.... copied from the Childs thread ...

jschmeagol: A personal favorite of mine too John. For those of you that rely heavily on Win Shares his numbers aren't as good as Doyle's. But remember that he should get a schedule boost.

John Murphy: Correct, which would help Childs peak and career-wise.

This is true. Though Childs best season was long (1892) and his second best season was in a very weak league (1890-AA). Still, he is helped by a schedule adjustment.

Raw WS totals are:

LD: 289
CC: 238

Adjusted for 154 games:

LD: 295
CC: 270

(Joe gets 310-284 for 162 games in the PA thread)

154 game adjusted WS seasons, sorted:

LD: 33-30-28-27-25-22-21-20-19-18-17-16-15-07
CC: 37-33-32-28-24-23-21-21-19-13-12-07

Childs looks better at the top of the list at first, but that top season is the 1890-AA. The year after they lost both Brooklyn and Cincy and the year of the PL. There is a lot of league quality debate out there, but I think there is a consensus about the 1890 AA. How much to discount that season, I don't know. Drop it down to 28 and they are pretty even again. Drop it less and Childs still has a slightly higher peak.

For those that discount 1907-1920 NL, then that of course hurts Doyle.

Doyle's extra two seasons are basically the career difference right now.

Joe has Doyle up by ~0.5 PA. (doesn't take into effect league quality issues on either end there I believe.

I have Doyle 3rd and Childs 5th in 1940. I have them VERY close. I could pick my own nits and say that it's because I don't put too much stock in the NL discounts of the teens (but others may disagree with me on that). But again its quite close.

Adjusted for 154 games:

LD: 295
CC: 270

I give a little more of a boost to Childs because infielders were more roughed up during era, so there shouldn't be a gap there, IMO.

My problem with Doyle is not Doyle himself but the phenomena of outliers at many of the positions during the Deadball Era. Was the competition affecting the standard deviation between the best and worst players? Bobby Wallace and Max Carey were hurt in my eyes for the same reason.

I discount Childs' 1890 heavily, but also hit Doyle for a weaker NL.
I have been voting for Childs annually and not Doyle, but I strongly advise everyone to seriously reconsider each, as I will.
The other candidate at this slot is Bill Monroe, who may once again have suffered from bad timing in relation to the Frank Grant push..

Bobby Wallace and Max Carey were hurt in my eyes for the same reason.

I wasn't fond of these guys either because of their hitting numbers.

Carey never made my ballot... Wallace got onto my ballot due to longevity, fielding and I'll also admit that there was influence from the consensus when I was a newbie voter.

Doyle could hit.

Doyle could hit.

In my system of adjusting offense for context, there's almost no offensive difference between Doyle and Sam Thompson.

My struggles with trying to get a real estimate of Bill Mazeroski's fielding led me to look at another 2b high in my comparison set, Larry Doyle. Basically, I compare assists + an estimate of GBs for the player and his team against the rest of the league. Here's Doyle's results during his 1912-19 prime I have identified, using an OPS+/ERA+ kind of scale.

1912 80
1913 77
1914 79
1915 86
1916 81
1917 80
1918 57
1919 67

The extremely low numbers at the end reflect a shortage of playing time. But Doyle's highest score isn't much better than Fox's or Mazeroski's worst score. Doyle looks like he was a statue in tne field - although he's somewhat hurt by his era, where GBs were more common than FBs, a reversal of Fox's and Mazeroski's environment.

Looks sorta like Maz' OPS+. OK I exaggerate, a little.

Well... I was starting to have doubts, whether I had Doyle too low... and #6 reminded me why I have him where I have him.

Larry Doyle has the same OPS+ as Edd Roush.

I see that Laughing Larry won an MVP in 1912.
That year he edged out Honus; it was the closest Honus Wagner ever came to winning an MVP.
Talk about some misguided voters!

That was also the year that Cobb batted .409; had 226 hits, stole 61 bases and finished 7th in the AL race.
Joe Jackson picked that year to bat .395, hit 44 doubles and 26 triples, slug .579, have an OPS of 1.037 and he came in 9th!
He too, Joe that is, never won an MVP.

I wonder who did the MVP voting in those days and what were they looking for in an MVP?

Doyle and Fox are candidates caught in a curious juxtaposition as I see it. Both seem marginally qualified for the HoM but for exhibiting entirely different skills/strengths, yet each won an MVP in their respective league.

I think I like Grich a lot better; I'll like Biggio, Kent and Alomar a lot better in future years.

Deadball Era Resources (Players - Contemporary Awards & Honors) is a good place to begin learning about the Chalmers Award.

By the way, does anyone know how/why he got the nickname of laughing Larry?

I think I like Grich a lot better; I'll like Biggio, Kent and Alomar a lot better in future years.

Well, I liked Grich better, too, but he's already elected and thus off the board. Biggio/Kent/Alomar/Knoblauch/Whitaker/et al aren't eligible yet - I'll probably take many of them quite seriously.

The "MVP" (actually, the Chalmers award) was as far as I know voted on by newspaper reporters. 1912 was only the second year of the award. There was a rule against a player repeating, but that was irrelevant to the 1912 NL election (only Wildfire Schulte would have been ineligible). The reporters then, as now, look fist to winning teams, and the Giants ran away with the pennant. So you start with a team (in this case, the Giants), then you have some narrative reason why you should concentrate on a particular player. As I see it, that could have been Doyle or it could have been Mathewson - and the writers chose to focus on Doyle. (Try explaining how the "designated Twin candidate" for 2006 became Morneau rather than Mauer or Santana before you launch into speculating about other generations.) In an 8-team league with a 154 came schedule, each team is scheduled to play each other team 22 times, so the beat reporters would have a fair amount of personal observation to go on. The one offensive statistic they would have valued above all other things would have been BA. (Doyle didn't lead the league but he did lead the Giants.) They also made a (non-statistical) fetish over defense (and general athleticism). Some of this may have been misdirected (cf Hal Chase), and you know about all the problems associated to non-statistical subjective evaluation of defense. But if the writers had seized upon the notion that Doyle was a "statue" who was hurting the Giants on defense, they probably wouldn't have voted for him.

Had the Pirates won the pennant, or even if it had been a close race between the Pirates and the Giants, I suspect that Wagner would have won. He had two things weighing him down: the first is the fact that the Pirates finished 10 games back, and the second is the comparison to Wagner circa 1908. The latter is, of course, completely unfair. Even after the long steep downward slide from "god of baseball" to "arguably the best player in the league," it would be fair to focus on his current production and not his past - but that's not the way human nature works. The same factor (all-time great but maybe not what he used to be) may also have influenced the writers away from focusing on Mathewson as the "designated Giant."

In my offensive system, I have 1912 as a "44" for Doyle - a very good year, but only his own third best, after the 59 in 1915 and the 48 in 1911.

Larry Doyle has the same OPS+ as Edd Roush.

Not just OPS+, and not just any rate stat: I have Roush and Doyle in essentially a flat-footed tie in all aspects of my offensive system.

Arguments that have been made for Roush over Doyle:

1. Defense. A good CF versus a bad 2B. The only thing: we've elected more outfielders, including more CF, than we have infielders. And the CF of Roush's time hit better than most infielders.

2. League strength. A tough way to draw a line between Roush and Doyle, since they shared leagues for 5 years, and since Roush had a year and a half in the FL.

Note that both were affected by the 1918-1919 short schedules.

By the way, does anyone know how/why he got the nickname of laughing Larry?

Couldn't find an answer. I do know that writers back then liked alliteration.

I just found the SABR Bio on Doyle (see bottom of bb-ref page for a link). It focuses a bit too much on his early career (barely touches on his decline and retirement at all) but it is not a bad read.

Well, I had Grich as a slightly better hitter (and, of course, he has a massive glove advantage). But having such a gap between them as I had in my latest ballot (Grich #2 with Rose boycotted, Doyle in the 20s) is starting to bother me a bit, so I would like to read more about him.

