Benny Friedman NPR, and a word on statistical completeness

JameisLoseston
Posts: 391
Joined: Wed Oct 16, 2019 12:39 am

Benny Friedman NPR, and a word on statistical completeness

Post by JameisLoseston »

I recently picked up my copy of Rupert Patrick's excellent A Statistical History of Pro Football. Being quite a fan of his Normalized Passer Rating system, and since his work begins with the 1932 NFL season, I decided to calculate it for pro football's first undisputed GOAT, Benny Friedman, for each season of his career and his career as a whole. All this only took me a day to do, but I hope it is sufficiently comprehensive; get ready for a lot of numbers and talk about what they mean. I presume this information will prove useful to the early NFL enthusiasts here.

For those who don't have Rupert's book, first of all, you should, and second, Normalized Passer Rating (NPR) is an era-adjusted metric that modifies the classic passer rating formula to compare QBs against league averages in specific years, rather than a constant baseline which has become obsolete. Therefore, NPR balances QBs from all eras relatively equally, instead of heavily favoring modern passers. It does tend to favor great QBs of earlier eras, and I'll explain why shortly; for this reason, I came in expecting Friedman to land at number one all-time.

Stats used to calculate these ratings were obtained from jt-sw.com. They are to be considered incomplete for the majority of these seasons, but I will address this issue extensively in this post.

1927: 101.6
1928: 106.71
1929: 143.98
1930: 137.85
1931: 123.3
1932: 60.32
1933: 123.27
Career: 121.95

All of the figures used for calculation can be seen on this spreadsheet:
https://docs.google.com/spreadsheets/d/ ... p=drivesdk

So, according to NPR, this makes Benny Friedman by far the most efficient quarterback of all time relative to his contemporaries. He's actually the only passer rated over 100 in this system; the #1 career NPR since 1932 is Otto Graham (96.1), or Sid Luckman (94.4) if you prefer to isolate Graham's NFL stats. He is so far ahead that his entire career would be the third-highest rated season in history, behind 1943 and 1941 Luckman. On that note, Friedman's best two seasons replace Luckman's best two on top of that leaderboard, although by not as enormous a margin as he leads the career ranking. 1932 was his only season below 100 (I didn't bother doing 1934 separately because he barely played, but it's in his career stats), but I was definitely surprised to see 1933 rate above 1927 and 1928.

Those who understand how era-adjusted passing metrics work will likely not be altogether shocked by much of this, because speak of the devil, the Luckman Effect is, of course, very much operative in all of it. That is, in earlier and smaller eras of the NFL, years with one or two excellent QBs and otherwise mostly terrible ones were frequent, and the abysmal league averages in those years tend to artificially inflate the NPR of anyone who's any good at all. Friedman, even more than Luckman, was a very big fish in a league full of very small fish at QB, heavily inflating the distance between him and the league average. Luckman and Baugh at least had each other in the league to keep the average somewhat afloat, but Friedman was truly the only game in town, and the average without him was utterly destitute. So that is one reason why it is hard to take NPR at face value in this instance. I expected to see Friedman land at a comfortable #1 for this very reason, but I have to admit the magnitude of the result did surprise me. I can't say it changes my perspective on him, though, given that I already considered him a legend.

It is also unclear how much of Friedman's recorded stats are incomplete, which obviously complicates any attempt to calculate ratio and efficiency stats for such players. However, we can estimate this factor by comparing the leaguewide stats in the unofficial years with those of the official years, and interpolating the difference to just Friedman. So I decided to compare the total numbers for 1927-31 to those in 1932-34, to estimate how much is missing from the former. Per the method for deriving NPR, Friedman's output was removed originally from the league stats for 1927-34, so he has been added back in for this exercise.

However, there is a clear problem here: teams often played different amounts of games during this era of football, and there were also different numbers of teams in the league each year. So I painstakingly counted up every individual game played between two NFL teams for these years, and multiplied the result by two to get an accurate count of "team-games" played per season; the figures are listed in the spreadsheet, under the games played column. With this, we can estimate how incomplete 1927-31 stats are, by noting how much lower the per-game passing stats are for those years compared to 1932-34, assuming per-game passing volume did not change notably over this span. Here is what I got:

Completions: 4.18 C/G 1927-31 (662 games), 4.35 C/G 1932-34 (330 games). Therefore, about 0.17 completions per game, or around 110 total leaguewide, are missing from the record. Friedman accounted for about 13.4% of all completions in this time, so we're missing maybe 15 of his completions, total. Insignificant.

