Pro Football Reference Blog

Ha, ironic this comes up just after the e-mail I sent about something I've been working on for opponent adjustments for passing numbers.

I plugged that Namath game in, and wow, that blows away anything I've seen so far. Of course I'm working on a limited sample and don't have 40 years to go off of. The highest game I've previously seen was Brady's game in Miami in 2007. You have it ranked 45th here.

The one that jumps out at me is Craig Morton at No. 4. Five touchdowns and 349 yards -- on only 17 attempts? Only four incompletions?

Then I looked up the box score:

http://www.pro-football-reference.com/boxscores/197012200dal.htm

The real star wasn't Morton, it was Bob Hayes. Six catches, 187 yards, 31.2 yards per catch, and four touchdowns (of 38, 38, 15, and 59 yards). The Cowboys threw seven incompletions on the day (Roger Staubach went 2-for-5). If every single incomplete pass was thrown Hayes' way, he still averaged 14.4 yards per target. No idea what opponent adjustments would do (Houston won only three games that year) but still...

One thing era-adjusting does wrongfully--back in the 60's & 70's, teams threw less frequently, leading to smaller sample sizes for passing--which hyper-inflates some games (Namath's was ridiculous in whatever era you put it in). They also threw higher risk and longer passes, which means you have a great chance for a GREAT day--or a HORRIBLE one. Teams/QB's today use short passes like the 60's/70's teams used the run--with much better results.
Once FO gets more game logs into their system (they go backward about 1 year per calendar year), it will be interesting if these numbers hold up. Not doubting your regression, Neil, but FO gives bonus YAR for FIRST DOWNS--meaning more completions, probably more first downs--and more "bonus" YAR. They also give "bonus" YAR for TD's--which are more prevalent now. BTW, I expect some of those games in your top group to go down because of this--as well as the penalty for incompletes/interceptions.

Dear Neil,

We already know that your method of normalizing stats to the modern era skews those stats heavily in favor of the old-time QBs. You've described your own method as a "monstrosity"---which is why I don't know why you continue to use it time and again.
The links that you've pointed us in Football Outsiders clearly state that DYAR (ditto for YAR) cannot be figured out prior to 1993 because it requires knowing the result of every play from the games. You claim that you have come up with a formula for overcoming this problem, but I don't see Football Outsiders or anyone else pounding on your door in an effort to obtain this formula which would revolutionize football statistics and make you a very rich man.
All of this aside, if I'm just an ordinary fan trying to interpret your chart (which I understand that you cannot be blamed for as it is a product of FO), there is a part of it that I just don't understand. Let's take a post-1993 game---No. 12---Peyton Manning's 1-4-2004 game vs. Denver. Manning had 377 passing yards in the game. You and/or FO give him credit for 354 YAR. What this is saying to me (the ordinary fan) is that Manning's replacement(Jay Sorgi or whoever) would have had only 23 yards passing? It doesn't make any sense--- seeing as FO tell us that an average replacement QB is only 12.5% less effective (valuable) than an average starting QB in the NFL---which means that whoever would be the replacement for Manning in this game would be expected to throw for around 330 yards (if both Manning and his replacement were "average"). Now, admittedly, Manning is much better than an average starting QB and I don't know how either you or FO gage abstract factors like "effectiveness" or "value" (I have to guess here) nor do I know what types of bonuses or penalties are given out for TD passes, INTs, etc., by FO, but 23 yards for a replacement QB just doesn't add up no matter how you dice it. To me, Manning (if he was "average") should receive about 47 YAR for his effort---and that all depends on whether or not FO's 12.5% figure is correct---which we can't tell because it is arbitrary.
Any of us can and often do make mistakes. But to continue to make the same mistake over and over again---after admitting your mistake---then it becomes a "monstrosity".

Adjusting for era seems pretty useless for an individual game. Who cares what passing yards were worth in the 1970's as a whole when looking at one game in which the ball was simply thrown all over the place? I realize that we should account for significant rules changes that impacted passing, but such efforts seem sloppy at best when applied to individual games.

Something I think needs no adjusting for anything (at least for 1970-present) would be the QB's rushing numbers. Most QB runs are scrambles on broken pass plays. Then you have the QB sneak, which is a minimal gain play. The scrambling QBs typically have pretty high career YPC's, which would be even higher if not for kneel down plays.