Along with no repeaters, I think the writers could only vote for one player on any given team, so if a voter voted for Mathewson, he couldn't vote for Doyle and vice versa. This could and probably did skew things.

On the Doyle-Roush comparison:

Yes, their OPS+ are equal for their careers, but Roush during his prime was quite a bit of a better hitter than Doyle. Here's a comparison of their age 21-33 seasons, which is all of Doyle's career except for his rookie season. It drops out Roush's rookie season as well, plus the last four seasons of his career.

Doyle
1697 g, 7131 PA, 127 OPS+, .295 EQA, 521 BRAR, 367 FRAR, 888 RAR total (Doyle is -115 FRAA)

Roush
1556 g, 6441 PA, 137 OPS+, .304 EQA, 547 BRAR, 329 FRAR, 876 RAR total (Roush is -17 FRAA)

So, Doyle wins on durability and very narrowly on defensive value. By WARP's measure, a poor defensive second baseman is slightly more valuable than an average centerfielder. But Roush's superior offensive performance makes him a more valuable hitter, despite playing about a season less of games because of holdouts and injuries.

Doyle is around 1.3 WAR more valuable than Roush over 13 seasons, by these measures. However, Roush has the higher peak in his top offense seasons, and he has two average and two poor seasons in addition at the end of his career. Also, if one adjusted the seasons shortened for WW1, that would help Roush catch up, as those were his two best offensive seasons, while Doyle's performance during the war years was not outstanding.

A win shares analysis would show Doyle more favorably, I expect, because its low batting replacement level would throw more win shares Doyle's way.

Heck, why don't I do the math?

WS, age 21-33
Doyle 234.3 bws, 49.5 fws
Roush 225.5 bws, 49.2 fws

As expected, win shares favors Doyle on offense, but gives Roush more credit on defense, which is somewhat surprising to me. Here Doyle is ahead by about 3 wins, rather than 1.3.

I trust WARP's batting replacement level much more than win shares' zero point these days as a measure of value, but it's worth looking at both systems.

I would be interested to see if an analysis based on RCAP and FRAA would yield similar results.

Overall, I have to say that Roush is clearly ahead. He obviously leads Doyle on career value, their primes are almost equal, and Roush's peak is better.

But they are a bit closer than I had realized: I may move Doyle up a bit in 1993--not that he's in danger of breaking onto my ballot, but he should be higher than he is.

I've figured it as Grich hit like Doyle, fielded like Gordon.

By the way, does anyone know how/why he got the nickname of laughing Larry?

According to R. J. Lesch in Deadball Stars of the National League, Doyle had a "kindly nature and sunny disposition." Of course, his most famous quote attributed to him was "It's great to be young and a Giant." Sounds like a merry guy (does anybody still refer to a person as "merry" anymore?)

merry, droll, gay . . .
even 'sunny' is out of style anywhere but the Hall of Merit


Deadball Era Resources (Players - Contemporary Awards & Honors) is a good place to begin learning about the Chalmers Award.

Good one.

Doyle's as good a time for me to ask this stupid, lurker, question as any. Could anyone hip me as to really understanding the fielding numbers on baseballreference.com? I read the FAQ there, but it only left more questions. I don't want to threadjack, but if someone could throw me a link to real information in this regard I'd appreciate it. I see various models, but none of them make complete sense to me (e.g. basing range on groundballs assists but without taking "flyball pitchers" into account and such).

Could anyone hip me as to really understanding the fielding numbers on baseballreference.com?

I thought bb-ref just had unadjusted numbers. Games, (innings when available), Assists, errors, putouts, double plays, Fpct. Range Factor is just (A+PO)/9 or (9 * (A + PO)/ Inn) depending on how much detail there is.

No adjustments for park, era or pitching staff (g/f or k/9).

bb-ref has been very conservative with fielding data. Are you talking about another site (baseballprospectus, winshares, uzr, etc)?

Thanks for all the help guys.
Paul, I'll be sure to read up on the Chalmers award.

One of my RCAA offensive charts. I think I've taking a level-of-competition adjustment for the Federal League for Roush, but it's out in the lesser years and doesn't actually make a big difference. Roush has more years than Doyle but also missed more time within years, so the playing time isn't all that different.

David: No, I was talking BB-ref, but you brought up UZR and I'm gonna study that until even a Humanities teacher such as myself can understand it. I may need a full weekend or two for that to happen. Thanks.

bump

X-posted from 2005 results.

I think I asked the question "where are the 2B who hit (in the 1910s NL) like Larry Doyle?" and the answer remains, there aren't any...except Larry Doyle. Whether as a group they hit pretty well (relative to 1910s 1B) doesn't mean that the best of them is unqualified for the HoM. Doyle was far and away the best of that group. Where or where is the 1910s NL 2B that hit like Larry Doyle?

Actually I have been working on documenting this whole question and there's more than I can get organized and posted right now but I'll just generalize. There are 10 peer 2Bs with enough innings to have a WS rating and only one of them, Johnny Evers, rates more than one full letter grade better. Doyle was a C+, Evers an A-. Everybody else--Cutshaw, Egan, Gilbert, Huggins, Knabe, Miller, Rawlings, Hornsby and Ritchey--are no better than B+ and most a B and B-. Other 2Bs played more and have letter grades elsewhere, and therefore played 2B for about 1 year each on average--Groh is an A- at 3B, Ivy Olson a C- SS, Doolan an A+ SS, Heinie Zimmerman a C+ 3B, Byrne at C+ 3B, Stock a C at 3B. My point being--was Doyle that bad and was the FAT (freely available talent) that good with the glove?

And is the differential on defense equal to the following: Doyle in the average year was 27 OPS+ points better not than FAT but 27 OPS+ points better than the average 2B. E.g. in 1910 he was 128 (only his 7th best year) and the other regulars (i.e. played the most games at 2B for their team whether they were batting title eligible or not) were Evers 115, Dots Miller 69, Knabe 89, Egan 82, Hummel 115, Shean 71 and Huggins 114. The average (technically the mean) is 114 + 89/2 (obviously). I use 1910 because it is the most typical year (differential of +26 versus mean of +27). Again, it is nowhere near Doyle's best year.

Maybe the reason for the following is that the FAT was so good that teams just couldn't resist going out and getting some of that FAT, year after year.

Pittsburgh 2B (1907-1920) Abbatichio 114-108 Miller 115-69-99 A. McCarthy 84 Viox 142-106-111 Farmer 103 Pitler 75 Cutshaw 116-79-71

Cincinnati (1913-1920) Egan 95 and Groh 107, Groh 120, Groh 123 and Olson 61 and Rodgers 91, Loudan 85, Shean 61, L. Magee 123 Rath 96-82. Prior to 1913 Egan held the fort for 5 years.

Philadelphia (1907-1920) Knabe 114-84-82-89-80-85-85 Byrne 85 Niehoff 77-95-93 McGaffigan 56 Pearce 26 and Paulette 78 Miller 68 and Rawlings 67

These are not atypical.

The idea that E. Collins had some impact on FAT in the NL--or on Larry Doyle's value in an NL pennant race--also does not stick for me. Sorry, more later.

137. David Concepcion de la Desviacion Estandar (Dan R) Posted: October 06, 2007 at 09:18 AM (#2563957)
sunnyday, thanks for your response.

