SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all PFR content, so you can quickly and easily find the content you want.
Also, our existing PFR blog rss feed will be redirected to the new site's feed.
Pro-Football-Reference.com » Sports Reference
For more from Chase and Jason, check out their work at Football Perspective and The Big Lead.
ShareThisI shouldn’t post this: Approximate value top 200
So I finally wrote down all the details of the approximate value (AV) method. You can find them at this ever-evolving page.
Sometimes, for example if you want to assess a trade or determine the top draft classes of all time, you need a metric that is capable of comparing players across positions and eras. In baseball and basketball, lots of stats have been cooked up to do this, and they can do so with a reasonable degree of precision. In football, no such stat exists. In most cases, people use "starter" or "number of years as a starter" or "number of pro bowls" as the metric when they have to compare across positions.
AV is intended to be an improvement over those metrics, and nothing more. It is not Not NOT an ubermetric whose purpose is to decide once and for all who the best players in NFL history were. So I should not Not NOT rank all the players in post-merger NFL history according to AV and present them in a list. To do so would be to imply that I've done something I haven't done. To do so would be to invite links from team message boards where the first post is "some guy posted his list of the best players of all time" and the second post is "I stopped reading as soon as I saw _______ ranked ahead of _________. What a bozo!" And I have to admit that I'm overly thin-skinned about that sort of thing.
But I'm going to do it anyway, for a couple of reasons. First, it's May. I don't think that one requires any more explanation. Second, I'd like feedback on groups of players that people think are over- or under-valued (by the way, allow me to agree in advance that mediocre quarterbacks with long careers are overvalued and that truly great cornerbacks are undervalued --- I'm just not sure right now how to fix those problems). This post will probably lead to some improvements in the system. Third, I've learned a lot from tracing down the reasons why some guy I've barely heard of is rated so high. You can too. Follow the links. Even if you ultimately don't agree with the ranking, you might learn that some players were better than you remember them, and others weren't as good.
Now let me explain exactly what the list is. Remember that AV puts a number of every player-season. My opinion is that most people mentally rank players by counting all the players' seasons, but weighting their best seasons more. In order to mimic that, I've defined each player's approximate career "value" to be:
100% of his best season, plus 95% of his 2nd-best season, plus 90% of his 3rd-best season, plus, ....
So, for two players with the same career AV total, the one with the higher peak will be rated a little higher. And junk seasons at the end of a player's career count for almost nothing.
Players are only eligible for the list if their career started in 1970 or later. Here it is:
1 Reggie White 165 2 Jerry Rice 153 3 Dan Marino 152 4 Brett Favre 149 5 Bruce Smith 144 6 Lawrence Taylor 142 7 John Elway 141 8 Peyton Manning 138
NOTE FOR PEOPLE WHO SKIPPED STRAIGHT DOWN TO THE LIST WITHOUT READING THE POST: the author of the algorithm that generated this list does not agree with the entirety of the list. He, in fact, acknowledges that it's obviously wrong in spots. He does think it's pretty respectable, given the difficulty of comparing players across positions and eras using only objective data. He also thinks it'll give us something to talk about here in May.
We now return to the list.
