
Again, it isn't an arbitrary number. When your All-Pro RB gets 20 carries in a game, that implies an establishment of the running game.


It really doesn't. What it implies is a gameflow that facilitates the RB getting 20 carries. Additionally, the number itself is arbitrary. Why not 19? Why not 21. It's just round and fits your narrative.

QUOTE
The backups get a few runs, the QB has a called sneak or two, and that works out in most games to at minimum, a relatively balanced attack. In fact, I would argue that pass completion % and rushing attempts (or better yet, rushing attempts as a % of total play calls) correlate positively, as the defense doesn't know what to play for.

I would only wager that they correlate positively as a result of a higher completion % giving the offense a higher % of run-friendly down and distance situations.

QUOTE
As has been demonstrated many times on this board, the defense usually knew exactly what to play for when Reid was here.

This has literally never been demonstrated with any substance. As many people have claimed that Reid is one of the most creative playcallers in the NFL as have said he's predictable. Additionally, as many have claimed Chip to be predictable as said the same about Reid.

QUOTE
I've used the number 20 from our featured backs (not a total of running plays) for simplicity, I didn't pull the number out of my ass. I'm relatively proficient and experienced in statistical analysis, or at least I can claim so based on the fact that several companies have hired me for that reason.

You chose it for simplicity, meaning it's not relevant. As I pointed it, it's irresponsible to run a back like McCoy 20 times per game. His numbers this year are increasingly proving my point.

QUOTE
Who woulda thunk? You mean the 6 time Pro Bowler with just under HoF numbers could actually complete passes, and didn't suck?

I've never said McNabb sucked. I think he was an inconsistent passer. The fact that his career completion % was 59%, his best season ever was 64%, and that most people agree that accuracy wasn't his biggest strength is why I found that specific piece of data to be interesting.

QUOTE
Your data set is useful, interesting, and relevant, but it doesn't tell the whole story.

Of course it doesn't. It's just far less dependent on other variables than your arbitrary running number.

QUOTE
For instance, the 'Winston Justice game,' our 16-3 loss to NY in 2007. He completed less than 50% of his passes (15/31), but he was sacked 12 times. While sacks obviously don't directly factor into pass completion %, he was under heavy fire all game, and he was also throwing to Reggie Brown and Kevin Curtis. Buckhalter carried 17 times for 103 yards, 6.1 yards per carry. What would've happened to McNabb's completion % if Buckhalter was given more carries, and he wasn't getting hit every other play?

Buckhalter probably would have been in a position to get more carries had we not taken 12 sacks in the game. Those 12 sacks inevitably put us in suboptimal 2nd and 3rd down situations throughout the game.

I also suspect he would have had more carries had we not gone into the 4th quarter with a 16 point deficit. Once again, gameflow.

QUOTE
The most intellectually honest way of looking at the entire picture with the data we have access to IMO is to look at completion %, the amount of running/passing plays called as a % of total plays called, turnovers, points scored by defense and special teams, and the amount of drops from receivers, and how those 5 factors correlate to winning percentage. That still wouldn't be absolutely perfect because it's impossible to determine the exact number of called QB runs as opposed to scrambles, as well as the 'running out the clock' factor, but it's the best way I can think of. Perhaps only using data from the first 3 quarters? There's plenty of other factors you can use too, such as penalties, and starting field position. If you call 3 pass plays and get 3 incompletions, but start from your opponent's 20, you're still getting points.

Maybe one day when I have several hours to kill I'll run that. In the mean time, feel free to try it yourself if you want.

While I agree that a multitude of factors correlate to winning and losing, the reason I think completion percentage is so relevant is because of the way it directly impacts playcalling, 3rd down conversions, and ball control. The only real dependency is drops, which I suspect is probably statistically insignificant as it relates to this dataset.

It doesn't surprise me at all that 4 of Alex Smith's 5 lowest completion % outings of the season have resulted in losses. If you can't complete routine passes, you will struggle to have offensive success.

I'd look forward to seeing what you come up with if you ultimately run those numbers.
mcnabbulous
FWIW -- I decided to run the same type of scenario for carries just as an exercise. I did it for Westbrook's 5 years as the feature back and for Shady's career beginning with his 2nd season.