My stance more generally is that the overall level of positional offense has nothing to do with a player's value. Since every team has to field a player at every position, the way you win baseball games is by being better than your opponents at a given position, whether that means having a 90 OPS+ where they have a 50 or a 130 OPS+ where they have a 90. What *does* matter is *how* much better you are than your peers. Pie Traynor was the best white 3B of the 1920's, but the gap between him and the worst regular 3B in MLB in those days was not particularly large. David Concepción was the best SS of the 1970's, and the gap between him and the worst regular SS in MLB in his days was ENORMOUS. That's my definition of Merit. I find Doyle far short of Meritorious because he didn't exceed the baseline (the worst-regulars average) of his position in his time by a sufficient amount. This would be true regardless of whether 2B offense was at an all-time low or all-time high in his era. I mention the strength of 2B offense in Doyle's era because his hitting probably *would* be Meritorious (ignoring fielding for now) if he had been able to produce those OPS+'s in, say, the 1960s.

You note that Doyle exceeded his positional OPS+ by 27 points. He probably gave back about 10 of those points with the glove. Is there a single HoMer with a middling career length (13-14 seasons) who only exceeded his positional OPS+ by 18 points after adjusting for defense? That sounds like an argument *against* him, not for him.

The examples of teams with a string of poor-hitting 2B that you have selected continue well past 1915, which is when both the 2B worst-regulars average and Doyle's own offense began to drop markedly. The only consistently poor hitter at 2B that you mention during the 1907-15 period is Knabe, who was a wizard with the glove.

Eddie Collins does not have any impact on FAT as I calculate it, because he was never one of the worst three 2B in his league.

138. sunnyday2 Posted: October 06, 2007 at 10:20 AM (#2564004)
>Is there a single HoMer with a middling career length (13-14 seasons) who only exceeded his positional OPS+ by 18 points after adjusting for defense? That sounds like an argument *against* him, not for him.

That of course is the question. But I would add that for a peak voter, the key is not the mean of 27, it's the advantages vs. the mean 2B of +54, +47, +45, +40, +37, +32, +28, +26, +23 that counts. IOW for 5 years the edge was not 27, it was ? +37 OPS+ points and up to +54 points, +/- (and in Doyle's case perhaps minus) some defense. But as I also showed, the FAT that was populating the other rosters at that time tended to be B and B- gloves, not a bunch of A+ guys, which is of course consistent with what we know about where the position sat on the spectrum in those days before the DP became de rigeur.

I did cover Doyle's entire career as a regular, if it is inconvenient to consider his period 1915-20 what does that say about the method. His offense began to drop and yet he provided positional advantages of +47 points in 1915, +45 points in 1919, +37 points in 1916, +23 points in 1918--again, this is against the mean 2B as a hitter, not against the worst/FAT. Knabe must have been a wizard, holding a job for 7 years with OPS+ consistently in the 80s, but your point was that Doyle's cohort was a bunch of terrific hitters. My point was, NOT, and the defense wasn't out of this world either. Knabe's WS rating is B+ which is good and better than Doyle's C+ but hardly comparable to an A+ SS. Wizard in relative terms, somewhat deflated terms, in fact. Only 3 of 10 of Doyle's cohort who have a WS ranking are B+ or higher. The other 6 are B and B-. How many runs/how many wins is half a letter grade worth against a C+ 2B when 2B was not a terribly important position? Not many.

I'd be willing to bet that we are close to electing any number of hitters whose positional advantage is less than Doyle's consistent 30-50+ for his peak/prime and 27 for his career. Bob Johnson comes immediately to mind. Does Bob Johnson provide a positional advantage vs. the mean LF (not the FAT but the mean ML talent) of 54, 47, 45, 40, 37, 32, 28, 26, 23 with a mean of 27? I would be totally shocked if that were the case. No wonder we're electing too many bats. We don't even seem to demand a positional advantage, just a high OPS+.

If I have time I will compare Doyle to Concepcion, but time is something I don't have much of right now. I agree with you that we need to be considering positional advantage, not just the raw offensive production, and that a good middle IF or a good C can provide a ton of positional advantage. I just find that Larry Doyle fits that description, along with Rizzuto, Pesky and others. I haven't yet determined to my own satisfaction re. Davey. Why not Fregosi? He looks better to me as of today.

Continuing:

145. David Concepcion de la Desviacion Estandar (Dan R) Posted: October 07, 2007 at 07:08 PM (#2565995)
Wow, in the day since I've been away from the Internet this discussion just got seriously substantive. Great thread, guys. On the Merits:

Sunnyday:

1. You have to take into account that the standard deviation of fielding ability in Win Shares is far below what all the play-by-play systems show it is, and far below the level that is consistent with the standard deviation of team defensive efficiency. (The degree of this underestimation varies by position, and is particularly extreme for corner outfielders and first basemen, but for second basemen the stdev is 38% lower in WS than it is in Chris Dial's system). "How many runs/how many wins is half a letter grade worth" is the *exact* right question to ask. My research finds that for Doyle, it's about 0.9 wins a year below average.

2. OK, so the criteria you're looking at is best 5 years of OPS+ above positional average? I don't happen to have positional OPS+ data--where did you get it, out of curiosity?--but I can convert pretty easily from my BWAA. Looking at Reggie Smith, for example, I get OPS+ above position of +56 in 1977, +50 in 1978, +44 in 1974, +41 in 1969, and +37 in 1973, suggesting that the two were similarly valuable hitters at their peaks after adjusting for position (assuming your numbers for Doyle are right). Then once you factor in that Lil' Reggie has none of the fielding problems that Doyle had, and he has much career more outside of his peak, and that he played in integrated leagues with a low standard deviation while Doyle was playing in the EXTREMELY weak teens NL (compared to the star-studded AL), and it seems that Lil' Reggie leaves Laughing Larry in the dust...no?

3. I don't support Bob Johnson, but he very well might have had similar OPS+ above positional average. Shouldn't be hard to look up, no? (Just make sure to knock something off for the war years).

4. Fregosi is a serious candidate, right on my in/out line. He was a no-doubt-about-it superstar from 1963-70, albeit in an AL that was substantially weaker than the NL of those days. Definitely a higher peak than Concepción's. But Concepción has *so* much career on him--Fregosi really has *nothing* outside of those eight years--that he's substantially ahead from me. A pure peak or short-prime voter unconcerned with league strength would probably prefer Fregosi.

147. Eric Chalek (Dr. Chaleeko) Posted: October 07, 2007 at 08:17 PM (#2566307)

Dan, I think you could be double-dinging Doyle and some other players on the unintegrated question, within the HOM context (not as a general rule). I'm not sure how much sense this will make, but I'll try. In the NgL era (about 1900-1950), there's a big bulge in the number of electees each year and a consequent bulge in the number of total electees concentrated in the the NgL era when the NgLs were at their peak and our records are best in tact for them.

So what's happened, in essence, is that we've elected two parrallel groups: the best MLB players of the NgL era, and the best NgL guys of the NgL era. We essentially avoid the question of what an integrated league would even look like by electing represntative samples from both simultaneously. We elect from three (or four) leagues instead of two, meaning that the comparisons of MLB players to league still works by dint of MLB election slots being apportioned in similar proportions to other 8-team leagues. So applying the unintegrated argument to, say, Bob Johnson may be hitting him with an extra demerit relative to the number of election slots apportioned to his era.

Now your rejoinder seems likely to be about the quality of the league rather than about HOM representation (that integrated leagues were just better, and indeed were), but for the purposes of HOM representation and voting, we could instead posit that Reggie Smith was playing in the near numerically equivalent environment of three to four leagues (24 to 26 teams versus 16) just as Johnson was playing within a three-to-four major league system throughout his career. Our perception is different of those leagues, but the demographic reality is pretty close for each.

Now returning to Doyle. The question is, indeed, murkier. We have less reliable stats before 1920, and we're not entirely sure how many Black teams there were, nor how good they were. But we've nonetheless HOM'ed a bunch of Black players from the Doyle era: Williams, Johnson, Lloyd, Hill, Santop, and pieces of Torriente and Charleston's career (off the top of my head, sorry if I forgot someone), an era in which we were electing 1 player a year for much of the time. So while maybe not as compelling as the Johnson example, the point remains in play: we have built in representation, which has the effect of avoiding the question of whether an integrated league is better or worse than a latter-day integrated league.