Attempts: 10.48 A/G 27-31, 12.91 A/G 32-34. This is very significant. We're missing around 2.5 attempts per game from 1927-31, around 1600 total - and almost all of them are incompletions. This makes sense, because complete passes would be much more likely to be noted in post-game recaps. Friedman accounted for about 10% of total pass attempts in these years, so we're missing about 160 of his attempts, 145 of which should be incomplete. His true Comp% is not as high as shown. Adding 15 completions and 145 incompletions to his stats gives a career percentage of 44.6 instead of 50.3.

Yards: 67.30 Y/G 27-31, 63.68 Y/G 32-34. Not significant in terms of incompleteness, which is expected considering the minuscule difference in completions. The weirdness of the result seems to be due to a passing volume drop in 1932 specifically. Impossible to estimate how many yards Friedman is missing, but it ain't much. If we multiply his career Y/C by the 15 or so lost completions, it's only about 260 more yards. His Y/A, however, will be lower due to the extra incompletions; 7.66 instead of 8.74, specifically, quite a big difference.

Pass TDs: 0.52 TD/G 27-31, 0.47 TD/G 32-34. One would expect TD data to be mostly complete, and this appears to be the case; they are the only stats considered official by PFR for Friedman, so I think we can assume they aren't missing anything. Again, 1932 is to blame for the discrepancy, as is Friedman running up 20 TDs in 1929, which was about 30% of the league total that year (Ruthian indeed).

Interceptions: 1.39 INT/G 27-31, 1.65 INT/G 32-34. Again significant. Unless QBs were just better in 27-31 (hint: they weren't, besides Friedman), we're missing about one INT per 4 team-games, or 165 INTs leaguewide over these 5 years. Friedman was a much better QB than the rest of these JAGsters, however, and only accounted for about 5.6% of all picks from 27-31, so only about 9 of those missing ones should be his. His career INT% actually decreases from 7.77 to 7.55, accounting for the missing pass attempts.

So overall, one may expect the NPR calculations for Friedman above to decrease as more data is uncovered, but not to a particularly large extent, especially since all of the adjustments I just applied to Friedman also affect the league as a whole. Sure enough, filling in the estimated incomplete stats for both Friedman, and the league minus Friedman, gives him an NPR of 119.17, a very small decrease. Either way, by all educated guesses, the 1927-31 pasaing stats are somewhere around 90 percent recorded, so the provided NPR figures for Friedman are very close to accurate. Oh yeah, dude could run, too.

I hope you enjoyed, and that you gained a new appreciation of Friedman, if you didn't already have one. So what did this exercise convince you more of: that Tom Brady couldn't carry Benny Friedman's cleats, or that NPR is ill-equipped to analyze him due to the horrible league context he played in? Discuss away.
User avatar
TanksAndSpartans
Posts: 1153
Joined: Wed Nov 18, 2015 1:05 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by TanksAndSpartans »

Good post! You should put this together into an article. Rupert's book is also on my list to pick up.

For this part:
JameisLoseston wrote:It is also unclear how much of Friedman's recorded stats are incomplete
We actually know - the origin of those stats isn't here:

http://www.jt-sw.com/football/pro/stats.nsf

It's David Neft who compiled them from newspaper accounts (In my opinion, the stats really shouldn't have been posted online without citing the source). You have to be careful with the online version of the stats too, there are some really bad typos. The original work can be found in several books and it calls out the number of complete and incomplete games clearly.
Bob Gill
Posts: 559
Joined: Wed Oct 08, 2014 7:16 pm

Re: Benny Friedman NPR, and a word on statistical completene

Post by Bob Gill »

TanksAndSpartans is absolutely right about the source of those stats for Friedman. From 1927-31, his peak seasons, Neft found "complete" stats for 42 of Friedman's games, and in those he connected on 291 of 610 passes for 5,038 yards and 40 touchdowns. In his other 19 games in those seasons period, for which Neft retrieved partial stats from game stories that detail touchdown drives and other key plays, Friedman completed at least another 79 passes for 1,683 yards and 18 more TDs.

Note that he completed 47.7 percent of his passes in those complete games. Since these games are pretty much a random selection, we might assume that's an accurate overall figure for those years as a whole. (That web site's inclusion of "incomplete" games gives him an unmerited boost of 3 percent.) Also, in the complete games he averaged about 7 completions, but in the incomplete games he averages only 4, so we might assume he's missing 50 or 60 others in those games.

Of course those are just assumptions, but since his TD passes came at almost identical rates in the complete and incomplete games (40 in 42 games; 18 in 19 games) they seem reasonable to me.
JameisLoseston
Posts: 391
Joined: Wed Oct 16, 2019 12:39 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by JameisLoseston »

Good post! You should put this together into an article. Rupert's book is also on my list to pick up.
Do you mean the Coffin Corner? I did happen to wonder if this kind of research would be worthy of publication. What is the process for submitting work to be considered?

