I forget who said it, but there are three kinds of lies: lies, damn lies, and statistics. I always wondered how they calculated passing efficiency - thanks for the post.
Behind the Numbers (9/16)
UPDATE: Part Two is here. Includes an alrernative to Passing Efficiency.
Today's Stat: Passing Efficiency(Full NCAA Rankings)
Players of note: Ryan Mallett, Arkansas (1st, 210.25); Jimmy Clausen, ND (3rd, 196.31); Kirk Cousins, MSU (6th, 186.71); Tate Forcier, Michigan (21st, 161.69); Terrelle Pryor, OSU (79th, 116.92)
Why it's important:It's pretty much the golden standard for measuring the (wait for it) efficiency of a quarterback. It's not flawless by any means, but overall is a pretty good indication of how good a quarterback is. Once there's a good sample size (at least 100 attempts), it's pretty safe to say that a player in the top 20 of the efficiency ratings is a good quarterback, and a player outside the top 50 isn't quite as high-caliber.
Why it's flawed:Passing Efficiency measures just that -- efficiency. How efficient something or someone is usually boils down to how much of 'x' they can do in 'y' amount of tries. It's no different in the world of college football. The equation for Passing Efficiency in College Football is as follows:
(Completions x 100) + (Yards x 8.4) + (Touchdowns x 330) - (Interceptions x 200)
So while all of that stuff on top is really important, it really boils down to how many passes the quarterback has attempted. For example:
Quarterback A plays basically the whole game and racks up some pretty good numbers, but in the red zone gets bruised up and comes out for a play.
Quarterback B comes in for that one play and throws an eight yard touchdown pass, and is right back on the bench, and remains there for the rest of the game.
Quarterback A's stats: 28/35, 310yds, 3 TDs, 1 INT
Quarterback B's stats: 1/1, 8yds, 1 TD
Go ahead and take a stab at each quarterback's rating. Or just scroll down a bit and look at the actual answers, you cheater.
Quarterback A's Efficiency Rating: 246.2
Quarterback B's Efficiency Rating: 497.2
Quarterback B, the backup who came in for one play, isn't necessarily a better quarterback than Quarterback A.. there's actually a good chance that he's a good deal worse. His efficiency rating, however, is more than twice that of Quarterback A, who had a damn good day throwing the ball. However because that one pass attempt that he did have was a successful one, his Efficiency Rating is about 287 points higher than the current highest rating in Division 1.
Applying this to current statistics:Ryan Mallett: 17/22, 309yds, 1 TD (210.25)
In the one game he's appeared in so far, Mallett has only attempted 22 passes (remember, the smaller the sample size the more skewed the rating), and completed 17 of them. A 77% completion percentage is second only to Sean Canfield (OSU, NTOSU), who has the 14th highest efficiency rating. He only has the one touchdown and has yet to throw a pick (not as important as you'd think, as you'll see later). Not stellar numbers by any means, but he did pretty well against Missouri State.
Jimmy Clausen: 40/60, 651yds, 7 TDs (196.31)
Not too much to say here, the efficiency rating is pretty well deserved so far. Quite the interesting comparison to Mallett's numbers, however. Clausen's numbers are obviously superior in every way but completion percentage. Clausen is clearly the superior quarterback here, yet because of the small sample size in Mallett's case, he has the higher rating.
Tate Forcier: 36/53, 419yds, 5 TDs, 1 INT (161.69)
Tate's numbers compared to his rating are also pretty interesting. He actually has a higher completion percentage (67.9) than Clausen (66.7), has a respectable touchdown percentage (9.43% of his passes are touchdowns, compared to Clausen's 11.7%), and only has the one interception. However even if we take that interception away (it wasn't even his fault!), Forcier's rating doesn't improve too dramatically. If the pass fell harmlessly to the ground, his rating would be a 165.5, good for 17th. If the pass was completed for a 15 yard gain his rating would be a 169.7, putting him in 16th.
The TakeawayQuarterback Efficiency Rating is an effective way to rank the overall efficiency of quarterbacks, especially later in the season once there is a decent sample size of attempts to go by. Until then, however, it's a stat that's easily skewed by a few attempts going for big yards and touchdowns. We all know Quarterback A in the example above had a better game than Quarterback B, but the formula for efficiency rating doesn't. Quarterback B did complete 100% of his passes, and 100% of his attempts went for touchdowns.. the thing is there was just the one attempt. Therein lies the flaw.
Just for fun, try to guess which stat line would garner the higher efficiency rating. Answers are at the bottom of the post.
Situation 1 Situation 2
A. 25/30, 250yds, 2 TDs, 2 INTs l A. 30/40, 300yds, 1 TD
B. 15/17, 140yds, 1 TD l B. 10/12, 100yds, 2 TDs
A. 20/24, 200yds, 2 TDs, 4 INTs
B. 20/40, 250yds, 5 TDs
Behind the Numbers will be back soon with another look at a stat from the world of College Football. Any stats you want to be examined a little closer? Or even just a stat you've been interested in for a long time? Let me know in the comments and I'll do my best to get to it in the next few installments of BtN. Thanks for reading!