Your mileage is likely to vary.

148. David Concepcion de la Desviacion Estandar (Dan R) Posted: October 07, 2007 at 10:30 PM (#2566804)
Sunnyday:
I just double-checked Smith's vs. Doyle's position-adjusted offense. First, I screwed up Smith's 1973 because I didn't account for the DH, but he also didn't play a full season that year, so I guess those two factors probably cancel out. Second, I did Doyle by the same methodology (BWAA per year minus positional average BWAA per year times 11.43), and get notably lower numbers for his top five than yours: +46, +45, +43, +37, and +36, which is way worse than Smith. How are you getting your results? Assuming that the +54 you cite is 1911, that would suggest that 2B were exactly league-average hitters in the 1911 NL. Given that their EqA was .276, that's just about impossible.

150. sunnyday2 Posted: October 08, 2007 at 09:41 AM (#2567044)
For Larry Doyle, yes, his biggest year (+54 OPS+ versus the mean for NL 2B) is 1911. All I did was take all of the NL 2B in order:

Doyle 153 OPS+
Sweeney 120
Hummel 113
Huggins 99
Miller 99
Evers 86
Knabe 80
Egan 75

Or for his most average year(s), 1910 and 1912:

Doyle 128 and 132 (+26 and +28 versus the mean regular NL 2B)
Evers 115 and 139
Huggins 114 and 117
Knabe 89 and 85
Egan 82 and 72

Pitt--Miller 69 and McCarthy 84
Bkn--Hummel 115 and Cutshaw 91
Bos--Shean 71 and Sweeney 133

As noted elsewhere, for his career Doyle averaged +27 OPS+ points versus the mean regular NL 2B each year. Taking a typical year, in this case 1910, you get the following.

Doyle 128 OPS+ 21 offensive WS and 4 defensive in 151 games
Mean 102 OPS+ 12 offensive WS and 4 defensive in 139 games

That is the sort of advantage Doyle gave his team in a typical year.

I wasn't suggesting, BTW, that 5 year consecutive peak is the be-all, it was just one set of numbers to look at. But it is worth noting some of the assumptions that go into this analysis, aside from the fact that it uses WS. It also sets Doyle (or any player) up within his league on the basis that "a pennant is a pennant"--i.e. it is within the NL that Doyle's value has meaning. His value has no meaning versus the other league.

For the record, the AL 2B that year include Eddie Collins and Nap Lajoie, and some competent hitters like Frank LaPorte, Jim Delahanty, Larry Gardner (in 113 games in his first year as a regular and his only year at 2B), but also non-entities like Rollie Zeider, Red Killefer and Frank Truesdale. Net: better than the NL and undoubtedely pushing the mean up a few OPS+ points and maybe 2 WS.

So what's left is 9 offensive WS within the league and maybe 7ish versus the MLs, and a break-even with the glove. Note that as a C+ glove Doyle is worse than every peer for whom we have a letter grade. What evens this out is the simple fact that there are a few others for whom we don't have a letter grade, and the fact that Doyle had more playing time that all of the better opponents and 151-139 games on average. How many WS is a back-up 2B going to get replacing Johnny Evers for 30 games or Jim Delahanty for 50 games? Not too many. So I would say that Doyle's defense actually cost his team relatively little in practice. Not to blow it off completely, but Dan's own calculation says 0.9 versus FAT. It might be 1 to 1.5 WS versus the average team (not Evers or Huggins alone, but Evers or Huggins and their replacements). That seems to me to be pretty much the limit.

151. sunnyday2 Posted: October 08, 2007 at 09:45 AM (#2567050)
All of this BTW is meant to test DanR's methods in ways that make sense to me. That is, asking how much of an advantage does a player provide versus others at his position (which is what DanR is trying to calculate, I think) but, in my case, within the league, not versus all ML talent. In WS. And comparing the test players versus the mean rather than against FAT.

The obvious problem is I can't calculate these comparisons by hand for more than just a few players, but I am working on a few other players.

Just in the case of Larry Doyle, it seems that he was worth about 27 OPS+ points and 9 offensive WS in a typical year and gave away no more than maybe 1-1.5 WS on average with the glove.

The key question is whether this is a great performance, or whether it is just very good, or maybe only a good performance and all I can do is compare Doyle to a few other players this week. That is what I will be doing.

152. sunnyday2 Posted: October 08, 2007 at 09:47 AM (#2567052)
PS. 5 years peak is not the be-all or end-all here at all. It was just one of the data sets I happened to look at. What I'm really trying to do is look at careers and then at "typical" seasons for a more finite understanding of what is really going on.

And finally to clarify, the average for NL 2B may well have been below league-average (i.e. DanR says their EQA is .276). My numbers are for the mean regular 2B, which would of course be a higher standard. My interest was in how much of an advantage Doyle (and others) provide versus the guys who are also likely to most impact a pennant race, head to head.

.276 is WAY above average, sunnyday--average is .260!

As I just posted in the other thread, Doyle's OPS+ was 46 points higher than the average starting NL 2B in 1911, not 54.

Note that OPS+ as calculated at baseball-reference excludes pitchers.

EQA as calculated at baseballprospectus includes pitchers.

So a position player whose OPS+ is average will generally have an above-average EQA.

(You all know this, of course, but it's another factor worth remembering in discussing how good second-base offense was at this time, so as to avoid apples vs. oranges comparisons.)

Right, I hadn't forgotten that. But on an apples-to-apples basis, using Sunnyday's own criteria--OPS+ above the simple average of starters at the same position in the same year in the same league--Doyle's 1911 was +46, not +54.

I was using the mean.

Are we not talking about the same thing, sunnyday??? What are you comparing Doyle's OPS+ to in 1911? The simple mean of the OPS+ of the 8 starters at 2B in the NL that year was 108, making Doyle +46. Heinie Zimmerman (124 OPS+) was the Cubs' starter, not Johnny Evers (86).

Xing from 2005 results and yes, I accept the correction for 1911.
--------
The 1-1.5 WS is the actual margin by which the "mean" 2B outperformed Doyle, looking at the actual WS earned by each regular 2B at his position in 1910. He earned 4 defensive WS in 1910 (rounded), Evers and Knabe 6, Huggins 5, Dots Miller 2, and so on. So he was not 2.7 WS worse than average (or than the mean) on a seasonal basis. Of course, playing time enters in. I suppose that with rounding he could be 2.7 WS worse than the best defender for the season but that would be pretty much the worst case.

I'm not arguing that 2.7 is wrong, just that it overstartes his actual performance per season in which neither Doyle nor anybody else actually played 162 games, and many played less than 130-135-140. And recognizing that this was more of an offensive position and as a result, regulars like Egan and Huggins and Dots Miller were B- gloves themselves. Evers was the only A (A- actually) glove in Doyle's NL, and Evers was done after 1916, so while I suppose he might have given back more than 1 or 1.5 WS in some other years, I can't believe it would be a lot different even if his defense was in decline before his offense was (and his OPS+ was 116-137-107 in his final 3 years). One thing about Doyle is that there was little if any ramping-up or decline (OPS+ 83, then 134 his first 2 years). So his career was "shortish" at 14 years, but relatively few players had 13 years of prime like he did. (OK, OPS+ 99 in 1917).

But bottom line, even after giving back 1-1.5 WS per actual season, he is still +9 WS including offense (+10) and defense (-1). Earlier it was stated that if he was +28 OPS+, he probably gave back 10 OPS points with the glove. I had a hard time visualizing what this means, so I wanted it in WS. So a 28/18 ratio seems to be somewhat overstated.

I realize these are WS and not every likes WS. And again, this particular year was selected for a closer look because his OPS+ advantage of +26 in 1910 is about the mean for his career. He advantaged the Giants (or Cubs) by ? 27 OPS+ points in half of his seasons and ? 27 points in the other half, though in his case he was below the mean only one time, in his rookie year. Correcting 1911 to +45 wouldn't change any of that.

