Welcome to the News Editorial section of PackerChatters.com where you will find Green Bay Packers news updates throughout the year. Packer fans editorial's, pre and post game reports, draft talk and more.

Wednesday, October 12, 2005

Are 'Stats' garbage?



by Rick
For PackerChatters

Hrm, I did not mean to call all stats 'garbage' but to call passer rating a 'garbage stat.'

First of all, just about every stat can be dismissed as garbage because sometimes they (stats) don't reflect what happens on the field adequately. But my two primary responses to that criticism of the passer rating stat in particular are: 1) What other measurements do we have (or can we create) that are better? 2) Most of the time, the passer rating actually does reflect on-the-field performance; there is a great deal of validity between the number and the perception of a passer's performance, especially when measured over the course of an entire season.

What measurements do we have that are better?

Back in the 1930s, our own Arnie Herber was widely considered the best passer in the game, and the NFL needed a number to cement it. He could throw nearly the length of the field and he had a cannon on him. But he only completed about a third of his passes, and he threw plenty of interceptions with his high risk bombs. No matter. Arnie won 3 passing titles in the 1930s because, back then, all one needed to do was add up a guy’s passing yards to win the passing title. Throw enough passes, and you win.

Then the pinpoint accurate Sammy Baugh came along in the late ‘30s, and many thought he was the best. So, suddenly the short pass and completion percentage was in. From 1938 to 1940, the passing title was given to the quarterback with the best completion percentage. (Baugh completed 70 percent of his passes one year.)

Then, in the ‘40s, the NFL chose to rate passers by ranking them against each other (the best in a category gets a 1, the second best gets a 2, etc.). They did away with completion percentage as the only measure, and instead used completions, passing yards, touchdowns, and interceptions as categories.

That was scrapped in the 1950s, when the NFL went with the yards gained per attempt as its measure of who won a passing title. Problem with that was, there were plenty of role players or running backs who threw occasionally (like Paul Hornung), but who typically had high yards per attempt percentages, and they won the passing titles. Bobby Layne never won any passing titles in the ‘50s, but he was considered the best passer back then (until Unitas came into prominence in the late ‘50s). Something had to change.

Then, in the early ‘60s, they put a minimum number of attempts in there, and they changed to the current 4 categories and weighed them equally: completion percentage, yards per attempt, total touchdowns (it didn’t become a percentage until the ‘70s), and interception percentage. That seemed to more accurately reflect what happened on the field, as the Unitas’ and Starr’s and Tittle’s and Jurgenson’s could now win or come close to winning passing titles, and those passers were indeed considered among the best in the game.

There are plenty of people who criticize the passer rating system, especially because the decision to weight the 4 categories equally is far from perfect, and it's unfair to passers from different eras. But it's an evolving science, and there hasn't been anything better that has been able to come along and stick better than the current system we have. If something better is out there, it probably won't be long before we'll start using it.

Besides, a passer rating is just one number among many other ones that are probably far more important when considering the greatness of a player. Tom Brady now has led 19 4th quarter comebacks in a little over 4 seasons, and he has thrown thrown just 3 total interceptions in 9 playoff games. Johnny Unitas threw a TD pass in 47 straight games. Brett Favre has started 200-and-some straight games. Dan Marino threw for 5,000 passing yards and 48 TD passes in 1984. Joe Montana threw 45 TD passes in the playoffs and won 4 Super Bowls. Those are stats that are going to be remembered when pondering the greatness of those quarterbacks. Few people care what John Elway's career passer rating was just 79.9. He's still considered by many to be one of the top 5 to ever play the game. That's because it's understood that passer rating, while helpful, doesn't come close to giving us all the information we need to decide how good a quarterback is. And it shouldn't be expected to either.

Most of the time, passer rating does reflect the perception of on-the-field performance.

Last year, for example, which passers were recognized as having the best statistical seasons if you were just looking at their TDs, INTs, passing yards, or completion percentages, but never looked at their passer ratings? Peyton Manning. Daunte Culpepper. Donovan McNabb. Drew Brees. Ben Roethlisberger. Tom Brady. Brett Favre. Interestingly, all 7 of those quarterbacks had passer ratings in the top 10 in the league. Very rarely you’ll get some exceptions.

It’s kinda like the Pro Bowl. Every once in a while a guy will get selected who doesn’t deserve it, and another guy who does deserve selection will get overlooked. It’s a popularity contest, they say. Well, of course it’s a popularity contest. The guys who are popular do indeed get selected for the Pro Bowl. But how do they become popular? Because they’re good. The better a player you are, the more likely you are to get noticed. Similarly, there’s a heavy correlation between a passer’s stats and the perception of how good he is. The passer rating puts a stamp of numerical approval on the perception that Peyton Manning was 2004’s best quarterback, for example, and that Mark Brunell was one of the worst. And, not surprisingly, Manning had a 121.1 passer rating, and Brunell had a 63.9 in 2004.