So if a guy like Roger Staubach has 5 rushes for 30 yards in the 70's, I think that's exactly the same as a guy like Aaron Rodgers having 5 rushes for 30 yards in 2009. I wouldn't bother adjusting any of that.

Reply to No. 12

Dear Neil,
I wouldn't make up such jokes if I were you. I took your words very seriously---because I took your "100 Best QBs" srticle very seriously and based much of my own research on "wins above average" on your results which I found fascinating although I little understood or agreed with your math. I have quoted and praised your results many times and you can imagine how I felt when you told us that your research was a "flawed monstrosity" which, in turn, made my own research results flawed.
That said, your chart, or similar ones found in FO, doesn't mention anything about "value in field position"---it just mentions YAR. The links that you provide don't mention where or how 354 YAR for Manning comes from. It is just a figure standing there alone and the hapless reader is left to his own thoughts trying to figure it out. Furthermore, as I understand the concept of "value" as used by FO, they include it in a different metric---DVOA or VOA---which is the metric that they use in place of passer rating.
I saw Namath's game vs. Unitas in 1972. Neither team had any defense and the receivers were running wild. It was a very enjoyable, much remembered, but meaningless game. It doesn't have to be hyped up with a lot of nonscence about replacement level stats. I don't know why FO tries to impose this type of metric upon us fans as no one understands, yet alone cares, about some mythical replacement level QB. Why don't we use real stats in these studies?

Why not make it simple, and use the same formula you applied in the GQBOAT series? For each game, it would be (AY/A - LgAvg AY/A)*attempts.

For Namath's game, it would be (20.39 - 5.3)*28 = 422.6 yards over average.

All of these performances (#s 1-100) were indeed ones for the ages. But it would be nice to be able to discern between games that were blowouts versus games that were in doubt for most of the contest, just for curiosity’s sake. Dick Shiner’s performance (#87) in 1973 came in a 62-7 rout over New Orleans. Similarly, Fran Tarkenton’s 18 of 24 day in 1974 (#77) came in a 51-10 thrashing over a Houston team that was so bad the game was never televised.

But Tarkenton had, in my judgment, an even better day (#80) against Washington in 1975 in a seesaw battle, a 31-30 loss coming on the last play of the game (a missed FG by Fred Cox). His numbers in that game were impressive enough (27 of 37 for 357 yards), but given that the game was such a nail biter I look back on that outing as more impressive when factoring in “pressure situations.”

Also, in adjusting for opponents (or top-ranked defenses) I fear that it might overinflate someone’s stats or severely deflate another’s. The 1972 game between the Rams and Vikings had both defenses ranked at the top of the NFC prior to their Week 10 game, a 45-41 “javelin contest” between Tarkenton and Roman Gabriel. But there were so many blown assignments on defense for both clubs that day and sometimes that happens in football – and thus enables some of these outings to make the list. The Namath game against Baltimore in ’72 was extraordinary no matter how you slice it. But it would be nice to put some context to some of the one-sided affairs.

By the way, the Mark Rypien game (#3) has got to be an all-time high for any fantasy football league.

Jurgensen and Unitas are clearly affected by the 1970 cutoff date. By that point, both were already on the downside of their careers, and while both started for almost all of the 1970 season, neither of them started more than five games in a year after that. If the years prior to the merger were included, I'd think they'd both appear on the list at least once (Both have a few games in the 60s that look like they'd fit on the list, and Unitas may have one or two back in the 50s as well).

However, the lack of Elway is somewhat suprising, as is Montana only appearing once. I'd have thought his 6 TD, 472 yard game vs the Falcons in 1990 would have at least cracked the Top 100.

"However, the lack of Elway is somewhat suprising"

I actually don't think that's surprising at all. Elway was a great quarterback, but a mediocre PASSER. His career ANY/A+ was only 106, slightly above league average. He only had one 4000 yard season and never threw more than 27 TD's in a season.

I did a little homework to answer my question above about where Jim Kelly fits in among Hall of Fame quarterbacks. By my count, there are 25 Hall of Famers who were primarily quarterbacks. There were a handful more who played numerous positions, including quarterback. I did my best to isolate the "just quarterbacks" from the "also quarterbacks".