And how this compares to other backloggers, I don't know yet. I am out on a limb here, you know. I support Larry Doyle, and I am going to explore whether +27 OPS+ and +9 WS per year turns out to be a good record. I think it will.

With no consideration for league strength, Rizzuto is the #1 backlogger using my consensus estimator. The FAT level at the SS position was basically flat from 1915 to 1960 at around 3 wins below average per 162 games. So if Rizzuto's hitting was below average for the position in his league, then you were really dealing with a situation like the early-80s or late-90s AL, where you had a few superstars pulling the average way up. Sunnyday, have you tried comparing players to the median rather than the mean OPS+? That might give you better results.

I'll just put up charts for Rizzuto and Doyle here...I've filled in the war credit years for Rizzuto to my taste, your mileage will obviously vary. SFrac is percentage of the league average plate appearances per full season played, BWAA is batting wins above average, BRWA is baserunning wins above average, FWAA is fielding wins above positional average, Replc is the gap in wins between a replacement player at the same position in the same playing time and the overall league average, and WARP is wins above replacement (BWAA + BRWA + FWAA - Replc). aTTL is career totals ignoring sub-replacement seasons. All figures are standard deviation-adjusted, which *strongly* favors Doyle as the teens NL was a very low stdev league.

Rizzuto




3-year peak: 20.9
7-year prime: 40.2
Career: 60.8

Doyle



3-year peak: 14.0
7-year prime: 29.6
Career: 42.1

Well, if anyone's listening, I don't see how these two guys can share the same ballot, as Rizzuto simply blows him away. The argument for Rizzuto seems extremely strong. He's slightly underrated by OPS+ due to his OBP-heaviness, his good baserunning, his having lost three prime years to the war, and him being played in his decline phase when he shouldn't have been; correcting for those factors would give him a lifetime OPS+ of approximately 98. According to both Fielding Win Shares and BP FRAA, he was absolutely an all-time elite defensive SS. 17.4 career FWAA (with war credit) is *outstanding*, behind only Ozzie, Wagner, Dahlen, Tinker, and maybe Maranville at SS. (Marty Marion is at 15.4 before war deductions, Ripken is 13.7, Concepción is 13.4, and Belanger a surprisingly low 11.7, since he played in many fewer games than his rivals and was inconsistent). In addition to his huge MVP year, he played at a high All-Star level in '42 and '47, and at an All-Star level in '41, '51, and '52 (plus his war years). A 98 OPS+, historically great defensive SS with a terrific peak? Seems like he should be a lock to me.

As for Doyle...say wha? Again, 2B in his day was an offense-first position--the worst starting second basemen in his day were no worse than the worst starting 1B and not much behind the worst starting OF. Adjusting for his fielding and baserunning, his lifetime OPS+ "should" be 118, and his position was easier then than 3B is today. In terms of value, that sounds pretty similar to me to Matt Williams. Get real.

I'll post this on both players' threads.

Another way to look at this is to say that Doyle and Rizzuto both produced about the same value above overall league average in their careers: Rizzuto 25.0 wins above average, Doyle 27.3 wins above average. And then you remember that one was a shortstop, and one played 2B at the easiest time in the position's history. It's just not close.

I'm sure I'm the biggest idiot ever, but I'm still struggling to understand your method--or, I suppose, to accept your method--at the theoretical--or, shall we say, normative--level.

e.g. Doyle is 27 WAA for his career, Rizutto is 25 (and I think that includes his war years, as it should). How is it that 27 WAA doesn't belong on the same ballot with 25? Why isn't a WAA a WAA? Considering the mathematical rigor of your method, it seems odd to me that, like Bill James, you resort ot a softer, qualitative variable to get from the numbers to an actual ranking--i.e. one was a SS.

It's not that I don't understand that an average player is better than a replacement player (however defined). Specifically you claim that a an average SS in Rizutto's day was 36 wins better (no, sorry, 35.9) than a FAT SS, and an average 2B in Doyle's day was only 15 WAA better than a replacement 2B, because that replacement 2B just didn't need a whole lot of skills to be play in the MLs.

But so what if it was offense first? We're not talkin' skills, we're talkin' value. And Doyle brought 28 OPS+ points to this offensive position that the average 2B didn't bring, and didn't give much away with the glove. Sure, he wasn't great, it was an offense first position. But neither was the average 2B a great glove either, nor did it make a whole lot of difference. It was an offense first position.

And equally, what really matters in a pennant race is not how much better Larry Doyle is than Dave Shean, who played 148 games at 2B for last place Boston in 1910 much less Shean's back up Gus Getz, or how much better he was than Possum Whitted, Harry Pearce, Eddie Sicking and Gene Paulette, who together tried to man 2B for the past place Phillies in 1919. I would guess that one or more of these guys was FAT. What mattered was how Doyle stacked up against Johnny Evers, and how his offense-first game matched up to defense-first 2B (yes, there were some) like Otto Knabe (2nd place Philly 1913) and Morrie Rath (Cincinnati's champs of 1919). And I'm not trying to white-wash Larry Doyle, the fact is he and the Giants finished 2nd to Evers' teams in 1910 and 1913 and Rath's in 1919. But they won the NL pennant in 1911-12-13 and Doyle was the Giants top position player (by WS) in the first 2 of those years. He trailed Art Fletcher, T. Schaefer and G. Burns in 1914 24-23-22-21.

As a peak voter, I have always given strong consideration to where players fit into the MVP voting (that would be MY MVP ballot, of course) and all-star team selections. By implication, the cut that really counts is against the next best players in the league at the position. And Doyle of course scores well on that sort of analysis.

Rizzuto is a harder case for me because he did not provide his team with as much of an edge. But just like Gehrig-Foxx-Greenberg and the ABC boys and Yount and Ripken and so on, one can recognize that fairly easily and make adjustments, and I do. I never thought that Rizzuto belonged in the HoF (and that Doyle did). Now through this process (HoM) I've decided they both do.

Not to speak out of turn, but here is how I would answer sunnyday2's question about Dan R's system (Dan, please correct as needed)

Why isn't a WAA a WAA?

Essentially, because it's not adjusted for position.

Dan R's system adjusts for the defensive spectrum by looking at how far below league average hitting the FAT at a position is.

His system adds

Wins above a league-wide average hitter
Wins above a league-average fielder at his position
Wins above a replacement level player at his position. The defensive spectrum is shown, basically, by how much offense teams were willing to get up to get an average defensive player at a given position. This is a way of quantifying the "qualitative" difference between positions.

It's a similar method to using Runs Created Above Position + Fielding Runs Above Average at Position, except that it finds a way to use replacement level as the baseline rather than average, since that is a much steadier number than average is, since star gluts or droughts will pull "average" up or down from year to year.

DanR, back to your post earlier about normative and whatever.... I appreciate the mathematical rigor of your method. But there are numerous normative decisions that go into it, that are built into it, and to say that 60.8 and 42.1 "prove" that Rizzuto is better than Doyle, and that this is the only "correct" position (I'm paraphrasing earlier posts) there can possibly be on the question is over-stated. The normative field IOW is larger than you acknowledged.

I am sure you believe that you are right, but I think that a 3rd party could point out the various assumptions that go into your method and say, yes, there are defensible alternatives. E.g.