Now, that’s not to say that the passer rating statistic hasn’t become outdated considering the changes in the passing philosophies across eras. Back in the ‘60s, for example, having a 5.5% interception percentage (11 interceptions every 200 attempts) was considered dead average when the current passer rating system was created. That would mean that a quarterback who today throws the typical 500 times a season could throw 28 interceptions before he’d have sunk below the average mark in the ‘60s. In 2002, by comparison, the average interception percentage across the entire NFL was 3.05%, which would mean that the average quarterback threw just 15 interceptions every 500 pass attempts that season. Also, with today’s emphasis on short, safe passes over higher risk bombs, today’s passers are given undue credit for being superior than their predecessors (i.e., of the 20 quarterbacks with the best passer ratings, 13 are currently playing, including Jeff Garcia at #9, Brian Griese at #15, and Brad Johnson at #17). But comparisons between eras is another discussion. I could go on and on about that, so I’ll just leave it here.

That means it takes a bunch of things and mixes them all together without trying to weight them properly. Kinda like if I took singles, doubles, triples and homeruns and rated each of them on a scale and then considered them of equal value. I think that's foolish--homeruns are obviously of greater value than singles.

I don’t think that having a high TD percentage is far superior to avoiding interceptions in nearly the same way as homeruns are far superior to singles. TDs and INT avoidance are much closer to equal in their value, in my opinion. Now, I do agree if you are suggesting that completion percentage is given too much weight. It’s essentially given double weight because the yards per attempt percentage is very similar to completion percentage. If they de-emphasized completion percentage and just measured yards per attempt with TD and INT percentage, I think that would be a fairer system. Otherwise you’re heaping rewards on the weak-armed dinkers and punishing the strong-armed bombers. Other than that, I think the weighting system isn’t nearly as inequitable as saying homeruns are the same as singles in their value.


I think that TDs are generally a better indication of a QB's performance, because TDs matter most and they tell you two important things: how many points a QB can put on the board, and whether he's trusted enough to throw the ball regularily in the red zone.


Well, then what about a QB like Troy Aikman? Or Bart Starr? Neither QB ever threw many TD passes. And it had nothing to do with whether or not their coaches trusted them. Heck, Bart called his own plays. It’s just that when those QBs got inside the 10 yard line, they typically just handed the ball off several times because they had Hall of Fame runners who could get the job done too. Had 2 of the goal line plays been shovel passes to Fisher rather than handoffs to Davenport this past Sunday, Brett could have had 5 TD passes and an extra 15 points added to his passer rating. Do you think that shovel passes are that much more superior to handoffs that Brett deserves that much more credit for making the play?

The bottom line in my view is there's very little that is truly individual about a QB's 'statistics' they are a record of the offense's performance and all the other players out there as well as the gameplan and playcalling impact on it, and trying to seperate such things is very difficult, if not nigh impossible, and consolidating everything down to a number that mushes things of varying importance together (that will often by extremely affected by conditions of era) to get a single number isn't very helpful, though I must admit it can be convienant.

I think the baseball pitcher comparison is apt here. Yes, in order for Bob Gibson to have compiled a 1.12 ERA with 13 shutouts and 22 wins in 1968, he had to have fielders who could record outs for him and as well as drive in some runs to give him a lead. Like a quarterback has to rely on his line and his receivers (and his running back and the defense) to compile the passing statistics he does, Bob Gibson did have to rely on the other 8 players on the field to experience his individual statistical success. But does that mean that Bob Gibson’s on-field performance wasn’t the most important, most critical factor in whether or not the Cardinals won the game when he pitched?

Finally, I tend to think about a passer rating as something like an IQ score. An IQ test consists of several subtests involving content areas like processing skills, memory skills, experiential knowledge, and reasoning skills. These subtests are standardized are weighted equally and then the scaled scores are combined and divided to produce a number, our IQ. Our intelligence probably does involve much more than what these tests measure, and maybe our reasoning skills should be weighted much more than our recall skills when figuring our IQ score rather than equally. An IQ test isn't going to be close to perfect. But an IQ score probably does tell us something about how intelligent a person is. It isn't just a motley assortment of numbers that are derived from unrelated sources. It's not a reality, but it's a good guide, anyway. I think similarly about a passer rating. It's a score. A grade. And right now, it's probably the most convenient statistic we have to compare one passer to another. And convenience is good.
Comments:

<< Home

This page is powered by Blogger. Isn't yours?