9 Derrick Brooks 137 10 Steve Young 133 Anthony Munoz 133 12 Junior Seau 131 13 Rod Woodson 129 14 Ray Lewis 128 15 Mike Singletary 127 16 Warren Moon 125 17 Emmitt Smith 124 Joe Montana 124 Dan Fouts 124 Marshall Faulk 124 21 Ken Anderson 123 22 Walter Payton 122 23 Randy White 120 24 Bruce Matthews 118 Michael Strahan 118 26 Marvin Harrison 117 Jack Ham 117 Jack Youngblood 117 29 Chris Doleman 116 30 Zach Thomas 115 31 Warren Sapp 114 32 Barry Sanders 113 33 Jack Lambert 112 34 Boomer Esiason 111 Rickey Jackson 111 36 Derrick Thomas 110 37 Terry Bradshaw 109 38 Drew Bledsoe 108 Jason Taylor 108 Isiah Robertson 108 41 Ronnie Lott 107 Terrell Owens 107 43 Randall Cunningham 106 Thurman Thomas 106 Randy Moss 106 Jim Kelly 106 47 John Randle 105 Michael Irvin 105 49 Steve McNair 104 Tim Brown 104 51 Mike Webster 103 Rich Gannon 103 53 Dan Hampton 102 Vinny Testaverde 102 Sam Mills 102 Troy Aikman 102 57 Mark Brunell 101 Tony Dorsett 101 Mel Blount 101 60 Steve Largent 100 Randall McDaniel 100 LaDainian Tomlinson 100 Edgerrin James 100 Dave Krieg 100 Hardy Nickerson 100 66 Pat Swilling 99 Too Tall Jones 99 Rod Smith 99 Cortez Kennedy 99 Shannon Sharpe 99 James Lofton 99 Howie Long 99 Kevin Greene 99 Marcus Allen 99 75 John Hannah 98 Karl Mecklenburg 98 Steve McMichael 98 78 Steve Grogan 97 Curtis Martin 97 Ken Stabler 97 Wilber Marshall 97 Trent Green 97 Willie Roaf 97 Greg Lloyd 97 Franco Harris 97 Cris Carter 97 87 Henry Ellard 96 Orlando Pace 96 Jimmy Smith 96 Gary Zimmerman 96 Clay Matthews 96 Jim Everett 96 Richard Dent 96 Andre Reed 96 95 Will Shields 95 Neil Smith 95 Ricky Watters 95 Tiki Barber 95 Art Monk 95 Tom Brady 95 John Lynch 95 102 Richmond Webb 94 Brian Urlacher 94 Andre Tippett 94 Mike Kenn 94 Lomas Brown 94 Isaac Bruce 94 108 Eric Dickerson 93 Mo Lewis 93 Phil Simms 93 Harry Carson 93 112 Joe Theismann 92 Deion Sanders 92 Ted Washington 92 Randy Gradishar 92 Tony Gonzalez 92 Darrell Green 92 118 Donnie Shell 91 Mike Haynes 91 120 Clyde Simmons 90 Bryant Young 90 Ray Childress 90 Torry Holt 90 Dave Butz 90 125 Tom Jackson 89 Eric Allen 89 Bob Baumhower 89 Lyle Alzado 89 Jim Plunkett 89 130 Donovan McNabb 88 Aeneas Williams 88 Charles Haley 88 Cornelius Bennett 88 Roger Craig 88 Jack Reynolds 88 136 Simeon Rice 87 La'Roi Glover 87 Warrick Dunn 87 Daunte Culpepper 87 Lydell Mitchell 87 141 Cliff Branch 86 Ronde Barber 86 Leslie O'Neal 86 Tom Nalen 86 Gary Clark 86 Kevin Carter 86 Henry Thomas 86 John Riggins 86 Jeff Garcia 86 150 Joe Ferguson 85 Ken Norton 85 152 Doug Wilkerson 84 Tarik Glenn 84 Chad Brown 84 Louis Wright 84 Steve DeBerg 84 Jim Langer 84 Walter Jones 84 Bill Romanowski 84 Bert Jones 84 161 Michael Dean Perry 83 Stanley Morgan 83 LeRoy Butler 83 Mike Munchak 83 Jacob Green 83 Nat Moore 83 Jackie Slater 83 Steve Wisniewski 83 Jim Harbaugh 83 170 Brian Dawkins 82 Fred Taylor 82 Charles Mann 82 Ron Jaworski 82 Robert Brazile 82 Fred Smerlas 82 Levon Kirkland 82 John L. Williams 82 178 Jonathan Ogden 81 Keenan McCardell 81 Chester McGlockton 81 Willie Anderson 81 Lemar Parrish 81 Seth Joyner 81 Harvey Martin 81 185 Eugene Robinson 80 Larry Allen 80 Art Still 80 Irving Fryar 80 James Brooks 80 Carnell Lake 80 Bob Kuechenberg 80 Rob Burnett 80 Vern Den Herder 80 194 Joe Jacoby 79 Dan Dierdorf 79 Joe Nash 79 Chris Chandler 79 Kerry Collins 79 Jake Plummer 79 Mark Gastineau 79 Brian Sipe 79 Wes Chandler 79
This entry was posted on Sunday, May 11th, 2008 at 7:26 pm and is filed under Approximate Value, General. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
You know, that looks like it worked out pretty well.