WS vs. WARP
DanR's WARP vs. its other iterations
FAT vs. "replacement"
FAT or replacement vs. average or median or mean
timeline adjustments vs. league quality adjustments, not always the same thing
can we really quantify how easy or hard it is to dominate, based on timeline/league quality/pool characteristics
can we really quantify how easy or hard it is to dominate, based on intrinsic characteristics (rules, equipment, conventions)
do the 3 worst players really constitute FAT or are market inefficiencies significant enough to invalidate the assumption
do the 3 worst players really constitute FAT or are random fluctuations in the player pool significant enough to invalidate the assumption
do the 3 worst players really constitute FAT or is managerial/GM ineptitude significant enough to invalidate the assumption
does the quantitative rating of FAT really have anything to say about the quality of other players; if so, what or how much
we need to look at the ML as a whole versus each league in and of itself

Dan's work has inspired me to look more closely at positional advantages, though I don't have a way to calculate them for more than just a few players (they're manual processes right now, it probably takes an hour or more for each player). And of course it militates against the idea that no player is responsible for the quality of his opponents. That's the responsibility of the opposing GMs and managers and of dumb luck and other factors. And so we elect Gehrig and Foxx and Greenberg, despite the fact that none of them was as position-dominant (well, Gehrig was) as Gil Hodges or Steve Garvey.

So for me, Dan's work at this point is part of the mix. I'm not on board 100 percent and I've been said to be narrow-minded or closed-minded on that account. But, while I can't do the math anymore and don't care to I did my PhD work in research methods and I get the theory just fine. My issues here are not with the execution but with the theory. I am as open-minded as the next guy. Like I said, I didn't used to think much of Phil Rizzuto, not to mention Charley Keller and Johnny Pesky and Gavy Cravath and Elston Howard. I now believe that all 5 of them should be HoMers and HoFers.

Everybody here has a method. Some are better than others. Some have tried to "sell" their method over the years, some apparently could care less if other people understand their method, much less adopt it. Dan would like others to adopt his method, and others hvae adopted his method. Great. I remain convinced that, as rigorous as it is, it has been over-sold. It is not the one true religion.

Maybe my resistance to all of this is partly the result that I feel that we've all been harangued about this being "correct" and "empirical" and "proven," though in fairness I will say that more abrasive harangues have come not from Dan but from his disciple. I can't and won't be convinced there aren't valid methodological/theoretical/normative questions/alternatives or that hanging on to some of them is closed-minded. If I accepted your assumptions but not your math, sure, that would be silly. And that seems to be the expectation here because the assumptions have become the air that some people breath. But that doesn't mean they're not there.

Coming soon to a thread near you--Davey Concepcion and Graig Nettles.

>It's a similar method to using Runs Created Above Position + Fielding Runs Above Average at Position, except that it finds a way to use replacement level as the baseline rather than average, since that is a much steadier number than average is, since star gluts or droughts will pull "average" up or down from year to year.

And is it a fact or an assumption that using the steadier number is a better thing?

Maybe my resistance to all of this is partly the result that I feel that we've all been harangued about this being "correct" and "empirical" and "proven," though in fairness I will say that more abrasive harangues have come not from Dan but from his disciple.

Verily, I was but a child, wandering in the wilderness, until Dan sheweth me the light of salvation and guided me. Praiseth be to him, Child of James, of Palmer, of McCracken! Ye of little faith, fear his wrath!

Whoa! Another excellent thread!

Chris Cobb, you are 99% correct. The only slight tweaks are:

1. I separate out batting and baserunning wins above average, but that is a simple accounting question.
2. The replacement levels I use do not precisely measure "how much offense teams were willing to get up to get an average defensive player at a given position," although in 6 of 8 positions that happens to be the case. What they measure is the *total* gap, measured in wins, between a replacement player at a given position and a player at that position who hits at the league average and fields at the positional average. According to Nate Silver's research, FAT first basemen are actually *above* average fielders, which stands to reason when you think about it, while FAT SS are *below* average fielders, which is also logical. FAT players at the other 6 positions are roughly average with the glove. So theoretically, you would say that an average-fielding, league average-hitting 1B would probably be 8 batting runs above replacement and three fielding runs below replacement, for a total of 5 runs/0.5 wins above replacement, while an average-fielding, league average-hitting SS would be something like 26 batting runs above replacement and 6 fielding runs above replacement, for a total of 32 runs/3.2 wins above replacement. For the other 6 positions, those fielding runs above/below replacement are approximately 0.
3. I have no idea if my FAT levels are more or less stable from year to year than positional averages--that's not something I've studied. My guess is that they fluctuate *more* because the sample size I use to calculate them is smaller--just 3/8 of starters, instead of the entire distribution including backups. I compare to FAT rather than to average for the simple reason that average players don't grow on trees, they have value, and the idea of actually penalizing someone for playing at a below-average but above-replacement level seems absurd to me.

And sunnyday,

First off, I think we're coming a lot closer to understanding each other's methods and approaches, so I'd just start out by saying that this has been highly rewarding and very worthwhile. On to the content:

1. Chris Cobb is right. It's not qualitative, it's quantitative. Doyle is 27 WAA and Rizzuto 25, *before taking into account the fact that one played second base and the other played shortstop.* Compared to a hypothetical positionless player hitting and fielding at the league average, Doyle is 27 WAA and Rizzuto 25. But compared to the players who actually would have replaced them had they gone down, Doyle was 42 wins better and Rizzuto 61 wins better, because Doyle was a 1910's 2B and Rizzuto was a 1940's-50's shortstop.

2. I am NOT saying that an average SS in Rizzuto's day would have been 35.9 wins better in the same playing time than a replacement SS, nor that an average 2B in Doyle's day would have been 14.6 wins better in the same playing time than a replacement 2B. In fact, I don't agree with either of those statements. I'd have to calculate positional averages to get the answers to those questions, which I don't do. Rather, I am saying that a shortstop who fields at the positional average and hits at the *league* average would have been 35.9 wins better in the same playing time than a replacement SS, and that a second basemen who fields at the positional average and hits at the *league* average would have been 14.6 wins better in the same playing time than a replacement 2B. It's an important distinction, since the average shortstop is a below-league-average hitter, and the average 2B around 1910 was an above-league-average hitter.

3. Comparing to average versus comparing to FAT is a great example of a normative, rather than positive, question. There is no one factually right answer for which of those two measures better reflects value; it just depends on your preferences as a voter. I obviously think FAT is the way to go, but I would never and have never suggested that voters who use compared-to-average measures are conducting some sort of empirically incorrect analysis. That said, I *would* suggest very strongly that I think comparing to the median is *far* superior to comparing to the average, as it mitigates the effect of star gluts at a position.

4. The implicit assumption in your statement "Sure, he wasn't great, it was an offense-first position, but neither was the average 2B a great glove either, nor did it make a whole lot of difference" DOES seem to me to be empirically/factually wrong--namely, that the standard deviation of defense at right-spectrum positions is much lesser than it is at left-spectrum positions. That's what Win Shares thinks, and I think that is one of the system's greatest flaws, since that contention is NOT supported by the play-by-play data we have at our disposal from 1987 to the present. The standard deviation at 1B is indeed much lower--although that is in part because we don't have any good way to measure scooping--but the standard deviation at 3B, a mid-spectrum position, is quite high. CF is not as hard as SS, but if I'm not mistaken it actually has a higher standard deviation (I can double-check that if anyone's interested). A position's difficulty--which is best measured by position-switcher studies--is NOT the same thing as its standard deviation. Just because 2B was an offense-first position in the 1910's does NOT mean that the effect of a good or bad fielder at the position on a team's runs allowed was any smaller than it was in other eras. I think you are making a positive, rather than normative mistake, by suggesting that Doyle's poor fielding didn't matter much to his teams because 2B was an offense-first position. Poor fielders at right-spectrum positions can and do increase team BABIP allowed by just as much as poor fielders at left-spectrum positions.

5. I don't know what earlier posts you're paraphrasing, but I would never say that "Rizzuto was better than Doyle" is a positive/empirical statement! Of course it's a normative one, because it depends on how you define better! If your definition of Merit is simple batting wins above average, it's Doyle in a landslide, 32 to 5! By contrast, again, saying or implying or assuming that there's a strong correlation between a position's defensive difficulty and its standard deviation is, I believe, positively wrong.