Situation 1- A: 162.0 B: 176.8; Situation 2- A: 146.2 B: 208.3;
Situation 3- A: 147.5 B:143.75
Mark Twain for 400 please Alex...
Ahh, I see you've learned how to use the google.
How did they come up with those exact numbers to multiply in the formula. Example: why is a touchdown pass multiplied by 330 and not 430? Why are yards multiplied by 8.4? Etc...
I wondered the same thing.. I couldn't come up with anything though. Any help from anyone?
When the idea of passing efficiency was first thought up, they used statistics from previous seasons to establish constants which would give the average passer a rating of 100.
That's what the great google machine told me anyways, so it has to be true.
There are a BUNCH of different weighting factors that would give an average rating of 100. The particular weights that they use end up being a kind of statement of values (i.e. an interception is twice as bad as a completion is good).
And actually ANY combination of weights can be normalized - i.e. transformed so that the average quarterback gets a rating of 100 with the good ones high bad ones lower.
The google machine is surely right but about what the creators of the stat thought they were doing but it's not a complete explanation.
My guess is that they started with
(comp * 100) + (yards * 10) + (TDs * 300) - (Ints * 200)
and went from there.
I think this could be another flaw in the stat, then. Unless I'm reading it wrong, it seems like they threw some random numbers together to multiply by.
One of my favorite stats for baseball is wOBA. It looks at mounds of historical data and applied values to walks, HR's, singles, etc. based on how important they are for a hitter in creating runs. If similar research could be done and assign appropriate numbers to short completions/long completions, or interceptions or touchdown passes or the amount of yards thrown for, etc. then I think the stat would be much improved.
Just in general, I think there is a need for more advanced stats in CFB. It would be much harder to do than baseball b/c of how complex football is but I do think the Football Outsiders have done a good job so far with NFL data. Catch rate, for example, appears to be a good advanced stat for receivers. It is hard to design stats for offensive lineman, but rush yards in their general direction/sacks allowed/pancake blocks/percentage of times not allowing a sack could be a place to start for a possible formula. Corners and safeties allowing catches in their zones could be analyzed, in addition to their pass breakups, int's, tackles/missed tackles. A running back's ability to block is probably able to be quantified in some way. In addition, breaking tackles/yards after contact, yards per carry, "long runs," consistent positive yardage, touchdowns, fumbles, catch rate, etc. could be a place to start for a potential formula for a RB. For D Lineman, you would need to look at how many yards were allowed rushing in their general direction, pass hurries, sacks, tipped balls, how many times they were pancaked, tackles for loss, etc. For linebackers, missed tackles/tackles in open space could probably be quantified. I guess you could also look at how many rush yards were accumulated in their zone on a particular play/ pass yards were accumulated in their zone per play. There would probably also be a way to factor in what player they were assigned to cover during a particular way, and you could apply all that to a formula with tackles for loss, tipped balls, assisted and unassisted tackles, interceptions, fumbles caused, etc.
Just thinking out loud and rambling. I do think, eventually, stats will be taken to the next level in football.
A. 21/24, 200yds, 2 TDs, 4 INTs
This stat line is impossible by the way.
one player has only 3 incomplete passes, but 4 picks.
Ah, nice catch. That's meant to read 20/24. Thanks!
Great Post. But (you knew there had to be a but) you posted:
"It's not flawless by any means, but overall is a pretty good indication of how good a quarterback is."
We had a few threads at the end of last year discussing this (in relation to TT and why he [sniff, sniff] did not win the Heisman).
The formula does not include any statistics for running by the QB.
So, it may be a pretty good indication of how good a QB is at passing the ball but it is not very good in college football for overall QB evaluation because many QBs have lower ratings because they are expected to run.
It's good when taken into context. No one stat is perfect, but when multiple stats are looked at together, and in context (i.e. the number of attempts in this case), they are a pretty good indication.
Passing efficiency is fine, but it doesn't take into consideration other factors like: running ability, decision making at the line of scrimmage, and intangibles. It would be nice to be able to quantify these other factors with efficiency ratings to have an overall rating ro QB's. Actually, won/loss rankings maybe does that.
Stop with this "intangibles" junk. It is by definition impossible to take "intangibles" into account in any sort of statistic or formula.
Actually I agree. I guess my point was that it would be nice if we could quantify those things for an overall QB rating. I neglected to add the "but we can't" part.
It's true that it's impossible to take "intangibles" into account.
I'd go one further (or, perhaps, come at it from a different direction) and say if it truly is "intangible" then it has no impact. Anything that has an impact will have a tangible result that can be measured.
Anyone who believes otherwise is either still looking for the right stat (which is fine with me) or simply disagrees with the numbers (which isn't).
I almost went that far and said that. I agree with you - the concept of "intangibles" is way way overused and may not even exist.