My initial observations, in addition to the ones you made in the prologue:
1) I think QBs with a high peak but modest tenure are undervalued here. Intuitively, that might be because a QB with an extremely high peak can more relatively valuable during those seasons, because he plays QB, than a player at another position who has a few outstanding seasons but little else.
2) I think LBs with long careers are overvalued. My guess at this is that voting for All-Pro and Pro Bowl teams at LB is more based off reputation than most other positions (except probably OL), so players continue receiving high AV points after they cease being great. This might not apply to a ridiculous player like LT, but I don't think players like Derrick Brooks, Jack Ham, and Zach Thomas belong in the same range as some of the skill position players around them, like Steve Young, Marshall Faulk, and Barry Sanders. Again, only my opinion.
One honest question: does the AV formula correct for the number of Pro Bowlers or All-pros as compared to the number of starters in the league at a given position? For example, each conference has 3 Pro Bowl RBs and 3 QBs out of 15 total starters at each position, so 20% make the Pro Bowl. Meanwhile, only 3 guards from each conference make the Pro Bowl out of 30 starters, so only 10% make it. I tried to figure this out from the AV formula page, but couldn't quite do it.
Anyway, I think this is a great effort and a useful tool - it actually seems to be much more "accurate" than I was expecting.
Seeing Warren Moon above Joe Montana makes me happy.
It is not that journeymen are overvalued in this system, it is that their value is spread out over too long of a time for them to have made a major impact in a given season. It seems to me that the best way to prevent mediocre quarterbacks with long careers is to use two stats:
1) Approximate Value
2) Average Approximate Value Per Year
The most valuable players (in both senses) will have high numbers for each stat.
That should read:
"It seems to me that the best way to prevent mediocre quarterbacks with long careers from being overrated is to use two stats:"
Whitedawg,
First note that the AV formula only uses all-pro and pro bowl information when it "has to." Pro bowl and all-pro info is not used at all for QBs, RBs, WRs, and TEs. And if you except those position, I think the ratios of leaguewide starters to pro bowlers are pretty similar.
Also, there are three levels of all-pro-ness: 1st-team all-pro, 2nd-team all-pro, and pro bowl. For the first two of those, the positional ratios should be just right. So I think (hope) that the issue you point out shouldn't be causing too much trouble.
whitedawg,
I'm also a little uneasy about the LBs. And a few defensive linemen (Chris Doleman in particular) turned up higher than I expected, too.
It looks like maybe I need to tweak the front-7/secondary ratio just a tad bit more toward the secondary.
I wonder if this system is somehow overvaluing modern players. Nobody in the top 16 played before 1980. Surely, one of the top 15 players since 1970 was active in the 70s.
I agree completely with the idea that the formula overvalues linebackers and undervalues cornerbacks, in general. Sam Mills over Mel Blount is an example illustrative of this bias. I think DB value should be adjusted significantly upward, mostly at the expense of LB value. DL value seems about fine.
Another really, really important one: It's undervaluing tackles. At the top, it looks OK because Munoz slips into the top ten (although I could easily argue he should be significantly higher.) But the real problem is the way things go after that. If I didn't miss anyone, then the next best tackle is Willie Roaf. That's a legitimate choice for the second-best tackle of all time, but if we think he's the second-best tackle of all time, he should be rated MUCH higher than 78th overall.
Another way to think of it: There are 25 tackles who have made the Hall of Fame. Quick, how many of those guys do you think should be on this list?
I'm betting you said something like ten or fifteen. But actually, right now you have four: Munoz, Zimmerman, Slater, and Dierdorf, who just barely sneaks onto the very end of the list.
I think most people would happily take Ron Yary or Art Shell over most of the players in the 150-200 range, without hesitation. In the other offensive positions, you're getting people like Keenan McCardell and Chris Chandler.