6. Every single one of the questions you cite regarding my approach are completely valid. I've never suggested that my ballot is the only defensible one, nor that my system is the "one true religion." But every single one of the issues you cite is also normative--What is the best way to measure value? What is the best way to measure Merit?--rather than positive, e.g., What was the defensive spectrum as measured by average positional OPS+ in 1911? What comment did I make that said something was a positive/empirical question that you consider to be a normative question? When have I ever suggested my system was the only way to go? OF COURSE there are "methodological/theoretical/normative questions/alternatives." Dismissing any of them without careful study/review would be narrow-minded.

7. I said you were closed-minded for that reason--because it seemed like you were completely uninterested in looking at absolutely anything other than seasonal WS totals. Your posts over the last two days seem to have demonstrated a substantial shift in tone and content, which I greatly appreciate.

Oh yeah, and as far as 'zop is concerned, he and I just don't agree on whether the appropriate/most effective style in the group is to be collegial or combative. I suppose they both have their Merits. I would just note that 'zop was tremendously helpful in the development of my WARP system--the original ideas of using worst regulars to measure replacement level and a market salary estimator to weigh peak vs. career were his--so I'm very thankful for his contributions. If he's the voter most dedicated to my system, that's probably because a lot of it is based on his innovative ideas.

As noted, I'm coming around on Dan's method. Therefore I'm rather interested in his ongoing dialogue with SD2....

And is it a fact or an assumption that using the steadier number is a better thing?

IIRC, the steadier number is in part so steady because Dan runs a 10-year average rather than a discrete season-by-season calculation for FAT. If one were to use the 10-year average of RCAP or FRAA, wouldn't those numbers seem also to be much steadier than a year-by-year iteration of them? Or were one to use a one-season number for FAT, I think it would bounce around quite a lot.

Actually, I'd like to pick Dan's brain a second. I think Dan uses a ten-year average that includes the five prior and subsequent seasons in the calculation of FAT. My question is why the five subsequent years? The reason for this query is that in the moment, a GM assembling a team has only an inkling of what the next five years will look like. What GM saw Ripken playing SS in 1982? Anyway, most people until very recently looked to a greater degree at current scouting and previous performance as the main barometer of what players to sign and how to deploy them (that is, not rigorously projected future performance).

So does using the subsequent five years model what's really happening in MLB? I think the prior five does, because that's about as far back as anyone's likely to look and feel confident in their observations of a player. But I'm not sure if the subsequent five aren't crediting GMs with a vision they don't have.

Of course, that said, I could be misunderstanding why Dan uses those subsequent five years....

Verily, I was but a child, wandering in the wilderness, until Dan sheweth me the light of salvation and guided me. Praiseth be to him, Child of James, of Palmer, of McCracken! Ye of little faith, fear his wrath!

Has anyone ever seen 'zop and DanR in the same room??? ; )

I would just note that 'zop was tremendously helpful in the development of my WARP system--the original ideas of using worst regulars to measure replacement level and a market salary estimator to weigh peak vs. career were his--so I'm very thankful for his contributions. If he's the voter most dedicated to my system, that's probably because a lot of it is based on his innovative ideas.

Dan is, of course, too kind. If I'm the voter most dedicated to his ideas, its only because I saw him develop his system every step of the way, feel like I understand it because he's explained it to me so clearly and thoroughly, and have been exposed to his terrific results for a much longer period of time than the rest of the electorate, with many more opportunities to ask questions.

Eric, it's a nine-year average, actually--the year in question, four before, and four afterwards. My thinking is the follows:

1. My WARP is based on one absolutely enormous and completely unsubstantiated assumption: that the relationship between the FAT level and worst-regulars average at each position has remained steady over time. I think this is preferable to BP's assumption (that the relationship between the FAT level and the positional average is steady over time), since the latter is demonstrably false, but that doesn't mean mine is correct. Nonetheless, I have chosen to invest over a year of my time in this extremely shaky house of cards.

2. I find the gap between worst-regulars and FAT by comparing the worst-regulars averages for the 1985-2005 period to Nate Silver's FAT levels for that period at each position.

3. Since that gap was established by looking at a 21-year average--so, ceteris paribus, centered on 1995--it seemed logical to me to place the year in question at the center, rather than at the end, of the period used to establish the worst-regulars average for each season.

There's also a more practical issue, which is that 5 years isn't nearly a large enough sample size--remember that in the single-league 1890s I'm only dealing with 4-5 players per position per season!--and stretching back 9-10 years seems like far too much to me, as it would obscure the real changes in positional strength that can take place over the course of a decade.

Hahahah, 'zop--thanks, but let's not turn this thread into a circle-jerk session. (Can I get that past the profanity filter?)

Larry Doyle grew up in the same town as my dad, Breese, Illinois. I believe Doyle's family were miners, while my dad's family were farmers.

>Poor fielders at right-spectrum positions can and do increase team BABIP allowed by just as much as poor fielders at left-spectrum positions.

Our difference on this one can be attributed to my using WS, I suppose. I was just saying that there are fewer WS to go around, which means that an A+ 2B is not gonna have as much of a defensive advantage over a C+ 2B than an A+ SS is going to have over a C+ SS. In my experience, this can be for one of two reasons--there are simply more chances at SS than at 2B, or the chances at SS are more difficult at 2B and therefore the results more variable, (or both).

>'ve never suggested that my ballot is the only defensible one, nor that my system is the "one true religion."

Well, somebody has. I thnk it must have been Dick Thompson.

I've looked at a lot of different data. I was overly OPS and ERA oriented for awhile, I think, but I always thought that rates are over-rated. A bit contradictory. To me, the litmus test of a rate-based voter, which I regard as a mistake, was always John McGraw. (Ducks.) I am sure I use ink a lot more than most people, along with MVP and all-star voting (by which I mean MY lists of course, certainly not the BBWAA though I look at it as a sampling of contemporary opinion). And I do care about contemporary opinion. Mine is a pretty soft approach (I mean, I had Kirby Puckett in an elect-me position until I went back and got a bit more rigorous with my WS) but, yes, WS is the primary basis and everything else is in the way of tie-breakers. And of course there's the peak/prime thing. Albert Belle #2.

Bottom line--baseball is like life. Too many variables. Can't be quantified. Need bullshirt dump. And know it.

I just see Larry Doyle as the best (position) player on a team that won several pennants and contended for others, and also the best at his position in his league. That's basically it. Very much like Kirby Puckett.

sunnyday, but this is where I do accuse you of being closed-minded. Saying "our difference on this one can be attributed to my using WS" would be fine if it were a normative issue, but it's not, it's a positive one. Take a look at the standard deviation of fielding WS, and compare it to the standard deviation of all the play-by-play systems (or that of team BABIP, if you believe in DIPS)--it's far, far too small. Is it really too much to ask for you to look at the hard data on this? I can send you the numbers. This is a case where Win Shares most definitely does not "add up" as James likes to tout, it is a demonstrable flaw in the system, and it is one that I think is leading you to overrate Larry Doyle and underrate Graig Nettles.

And as for McGraw, my argument isn't just that he had shiny rates. It's that when you credit him properly for the in-context value of his hitting and use the replacement levels consistent with empirical findings--higher than WS's and BP WARP's in general, but lower for 1890s 3B than for contemporary 3B--his total contributions to pennant-winning are clearly Meritorious. Here are his ranks in league, measured in total wins above replacement not rate, at his position and in the whole league (and remember that this is almost entirely in the single-league 1890s, so these are really ranks in all of MLB--he's not "shielded" from the competiiton the way Doyle is from Collins and Lajoie):

1893 Position 1 League 12
1894 Position 4 League 10 (strong year for third basemen!)
1895 Position 1 League 11
1896 barely played
1897 Position 3 League 13
1898 Position 1 League 2
1899 Position 1 League 1 (and the second- or third- most valuable season ever by a third baseman)
1900 Position 1 League 3 (and a super tough league!)
1901 Position 3 League 11

So you have a guy who was the best player at his position in the major leagues 5 times and among the top 15 players in the majors 8 times, and had three MVP-type seasons, one of which was truly historic. It's the total peak value, not the rate, that makes him my #1 backlogger.