For each of the components that makes up the passer rating (completion percentage, yards per attempt, touchdown percentage, and interception percentage), I ranked each of these 25 HOF QBs from 1 (best) to 25 (worst).

Here were the results:

COMPLETION PERCENTAGE: 1.Young 2.Montana 3.Aikman 4.Kelly 5.Marino 6.Fouts 7.Moon 8.Starr 9.Dawson 10.Jurgenson 11.Tarkenton 12.Staubach 13.Elway 14.Baugh 15.Griese 16.Graham 17.Tittle 18.Unitas 19.VanBrocklin 20.Bradshaw 21.Luckman 22.Waterfield 23.Namath 24.Layne 25.Blanda

YARDS PER ATTEMPT: 1.Graham 2.Luckman 3.VanBrocklin 4.Young 5.Starr 6.Unitas 7.Fouts 8.Dawson 9.Staubach 10.Jurgenson 11.Tittle 12.Montana 13.Kelly 14.Namath 15.Marino 16.Waterfield 17.Griese 18.Baugh 19.Tarkenton 20.Layne 21.Moon 22.Bradshaw 23.Elway 24.Aikman 25.Blanda

TOUCHDOWN PERCENTAGE: 1.Luckman 2.Graham 3.Dawson 4.Baugh 5.Waterfield 6.Jurgenson 7.VanBrocklin 8.Blanda 9.Griese 10.Unitas 11.Young 12.Tittle 13.Bradshaw 14.Layne 15.Tarkenton 16.Staubach 17.Montana 18.Marino 19.Kelly 20.Starr 21.Namath 22.Fouts 23.Moon 24.Elway 25.Aikman

INTERCEPTION PERCENTAGE (low to high): 1.Montana 2.Young 3.Aikman 4.Marino 5.Elway 6.Moon 7.Kelly 8.Staubach 9.Tarkenton 10.Fouts 11.Starr 12.Jurgenson 13.Unitas 14.Dawson 15.Griese 16.Graham 17.Bradshaw 18.Tittle 19.Namath 20.VanBrocklin 21.Layne 22.Baugh 23.Blanda 24.Luckman 25.Waterfield

TOTAL OF THE FOUR RANKINGS: 1.Young(18) 2.Montana(32) 3.Dawson(34) 4.Graham(35) 5.Jurgenson(38) 6.Marino(42) 7.Kelly(43) 8.Starr(44) 9.Staubach(45) 10.Fouts(45) 11.Unitas(47) 12.Luckman(48) 13.VanBrocklin(49) 14.Tarkenton(54) 15.Aikman(55) 16.Griese(56) 17.Moon(57) 18.Tittle(58) 19.Baugh(58) 20.Elway(65) 21.Waterfield(68) 22.Bradshaw(72) 23.Namath(77) 24.Layne(79) 25.Blanda(81)

Note that I used regular season stats for this study.

While this is a fairly quick and dirty methodology, a safe conclusion that I drew from the results surprised me a little. The conclusion is that Jim Kelly actually belongs in the Hall of Fame, at least insomuch as his numbers compared favorably with his bronze busted compatriots.

To answer my question from before, based on this data, Kelly would probably start over a large percentage of other HOF QBs. Very interesting.

I'm sure Namath would have a few contend for the worst game as well.

How about this one on 10/19/75 vs. MIA?

8/24 for 96 yards, 0 TDs, 6 INTs

Michael vick has just earned the number one spot

I worked Vicks numbers from last night to be 307.15 as well. Believe it speaks to a flaw in the ERYAR, because of the following reasoning. A Qb is not a RB and many Qb rush attempts are the result of broken plays, coverage, et cetera. If the QB decides Running the ball is the best option, then the formula should simply place the same weight on every element, counting the play as a completion, and substituting fumbles for interceptions. Otherwise the formula would have to count loss yardage on sacks as running plays with negative resulting yards, because the alternative to being tackled while holding the ball in your hand is apparently giving the ball to the running back. This is lazy and unimpressive. I am a NY Giants fan and I believe that on a properly weighted basis that Vick had one of the 10 best games ever by a QB last night; it was totally cool to watch