Westbrook (Total 40-26 .606)
Less than 10
3-2 .600
10-14
13-9 .591
15-17
8-7 .533
18-19
4-3 .571
20+
12-5 .706

McCoy (Total 37-31 .544)
Less than 10
0-3 .000
10-14
8-11 .421
15-17
4-10 .286
18-19
5-3 .600
20+
20-4 .833
nephillymike
Interesting discussion.

A few thoughts:

It is not a very difficult thing to do, but I think you can do a comparative analysis for the Eagles games in any subset you choose and look at both teams in various stats to see who won the game more. I think that rushing yards, while probably a decent correlation to win % is misleading because teams rush the ball a lot after they get a lead to run the clock down.

I'll throw out a few stats and guess as to their correlation with victory:

Point Differential = Perfect rolleyes.gif 100%!!
Yards Differential = Moderate Correlation 65% (guess)
Turnover Differential = Very High 80% (actual)
Total yards Offense = Moderate 60% (guess)
Total Yards Allowed Defense = Moderate 60% (guess)
Home Team = Moderate 57% (actual)
Rushing Yards = High 70% my guess
Yards Per Carry = High 75%, my guess, but stats says moderate see below
Least Sacks Allowed = Moderate 60% my guess and stats backs that up see below
Completion % = High 70% my guess, but stats say moderate see below
QB Rating = Very High 75% my guess and stats back that up see below
Points Differential by Def/ST = Very High 80% my guess
Point Spread = High 67% (actual)


"Passing Efficiency explains 62% of the wins variation in the NFL, while run efficiency explains 4.4 %"

http://thepowerrank.com/2014/01/10/which-n...n-the-playoffs/

"Turnover differential explains 44% of the variation in wins in the NFL"

http://blog.minitab.com/blog/the-statistic...-the-nfl-season

For how the components of passing correlate to wins:

http://www.footballperspective.com/correla...tats-with-wins/

The analyses mostly point to how much of a teams win likeliness can be accounted for by X. The turnover differential measure actual W-L by this, which I like. I couldn't find data that converted those other measurements to win% like the turnovers was.

However, if one had time, they could do this for all the Eagles games....................................

My conclusion based on short research:

Turnovers is highest correlation, special teams and defense scoring is 2nd and passer rating is third.
nephillymike
QUOTE (mcnabbulous @ Nov 22 2014, 09:42 PM) *
FWIW -- I decided to run the same type of scenario for carries just as an exercise. I did it for Westbrook's 5 years as the feature back and for Shady's career beginning with his 2nd season.

Westbrook (Total 40-26 .606)
Less than 10
3-2 .600
10-14
13-9 .591
15-17
8-7 .533
18-19
4-3 .571
20+
12-5 .706

McCoy (Total 37-31 .544)
Less than 10
0-3 .000
10-14
8-11 .421
15-17
4-10 .286
18-19
5-3 .600
20+
20-4 .833



I'd be interested in comparing to what the other team did that day.

Example, if the eagles ran for more YPC, if Eagles had the higher passer rating, if Eagles had better turnover differential etc
mcnabbulous
I've been a bit under the weather today, so I've spent way too much time watching football and looking at this stuff. I did one more deep dive to get a breakdown of Donovan's completion % and it's correlation with the number of non-QB carries in each of those games. I used the same ranges as before.

Less than 50%
19.21
50-55%
18.22
55.01-59.99%
20.43
60-65%
21.22
65.01% +
23.13

So then it becomes the chicken and egg argument. I tend to think we ran the ball more when the completion % went up because we were in more favorable down and distance situations and likely moved the chains more often.

It sounds like HoP thinks the running led to more unpredictable playcalling, which benefited Donnie's accuracy.

Other observations:
- That was tedious
- It was kind of interesting going back through those old box scores. Had completely forgotten about some of the guys that got carries for this team (e.g. Chris Warren)
- The most run heavy game I came across was Donnie's broken leg game against Arizona (43 carries)
- On the opposite end, there were two games where we only had 7 carries. You won't be surprised to learn that we lost both.
- Donnie was awesome in 2004
- We were really good in 2002, but Donnie was wildly inconsistent before his injury.
- Andy seemed more comfortable giving Deuce heavy workloads than Westy and McCoy. No specific numbers to back that up, but it seemed like he had several more 25+ carry games compared to the other guys.
- If there is any interest, I can share a more detailed breakdown of any of those numbers.

This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2017 Invision Power Services, Inc.