Saturday, October 27, 2007

Tonight we're gonna Party like it's 1695!

Remember when Dr. Z was cool? I don't.

We have fun on this blog with Dr. Z's works. In fairness to him, he's not really a serious journalist. He's reached the point where you either accept his rankings as absolute truth no questions asked, or you realize that they are complete baloney and you are better off taking teams out of a hat and listing them as the pills fly.

Senile or not, there is absolutely no excuse for this piece of crap. I wish not to discuss the nuiances of a football game with anyone who can't see from this piece that Dr. Z has simply lost his mind.

He basically tries to denounce passer rating as a stat as utter garbage.

[I]t's a prehistoric monster that no one understands, an illogical piece of antiquity that influences so much of the game when it shouldn't. It affects what is written, what is discussed, what becomes the basis, in some cases, of salary structure and bonuses for players and coordinators.

Okay. Settle down kiddos and take a deep long breath. If you are reading this blog, you are certainly not a member of the sample universe that Dr. Z is addressing. Therefore, you are a no one to him. Because let's face it, if you are anyone, you aren't allowed to understand passer rating. Anyway, for those of you who need a refresher course: here's a basic overview of how passer rating is calculated.

Passer rating is a cumulative statistic that is split equally into 4 parts: Int/Att, TD/Att, Comp/Att, Yards/Att. Scores can range anywhere from 0.0-158.6 Each component of the stat will range between 0.0-39.7. The 4 components are added together and gives you a cumulative number. League average tends to hover around 78.0 depending on the year.

Passer rating has some major flaws. First of all, remember that the stat represents all 4 categories as equal components of QB performance. Yards/Attempt is generally seen as the most important stat, and should rank more than one quarter of the total output in a perfect system. Combined, TDs and INTs make up half the rating, but these number can skew the total BAD especially in a small sample. Additionally, Completion percentage already correlates somewhat to TDs AND INTs based on attempts, so its essentially stating the same thing 3 different ways.

A better formula would look something like this:

(Quantity INT/ATT) * .05 + (Quantity comp/att) * .5 + (Quantity Yards/Att) * .45

Where the completion percentage makes up half the total and the yards per attempt makes up 45%, leaving a measly 5% for interceptions per attempt to account for the select QBs (Favre) who always exceed their INT projection based on their completion percentage.

Here's the point. QB Rating is a main stream stat, and it gets it right in a lot of ways because of its use of the ever so important rate stats.

Now with that out of the way, back to Dr. Z totally humiliating himself.

Steve Young, who has the highest career passer rating in history, admits that he's "not quite sure how the system works."

Steve Young is an idiot. Maybe the greatest QB ever, but an idiot nevertheless. I wouldn't expect anything different, would you?

Charley Casserly, who as Redskins general manager was quite aware that some clauses were built into contracts that reflected the rating points, says, "No, I couldn't tell you exactly how they determine the ratings."

Charlie Casserly drafted both Heath Shuler and David Carr in the top 5 picks of their respective drafts.

Bill Parcells, whose 11-point dictum to quarterbacks came from years of study of the position, says, "I don't know how they arrive at their ratings and I don't care. I don't pay any attention to them. I have my own system for evaluating quarterbacks."

The key here is that he does in fact have his own system for evaluation. Dr. Z clearly does not, but curiously ignored that part of Parcells' quote. As stated above, QB Rating is hardly the be all end all, but in a large sample, you won't stray too far from reality if you use it.

Average grades were in the 60s and 70s. En masse, NFL passers in 1972 completed 51.7 percent of their heaves. A mark like that would earn a player a grade of 72.3. Average, in other words. Interception percentage, or number of interceptions per 100 passes thrown, was 5.3, league-wide. A grade of 70. Touchdowns per pass attempts averaged out to 4.5, a grade of 60, and yards per pass attempt came out to 6.82, which got a mark of 63.7.

Put all those figures together and you've got a number of 66.5 for a dead average player, hitting the norm in each category. Higher achievements, of course, would bring higher grades.

Okay. This is getting weird.