I knew about the typos, like 703 carries for Tex Hamer (should be 103), and one guy has 111 interceptions instead of 11. There aren't any for Friedman, though, and I culled them out of the league totals. Also remember hearing that these stats are derived from Neft's research, but didn't know there was so much detail about completeness in the primary source, and thank you to Bob for filling in the blanks; seems my statistical estimations were pretty close to reality, and about 85-90% of the total stats are recorded. Actually, those estimations of the missing numbers might even be more charitable to Friedman than mine were. I'd love to hear his per-game averages in the complete vs. incomplete games for all of the following: completions, attempts, yards, and INTs (TDs is the same, as Bob noted).
Brian wolf
Posts: 2981
Joined: Wed Nov 27, 2019 12:43 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by Brian wolf »

Friedman seems to be the Joe Namath of his time. I wonder how effective he was using his looks, ability, and ability to generate controversy into affective endorsements, especially with pro football trying to become more accepted by the public ?
User avatar
TanksAndSpartans
Posts: 1153
Joined: Wed Nov 18, 2015 1:05 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by TanksAndSpartans »

JameisLoseston wrote:Do you mean the Coffin Corner?
Yep - it looks like a good project to me - Friedman never seems to be in the conversation with the great QBs. I would check his bio by Murray Greenberg, check the article Bob wrote in Vol. 14, No. 2 (1992) - Statistical Leaders of the '20s, and see if his name shows up in other CC articles. I like to review what's already been said before I start - sometimes what others have said support your argument and also to site them if they already said part of what your saying. Once your happy with it, email it over: https://www.profootballresearchers.org/ ... orner.html
JameisLoseston
Posts: 391
Joined: Wed Oct 16, 2019 12:39 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by JameisLoseston »

Another question I would need to know before I try turning this into a CC article: were there any changes to the ball, the rules, or any other aspect of how the game was played, that would have made passing more efficient between the late 20s when Friedman was in his prime, and the early 40s when Baugh and Luckman came into vogue? The leaguewide baseline had undeniably improved by the 40s, but there were also still plenty of bums who were just as horrible as most of Friedman's peers (mostly on the Cardinals, lol). So I'm not sure whether Friedman should be portrayed as a relatively equal passer to Baugh and Luckman who just happened to precede them and therefore lacked quality competition, or considered as something greater due to any potentially intervening circumstances that may have facilitated the advances Baugh and Luckman made, conveniences that Friedman would have lacked. Such context would allow me to make one case or the other much more convincingly.

Honestly, just browsing PFR, it seems to me like something must have changed, because it wasn't just Baugh, Luckman, and Isbell putting up Friedmanesque numbers by then. You had players like Ace Parker, Tommy Thompson, Frank Filchock some of the time, even guys no one's really heard of anymore like Parker Hall and Charlie O'Rourke, all who could really give it a spin. Even Tony Canadeo, a halfback, was able to step in and provide something of substance when Isbell bailed (to be fair, he did have Don Hutson to throw to). You still had your Whizzer Whites who never should have been allowed to touch a football with the intention of throwing it, but they were fewer and farther between. Meanwhile, in Friedman's time, Jack McBride and Red Dunn were inconsistent and decent at best, and that was literally it. Otherwise, you had the likes of Curly Lambeau and Ernie Nevers, who were halfbacks trying to pass and played like halfbacks trying to pass; a couple more guys had a promising season, but then spontaneously combusted and disappeared. That feels like too big a difference to come down to mere coincidence or teams simply not having learned how to throw a football yet; something had to have happened to make it easier. I'd love to know what exactly.
User avatar
TanksAndSpartans
Posts: 1153
Joined: Wed Nov 18, 2015 1:05 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by TanksAndSpartans »

Hopefully someone can help with specifics, as I'm no expert on the evolution of the football itself or rules, but the ball is something Murray talks a lot about in Friedman's bio. I recall it being described as big, fat, rugby-like, etc. I think by '37, Baugh would have had a ball to throw that was more conducive to passing.

There were rule changes too. Friedman would have had to be at least 5 yards behind the line of scrimmage. I believe incomplete passes in the end zone were essentially turnovers. Probably a few others.

I think Nevers had a 300 yard passing game along the way. Johnny Blood felt Dunn was as good of a passer as Friedman.
Bob Gill
Posts: 559
Joined: Wed Oct 08, 2014 7:16 pm

Re: Benny Friedman NPR, and a word on statistical completene

Post by Bob Gill »

The ball did get slimmer between 1927 and 1940, maybe in a couple of stages. (By 1927 it may have already been slimmer than in Jim Thorpe's day.) I think the ball had reached its current shape by 1936, maybe 1933.