I should note that the system seems to like top interior linemen just fine. I think it probably needs to give much better bonuses to offensive tackles over the interior linemen, and it needs to credit the offensive line slightly more in general.
While not necessarily disagreeing with your overall point, there are only four tackles in the hall of fame who started their career after 1970. They are the four you listed above.
FWIW, Yary scores at 114 without his first two seasons. Those two years weren't much, but he'd probably be around 117 or 118.
I think TEs are under-valued. Three HoF TEs played their careers in the period under study: Caspar, Newsome, and Winslow. I'm not sure any of them qualify as "blocking" TEs either.
The problem may be that TEs have tended to have shorter careers, at least compared to elite offensive linemen - the three TEs above averaged 11 year careers while the HoF OL whose careers have taken place since 1970 have averaged 14 years. The only two TEs in your top 200 have already played 25 total seasons, indicating that longevity is rewarded a little too much for TEs (or shorter careers punished too much).
What happens if you tweak the TE multiplier up? Or keep it constant and give them all-pro points to TEs so elite TEs get a little more credit?
Definitely something worth looking into. It's not because the seasons were 14 games long in the 70s and 16 games long now, or anything like that. At the team level, I know the 70s aren't undervalued. They might be at the player level, but I am not sure what's causing it if they are.
It does seem odd. But no particular 70s players look undervalued to me*, and some --- Ken Anderson, Jack Youngblood, Isiah Robertson --- are higher than I think you'd see on subjective lists.
* unless you want to count Montana as a 70s player.
Also FWIW, Alan Page (who debuted in 1967) would be #7 or #8 on this list if he were eligible.
The one that jumped out at me was Eric Dickerson at 108 overall. Seems way too low. Especially when you see Jim Everett and Henry Ellard quite a bit higher.
In general, I think the top 10-20 looks real good and then things start to get a bit weird.
9: (Doug)
Fair enough, I wasn't thinking about the eligibility requirements when I made that post. Nonetheless, I encourage you to look at the general way a good tackle career is valued, both against other linemen and in general. I think there might be more guards on this list than tackles, total. After Munoz, the next few linemen are all interior.
A personal preference thing you might not agree with:
I think you should put a greater emphasis on yards per carry, to help people like Barry Sanders and Eric Dickerson, who most would say are too low right now. It also probably wouldn't hurt to put greater emphasis on adjusted yards per attempt or some other QB efficiency stat. That would help you knock down compilers like Bledsoe a little bit.
On the possible anti-1970s bias:
Maybe there just aren't any players debuting in the 1970s that should be in the top 15. The 70s is an unlucky decade, one that barely managed to miss out on Lawrence Taylor, Mike Singletary, Anthony Munoz, and the superstars of the 1983 draft. There are a number of good quarterbacks, but I certainly don't think anyone really loves Fouts enough to put him in the top 15.
I think Mel Blount, Joe Montana, and Walter Payton are the only three 1970s-drafted superstars you could really claim are undervalued.
Um, Drew Bledsoe is 38th and Tom Brady is 95th....that is really bad...and it has nothing to do with positions or eras...they are both QBs from the same era!
I know Andrew. Fred Taylor outranks Adrian Peterson, too. It's shameful.
Uh, seriously, Brady will probably pass Bledsoe by the end of the '08 season.
What's wrong with Fred Taylor ahead of Adrian Peterson?
Richie, I should learn to be less of a grouch. What I should have said was something like this:
It seems like great players on great teams are ranked too high compared to great players on bad teams. I think most people would agree that Payton was better than Smith, but Payton was great on some terrible teams, while Smith was great with other great players around him. It seems that there needs to be a stronger "overcame terrible teammates bonus."
And I agree with the others that linebackers are ranked too high. I'm a huge Bear's fan, and I have a hard time buying that Singletary was the 15th best player.
Seeing some of the players ranked below Steve Grogan is down right humorous.
I know the object is to make the number position independent so one can measure value across positions, but I'd be interested in lists broken down by position. Top 20 QB's, top 20 RB's, etc.
It'd also be interesting to see what the average value of each position is, and perhaps standard deviation... It might tell us which positions are likely have the greatest effect on a team's value.