After spending the entire first page of his article criticizing a system that "no one understands", Dr. Z goes on to show a pretty solid understanding of the system as it applies to stats in the 60s and 70s. One can only wonder where he's going with this.

Now here's the snapper. Achievements have gone way above the old standards, but Elias has maintained that same system for 35 years, with the same benchmarks and the same schedule of rewards. The passing game has changed dramatically, but [The] Elias [Sports Bureau] plods on, stuck in its standards of 1973, when its system came in.

That's it folks. Dr. Z. thinks QB rating is a useless stat, because it's old and hasn't changed in thirty five years. That's...blatent hypocrisy.

But that's not all, Dr. Z now goes on the offensive.

I said that their practice of including quarterback kneels at the end of the game in the rushing stats was wrong and misleading. It penalized the good teams, which won, therefore had QB kneels. It could knock a team's rushing stats down from 4.0 to 3.7 by artificial means. Just have an asterisk designation ... "Three kneels for minus three yards, not to be included in the official statistics."This of course is a good point, but sample size more than accounts for this. Not that I would expect Dr. Z to know anything about sample size, it's not like that concept has changed in the last 35 years, so it must be garbage.

The small sample stats would be a lot more accurate if kneels weren't included in the stats, but it really is much ado about nothing.

I screamed about spikes being scored as incompletions thereby penalizing the QBs from bad teams, which always were catching up, hence spiking the ball. Why should they influence a passer's accuracy?

Same deal here. Dr. Z is right that the short term stats would be more representitive of the job a guy did if spikes didn't count as incompletions, but like the above example:

1) Independant Organizations such as Football Outsiders have already created stats (See: DVOA) that measure more accurately than the basic stats.
2) Sample size all but eliminates rare plays such as spikes and kneels.

In Dr. Z's defense, his beef seems to be against the Elias Sports Bureau, but that doesn't give him a right to project his beef to any and all objective evidence.

As of this week, all the ranked quarterbacks in the league average 63.3 percent completions. In 1972, the year that keyed the standards put in, that was a stunning statistic. Only one passer even topped 60 percent, Norm Snead of the Giants at 60.3. A mark of 63.8 percent would have gotten you a rating in that category of 110.9, a Pro Bowl number. Guess what? It still does today. In other words, average equals excellent.

Okay. This is where Dr. Z culminates his pretty solid argument by drawing one of the most asinine conclusions in the history of logic. Outside of his complete and utter ignorance of era adjustments, he makes the blanket assumption that all "stat geeks" feel that there is a hard number that QBs must stay above to reach certain levels of achievement. This is every bit as stupid as arguing that a RB who gains 99 yards in a game had a comparable game to a RB that gains 101 yards under similar conditions, and then bitching out the guy standing next to him under the assumption that he might disagree with this logic.

Obviously a QB rating of 90 was more impressive in 1958 than it is now. Any stat relies on realistic interpretation to give it value.

And here's where the shit hits the fan:

Chad Pennington: 111.2
Jeff Garcia: 110.7
Ben Roethlisberger: 108.0
Byron Leftwich: 97.2
Sage Rosenfels: 91.4
Donovan McNabb: 91.1

In each case, the passer with those gaudy numbers lost ... repeat: lost the game. And yet many people rely on them to judge the quarterbacks. A safety-first mentality has been created. Throw the 8-yard checkdown on third-and-12; it'll work wonders for the rating chart. Avoid interceptions at all cost, don't be bold, take care. Remember, your contract is tied to it.

I would be willing to bet Dr. Z's house and his wife that QB Rating correlates STRONGLY to winning percentage. Thanks for giving us a sample of one week and trying to make a conclusive argument though.

For the record: 8 yards on 3-12 is significantly better than an interception, and measurably better than an incompletion. If it was simple to convert on third and long, this discussion wouldn't be necessary.

Also, avoiding interceptions=generally a good thing.

Anyway, that's whats wrong with the system, per Dr. Z. He doesn't understand it, and you nobodies who happen to understand can you sleep at night!

No comments: