THE BOOK--Playing The Percentages In Baseball

When I started reading that article, my first thought was “Aw, jeez.” Then I kept reading, saw what he was getting at, and realized he has some excellent points there.

It’s taken me a while to warm up to Poz, but the more I read of him, the more I like him.

I think many who read here might struggle with this group of seven 1B-OF-DH types.  Outstanding Offensive players, if you will.  Palmeiro, Bagwell, Walker, Martinez, Raines, Murphy, McGwire. 

name/win shares/rWAR/wOBA/MOPish?
Palmeiro/394/66.0/.380/No
Bagwell/387/79.9/.406/Yes
Walker/307/67.3/.414/Yes
Martinez/305/67.2/.405/No
Raines/390/64.6/.374/No
Murphy/294/44.2/.357/Yes
McGwire/342/63.1/.415/Yes

Now, I realize many folks don’t like some of those four things (some reject win shares, some reject MVP or MOP measures, etc.) But it’s worth noting that of this bunch only Bagwell and McGwire “nail” all four things.

It shouldn’t piss me off, but Trammell still not being in the HOF is very irritating.

You can’t show wOBA without adjusting for: (1) context and (2) differing career lengths.

I wonder if the rule would be a little different for pitchers.  Including offense, here are the 5+ WAR seasons from some recent pitchers, who most around feel here are solid “yes” to no-brainer guys:

Mussina: 6
Schilling: 6
Brown: 6
Glavine: 5
Cone: 5
Smoltz: 3

Even top- 10 all-timers Maddux and Johnson “only” have 10.

Is it easier for position players to accrue WAR during recent decades?  Maybe the fact that pitchers are not allowed to throw more than 240 innings or start more than 35 games anymore?

I also think POZ has the right idea. It always seemed to me that a player should have some pretty high peak value and also pretty high career value to get in. Maybe being an MVP caliber player at some point for at least 3 stratight years and having some kind of significant career accomplishment or milestone (like a certain number of hits or HRs). There might need to be some adjustment for position. It is probably hard for catchers to have high WAR seasons (or alot of them).

But what if we have a guy like Palmeiro. Never high peak value but who ranks extremely high in career WAR. Is that enough to overcome never being an MVP type of player? I would probably vote for him if he had not tested positive for PEDs.

Its funny now because I recall once arguing against him on Fanhome several years ago, saying something like POZ is now yet I was criticized for not recognizing Palmeiro’s overall career value.

I guess POZ is using WAR for everyone, including pitchers. Palmeiro does have 4 top 10 finishes if you just look at position players (none of them consecutive). But his career WAR, pitchers included, is 109th. Is that enough to overcome the “low” peak value?

Palmeiro does have 8 seasons of 4+ WAR and he had 3.6 in 1994. It seems like that would make 5 for a whole season. Then Palmeiro has 5+ seasons of 5 or more WAR and then he also has a 4.9. Pretty close.

Right, Palmeiro is right on the edge.

Historically speaking, the 50/50 point in WAR for HOF is around 55-60 wins.  That is, if you look at all players from 45 WAR to 70 WAR, you see that somewhere in the middle of that, you get as many HOFers as not.  Above 70 WAR, and they are virtually all in (sorry Lou Whitaker).  Below 40 WAR, and they are virtually all out.  Well, depending on the time period anyway.  I’m not going to hold it against the really old time period for its lack of data use.

So, Palmeiro should be a better than 50/50 candidate based only on WAR.  But if you include the fact that he has had a very long career, and very little peak, that’s probably held against him somewhat.

I had another thread a few years back that basically showed that WAR^1.5 or WAR-squared (season-by-season) gives a better representation of players currently in HOF.

So, WAR-squared basically is a combination of career and peak value.  I didn’t try it out on Walker and Palmeiro and McGwire and Trammell.  If someone wants to try…

That’s interesing about the range for 45-70 WAR. I don’t think I had seen it before.

You can see it pretty clearly in Rally’s page.  Here it is for pitchers:

http://www.baseballprojection.com/war/top500p.htm

Removing pitchers not yet eligible, we see Rick Reuschel at 66 WAR as the best guy not in there (hence our shock that it took Blyleven so long).  Kevin Brown at 65 WAR is next (if we ignore the real old timers).

If you focus on the guys with 50-60 WAR, you see it looks around 50/50.

Under 50 WAR, and you see some sporadic pitchers, each with his own story to tell.

It’s not as clear with position players:

http://www.baseballprojection.com/war/top500.htm

You get a similar story, though it’s alot easier for a position player with 40-50 WAR to make it to the HOF than a pitcher.  Realistically, it’s easier to evaluate a pitcher’s career (runs allowed) than a position player’s career (runs created, runs saved, position).

I came up with HOF odds based only on WAR a while back, maybe 2-3 years ago.  It was fairly rudimentary, but I’m sure someone here can come up with something better, especially if you include season-by-season WAR.

Thanks for those links. I have written about Reuschel. Not only is he 30th in career WAR for pitchers but he was in the top 5 in each year from 1977-80, including a 1st in 1977.

Does Adam Darowski’s “weighted WAR” accomplish what you are trying to do by raising individual seasons to a given power? He gives extra credit for seasons going above a cerain WAR.

http://darowski.com/hall-of-wwar/

http://darowski.com/hall-of-wwar/about/

I was thinking about wWAR too when I read this.  I like it, especially since he is now adding postseason-weighted WPA.

http://www.beyondtheboxscore.com/2011/11/18/2569018/upcoming-hall-of-fame-candidates-by-weighted-war

I copied and pasted Adam’s data into Excel (and some data showed up in Excel which is not seen on the website-I think this made for a more complete data set). Palmeiro is 170th among all players and pitchers. Reuschel is 161st.

Tango, Here are some WAR^2 of the players you mentioned. For negative WAR seasons, I put the value at 0.

Bagwell: 505
Trammell: 365
Larkin: 345
Walker: 342
McGwire: 321
Raines: 309
Palmeiro: 303
Murphy 270 (he had 6 negative WAR seasons)

pm: excellent stuff, thanks for doing that.

I agree with the zero-ing of the negative WAR seasons (not only in the squaring part, but also if you just add them up).

So, Trammell is really getting short-changed by the HOF voters.

We see why Raines is not getting the love, in that he didn’t have the huge WAR seasons.

It’s possible that the squaring is too much of course.  If a guy has a 9 WAR season, that’s the equivalent of two 6.3 WAR seasons.  I’m not sure a voter is (implicitly) seeing it like that.

I know Dan T. has done alot of work on this stuff.  So, if he’s around, he might tell us what he’s learned.

Here are the same results if you do WAR^1.5 which doesn’t overvalue peak as much.

Bagwell: 198
Trammell: 155
Larkin: 152
Walker: 148
McGwire: 140
Palmeiro: 139
Raines: 138
Murphy: 110

The rankings are the same except Palmeiro and Raines are switched.

I’m curious about Adam Darowski’s wWAR as well.  I’d say that he’s getting at the same thing but the idea of WAR^1.5 or WAR-squared (season-by-season) rounds out the “edges” better because you don’t have to arbitrarily say 5.9 WAR means one thing a 6.0 WAR means another - they’re dealt with incrementally, which seems preferable to me.

Tom, I’m also curious about your comment regarding the need to adjust for context and career length when citing to wOBA.  Is there an easy way to do that?  Or would you suggest (if you prefer the FG advanced stats like wOBA) that you just use wRC+ instead?

Right, wRC+ is wOBA which is Linear Weights.  They’re all the same thing.

wRC+ is a rate stat, indexed so that average is 100.  So, you can do wRC+ minus 100, times PA (akin to above average).  Or wRC+ minus 75, times PA (akin to above replacement).

Just do something other than letting wOBA sit there as a rate stat, without adjusting for context (which wRC+ does) and without adjusting for career length (which times PA gives you).

Re Tango, 18:  Thank you for explaining that.

I like that measure for guys who are mostly being measured for their offense (the DH/1B/corner OF guys, mostly), because it does a nice job of measuring career value that is adjusted for context.  As “an Edgar guy,” I like where it leads us too, because it nicely dispels conventional (and generally uninformed, from an objective point of view) thinking that Edgar didn’t have a long enough career and/or a good or sustained enough peak to be comparable to typical “offense first” HoFers.

I went through and calculated a value above average using the formula you suggested ((wRC+ - 100) * PA), and looked at the guys Poz identified in his article as being in mostly for offense (Name, career wRC+, PA, “VALUE” adjusted for both league/era context and career length).  In order of most to least “VALUE”:

1. Ted Williams: 189, 9721, 865169
2. Reggie Jackson: 139, 11416, 445224
3. Willie McCovey: 145, 9686, 435870
4. Harmon Killebrew: 142, 9831, 412902
5. Willie Stargell: 145, 9026, 406170
6. Ralph Kiner: 146, 6256, 287776
7. Orlando Cepeda: 131, 8695, 269945
8. Jim Rice: 128, 9058, 253624
9. Hack Wilson: 142, 5556, 233352
10. Tony Perez: 121, 10861, 228081

Average “VALUE” for these 10 HoFers: 383811

I personally think that’s a pretty good proxy of HoF guys who made it mostly for offense, but if you want you can add a few of the more recent HoF inductees who *sort of* fit that description:

Eddie Murray: 128, 12817, 358876
Dave Winfield: 129, 12358, 358382
Wade Boggs: 131, 10740, 332940
Tony Gwynn: 132, 10232, 327424
Paul Molitor: 125, 12160, 304000
Andre Dawson: 118, 10769, 193842

Murray, Winfield, Boggs, Gwynn and Molitor would slide into the first list in between Stargell and Kiner; Dawson brings up the rear.

If you add these six to the previous ten, you don’t necessarily improve the picture of what an “offense first” or “primarily offense” HoF inductee looks like, but you definitely get a better picture of what recent HoF voters have thought was “enough” for induction.  Add those six in, and the average “VALUE” for the 16 “offense first” players already in drops to 357099.  [The average for those recent six inductees alone is just 312577.]

Now, here are the guys that #2/SJ listed, that led to Tango’s comment about the need to adjust wOBA (or wRC+) for context and career length:

Jeff Bagwell: 149, 9431, 462119
Mark McGwire: 157, 7660, 436620
Edgar Martinez: 148, 8672, 416256
Rafael Palmeiro: 130, 12046, 361380
Tim Raines: 134, 10359, 352206
Larry Walker: 142, 8030, 337260
Dale Murphy: 120, 9040, 180800

Two from the current ballot that SJ didn’t list, but very well could have:

Fred McGriff: 134, 10174, 345916
Don Mattingly: 124, 7721, 185304

The first three easily and comfortably slide in near the top, if not at the top, of the proxy “standard” established by Poz’s “offense first” list.  The next three are about average candidates (or better, if you consider the pattern of more recent “offense first” inductees). Murphy doesn’t belong, at least based on this measure; it might be a different story if you combined a measure of 5 or 7-year peak in addition to the career value, but even then it is difficult to see how that moves Murphy much beyond the range of Tony Perez or Hack Wilson, who look to be the bottom of this particular barrel.  If McGriff were on SJ’s list, he too looks like an average to slightly below average candidate; Mattingly looks a lot like Murphy.

Just for kicks, here are some other candidates from recent times (both retired and active):

Barry Bonds: 175, 12606, 945450
Frank Thomas: 154, 10074, 543996
Alex Rodriguez: 148, 10634, 510432
Manny Ramirez: 152, 9774, 508248
Jim Thome: 145, 10127, 456615
Ken Griffey, Jr.: 133, 11304, 373032
Albert Belle: 148, 6673, 266920
Sammy Sosa: 123, 9896, 227608

I’m sure I’ve forgotten about a lot of guys.

Take it for what it’s worth.  Some of these guys might deserve some credit for being guys of outstanding character who played primarily for a single team, and/or for generally being thought of as being “clean” players in a “dirty” era (and some others may needed to be dinged or discredited for the opposite reason, to varying degrees depending on whether known or merely suspected, and to what degree of “certainty").  Others were fine defenders even if they played a position not generally known for defense, and some people may want to give them some extra credit for that.  That seems fine to me.  But this is a decent starting point.

I don’t know that this is any better than using WAR, though, because WAR already has positional adjustments and accounts for defense and other important factors (like baserunning).  Still, it might help provide some context with “offense first” candidates.

Good job.  Dawson, Murphy, Junior are all CF mostly, or at least have substantial fielding value, as does ARod.

We can quibble about the others, but it won’t change much of what you are trying to do.

Note that if you want to give the number more meaning:
wRC+ gives you the percentage of runs created.  So, 100 means 100% of average runs created.  If the average is 0.12 RC per PA, then you would do:

wRC+ minus 100, times 0.01 times 0.12

So, in your number list above, you multiply by .0012 to get you runs above average.

You had Raines at 352206, so times .0012 gives you +422 runs above average.

If you go here:
http://www.baseballprojection.com/war/r/raint001.htm

You see he’s +306 for batting, +121 for baserunning, +8 for DP, -18 for Reach on error for a total of… +417.

I mean, I didn’t expect it to be so close, considering that the two metrics (Fangraphs, Rally) are not identical, and we’re taking shortcuts left and right.

But, that’s basically how you get it.

I fooled around with Fibonacci Win Shares once or twice. If you have the Loss Shares and handle each year individually, it comes out nice. One good reason to go with Tango’s WAR-based won-loss records, which will achieve the same thing: a peak value without arbitrary cutoffs for peak years.

Anything that uses sum-of-squares or the like works. (James’s Fibonacci number uses sqrt(2)-1 as a floor for the percentage, halves wins at .500 and sqrt(5)/2 for where the points equal wins.)

You know, I almost didn’t include Dawson for that very reason.  I had occasion to look at his 1981 stats recently, and MAN, he was a beast at his absolute peak.  That year he played in 103 of 108 games in a strike-shortened season, but STILL posted a 7.3 WAR in those 103 games (including 2.1 dWAR, according to B-REF), which projects to 11 WAR over 162 games.  I’d say that high an absolute peak goes a long way to explaining the voters’ love for Hawk, in fact.

BUT, the reason I ultimately did include him is that he didn’t really play much, if any, CF after 1983 (his 6th season), and with a couple of exceptions, wasn’t much better than an average defender in most seasons after that (in fact, the total of his RField numbers from 1984 on is -5; he was +75 over his first six years as a CF).

I only included Murphy because he was on the list that SJ/#2 posted, that started this side-discussion and exercise.  He obviously did play more CF than corner (though not by as much as you might expect), but he was not the defensive CF that Dawson was, really at any point of his career.  But point taken.  In any event, I don’t really think the inclusion or exclusion of Dawson or Murphy moves the bar or changes the point of the exercise very much.

As for Griffey and A-Rod, I posted their numbers not because they fit the “offense first” group, but because they ARE great offensive players (regardless of position; obviously their position and defensive value added move them up considerably in any discussion) and because they came from generally the same era as the players SJ was discussing.  I did that for era context, and partly because I think some voters are reluctant to vote for people like Bagwell or Edgar because there are guys not yet eligible like Bonds, Manny, or Frank Thomas who were better in the same era.  The whole thing is clouded by PEDs, obviously, but I sense some reluctance to repeat the HoF voting pattern of the 20s and 30s, where virtually every Tom, Dick and Harry who got to 7000-8000 PAs in a heavy-offense era made the Hall.  PEDs makes it even harder to figure out who the elite of the elite were in the 1990s, but I get the sense that the voters definitely think these guys were not them and want to be careful about voting too many of them in before they get to consider the era as a whole. [Not saying I agree with that...just the sense I get in looking at lots of published ballots.]

I appreciate you explaining how to get from the “VALUE” number I came up with and runs above average.  The only thing I didn’t follow was how to figure out how many RC per PA average is.  I’m tired, and can probably figure it out tomorrow, but if there is a math shortcut you can give me, much appreciated.

Oh, and one more thing from “an Edgar guy.” The list that pm/#14 provided for cumulative seasonal WAR^2 didn’t include Edgar.  If it had, his number on pm’s list would have been 372, second only to Bagwell. 

And all of this basically just mirrors what WAR already tells us....

If there is 4.62 runs per game and there is 38.5 plate appearances per game, then 4.62/38.5 = .12 runs per PA.

It doesn’t matter what you use, because every player is going to be baselined to the same number.  point 12 just happens to be a useful number for meaningful context.