The rule that said a passer had to be five yards behind the line of scrimmage was abolished in 1933, and the creation of hash marks that same year also boosted offense, rushing AND passing. (Before that, whenever the ball went out of bounds it was spotted a yard from the sideline, which severely limited the options for play-calling, so the offense usually had to waste a play just to get it back toward the middle of the field.)

So Baugh and the others who followed close behind him (or went just before him, like Arnie Herber and Ed Danowski) had those advantages over the passers of Friedman's day. Also, don't just dismiss this idea from your post: "That feels like too big a difference to come down to mere coincidence or teams simply not having learned how to throw a football yet." I think to some degree it was exactly that. In Friedman's day he was basically the only one who could complete passes consistently enough to move the ball without much help from a running game. Other teams were no doubt impressed by the things he did, but it just didn't make any sense for them to adopt the same strategy.

By the time Baugh came along, though, that had changed. Though he was the best passer of his time, there were a number of others -- you named most of them -- who were close enough in ability that it made sense for other teams to copy what the Redskins were doing. Within five years of Baugh's debut, half the teams in the NFL had a passer who could carry their offense on a good day. I think it was one of the crucial periods in the development of the pro game.

Come to think of it, I wrote an article for the Coffin Corner on this very subject, back around 2009. I don't remember the headline, or title, but I suspect you can find it on the site. I'm sure it has more details than what I remembered here.
JameisLoseston
Posts: 391
Joined: Wed Oct 16, 2019 12:39 am

Re: Benny Friedman NPR, and a word on statistical completene

Post by JameisLoseston »

TanksAndSpartans wrote:Hopefully someone can help with specifics, as I'm no expert on the evolution of the football itself or rules, but the ball is something Murray talks a lot about in Friedman's bio. I recall it being described as big, fat, rugby-like, etc. I think by '37, Baugh would have had a ball to throw that was more conducive to passing.

There were rule changes too. Friedman would have had to be at least 5 yards behind the line of scrimmage. I believe incomplete passes in the end zone were essentially turnovers. Probably a few others.

I think Nevers had a 300 yard passing game along the way. Johnny Blood felt Dunn was as good of a passer as Friedman.
Whoa, those are actually some amazing facts! I knew something had to be amiss with how much passing had improved by 1940, but I had no idea the disadvantages were that severe in Friedman's time. These are the kind of things that can go beyond even what stats can say, and truly change my outlook on a player. Friedman, as they say in modern NFL parlance, was just built different.

Nevers and Lambeau, for a short time, were interesting early case studies in volume passing; they were essentially the progenitors of the "f**k-it-chuck-it" play style that gave rise to Davey O'Brien, eventually George Blanda, and endures today in seasons like Jameis's 30-30. Clearly, those teams were willing to tolerate the negative returns in the name of experimentation.

I think Dunn was about as good as Arnie Herber, who immediately succeeded him as Packer QB. Maybe better, since he had the same disadvantages as Friedman and didn't have Don Hutson for half his career. He's a great HOVG candidate.
In Friedman's day he was basically the only one who could complete passes consistently enough to move the ball without much help from a running game. Other teams were no doubt impressed by the things he did, but it just didn't make any sense for them to adopt the same strategy.
It's observations like this that make me feel like Friedman was really just unbelievable, even more than I already did. I've always said that there are two distinct, in many ways opposite, measures of greatness. One of them is to say that a player "changed the game" - he started something big, and inspired other teams and players to emulate and one day surpass him. Don Hutson is the perfect example of this, and Deacon Jones, etc. But the failure to change the game is an equal, perhaps even more defining barometer: when a player is unable to leave such a legacy, no matter how hard they try, because they were just too special. You have to look outside of football to find quintessential examples of this; Wilt Chamberlain and Wayne Gretzky are probably the clearest ones. They affected their sports in various ways, but no one ever learned how to do it like they did.

Friedman seems to fall square in the middle of these two extremes. On the one hand, the passing game opened up considerably shortly after his career, and his records fell within 15 years. His achievements probably had something to do with that forward progress. On the other hand, it apparently required the NFL to completely overhaul its approach to the pass play, on both the league and team levels, for that to happen; if the rules Friedman played under had remained in effect, it may have taken decades, or maybe never happened at all. Honestly, the kind of rule changes mentioned here seem even more drastic than, say, what we saw in the late 1970s and early 2010s that led to those respective passing booms. Of course, the NFL may very well not have existed for much longer if it hadn't made those early adaptations, too...
Post Reply