>Poor fielders at right-spectrum positions can and do increase team BABIP allowed by just as much as poor fielders at left-spectrum positions.

Sunny might have covered this, but I think this can't be true simply because of the distrubtion of left- and right-handed bats. If 1B and (at that time) 2B are right-spectrum positions, the fact that lefty bats are less populous than righty bats should strongly suggest that BABIP is less influenced by these guys than by the SS and 3B.

Right-spectrum means easier to play, not on the right side of the diamond....You may be right that the stdev of 1B/2B/RF defense has risen (and that of 3B/SS/LF has fallen) with the increase in LHB, I can check that, but the broader point--that defensive difficulty and defensive standard deviation are not strongly correlated--remains the same, and WS gets it wrong.

Right-spectrum means easier to play, not on the right side of the diamond....You may be right that the stdev of 1B/2B/RF defense has risen (and that of 3B/SS/LF has fallen) with the increase in LHB, I can check that, but the broader point--that defensive difficulty and defensive standard deviation are not strongly correlated--remains the same, and WS gets it wrong.

I'm fully aware of how the spectrum works Dan, but the positions in question (1B and 2B, which happen to fall on the right side of the diamond) don't receive as many balls in play since LH hitters are less populous than RH hitters. Therefore they cannot have as large an effect on BABIP as the SS/3B do. That's assuming that RH and LH hitters K and walk at about the same rates.

Since 1911 came up....
Let's just take the Assists as a quick proxy for the realtive number of balls in play at each position. In 1911 assists for 3B+SS are 6203, while 1B+2B are 4347. That's 1.42 times as many number of balls in play.
SS 3878: 37%
2B 3638: 34%
3B 2325: 22%
1B 709: 7%

The right-spectrum positions (that happen to be on the right side of the infield) can't have the same impact as the left side. Individually, relative to one another 2B can have more impact than 3B, though.

Assists are definitely not a good proxy for balls hit to 1B because a significant portion of them will be unassisted putouts.

That doesn't change the basic point, which is that yes we would expect, ceteris paribus, the standard deviation of FRAA to be largest at SS, a tiny bit smaller at 2B, significantly smaller at 3B, and very compressed at 1B. I can check my numbers to see if that is the case. What's relevant here is that 2B had almost as many chances as SS in those days. Even though 2B was much easier then than it is now, a bad 2B like Doyle could still really, really hurt his team. That contradicts sunnyday's statement that Doyle "wasn't great, it was an offense first position. But neither was the average 2B a great glove either, nor did it make a whole lot of difference." That argument is empirically false and derived from a demonstrable flaw in the Win Shares system, namely that the standard deviation of Fielding WS is far smaller than the spread shown by play-by-play systems, and the spread consistent with the known standard deviation of team BABIP allowed.

Sunnyday--was Doyle really the best position player on those teams?

Here is where I have him each year, along with the superior position players and the team's record and finish:

1908: 6th (Donlin, Bresnahan, Devlin, Bridwell, Tenney, Seymour)
1909: 3rd (Devlin, Bridwell)
1910: 5th (Snodgrass, Bridwell, Merkle, Devlni)
1911: 2nd (Snodgrass)
1912: 2nd (Meyer)
1913: 7th (Fletcher, Shafer, Meyer, Snodgrass, Burns, Herzog)
1914: 7th (Burns, Fletcher, Stock, Merkle, Bescher, Meyer)
1915: 1st (and Giants finished in last place)
1916: 3rd (Fletcher, Kauff)
1917: 1st (on 5th-place team with lowest OPS+ in the league)
1918: 5th (Burns, Fletcher, Youngs, Kauff)
1919: 4th (Burns, Youngs, Fletcher)
1920: 6th (Youngs, Bancroft, Burns, Frisch, Kelly)

So by my measure, the only years Doyle was the best position player on his team his team finished last in the league one year and last in hitting in the other. Looks to me like the answer for Doyle to the Keltner question "If this player were the best player on his team, could they win the pennant?" is a resounding No.

Looks to me like the answer for Doyle to the Keltner question "If this player were the best player on his team, could they win the pennant?" is a resounding No.

I see (and attempt to measure) this differently. A player is a moving target in most ways. In any given season, due to performance variation by himself and his teammates, he might or might not be the best on his team, let alone on a pennant winner. And, indeed, the idea of what a pennant winner looks like also changes from year to year. In the NL this year, a pennant winner is a 90-win team (or fewer!), but in the AL in 1998, it's a 110+ game winner.

All of which is to say that, to use the recent language of our discussions, it's a normative question that looks like a positive one, and therefore is up for many interpretations, particularly given the lack of specific contexts for measurement. Anyway, in my interpretation of this question, I try to separate the idea of his performance within the context of his team from the context of winning a pennant. Best on his own team, that's a whole 'nother Keltner question, and one I don't always find valuable, though it's just my opinion. I try to answer the pennant question as much in a positive vein as possible, but it's still arrived at normatively in that I try to answer three interpretive questions:
A) throughout history, how often are pennant/division-winning teams led by a player of X number of Win Shares (or Dan R WARPs, the measuring stick can be swapped, depending on taste)

B) what is the best expression of a player's ability to lead a team toward a pennant?

C) how does the player, using (B) compare to (A)?

I've done the legwork for the first one, and I've reported it before.

For the second one, my answer is that a player's fifth-best year is an effective representative of his typical contribution toward winning a pennant. I chose the fifth-best season for a few reasons.
-If you choose his best year, then you run into issues like Norm Cash or Adrian Beltre, where their best season is several lengths better than the second best. In other words, fluky years.
-If you take any of his best three years, you turn the question into a measure of absolute peak, and you continue to run the risk of overrating players at their very best without accounting for what a more typical year looks like. I suspect that a peak voter, particularly 'zop would disagree with my reasoning at this juncture.
-If you take a career average look at it, then you will undersell the player considerably in part because many pennant teams are led by peaking/priming players.
-If you take a prime look at it, you are getting closer, but even a simple average of top-ten seasons won't get you where you want. Al Rosen suffers a lot by such a method, for instance.

So I take his fifth-best year. It's at the end of the absolute nonconsecutive peak, it's likely close to the mean/median of his prime, but it avoids the Al Rosen and Norm Cash issues quite nicely. There's a practical reason too. Very, very few pennants are won by teams led by a player with fewer than 26 WS. That's just how it is in the WS world. Using anything much lower than 5th-best or that incorporates career or prime averages will fail to give many players any kind of rating because it will pull them down too much. And with such an expansive backlog, it's useful to have points of differentiation such as these, even minute ones.

So what about Doyle? How does he do here? About as well as a backlogger will do. Doyle's 5th-best 162-adjusted season is 26 WS. 15.9% of all pennant winners were led by a player with 26 or fewer WS. I use a 1-10 scale, and this scores as a 2 on my scale. A 2 is significantly better than the resounding No that Dan suggests. I'm not saying he's wrong/I'm right, but I will say that if you interpret the question in a different manner, then you may see Doyle in a more favorable light.

Again, normative responses that lead to positive answers to normative questions, so everyone's milage should probably vary.

Fair enough. Perhaps I went too far by expanding it to Keltner--I only intended to respond directly to sunnyday's statement that Doyle was the best position player on numerous contending teams. I think the bigger disagreement here is on my WARP's take on Doyle vs. WS's take, since my WARP hit him hard both for being a poor fielder and for playing the position at its all-time offensive high, whereas WS ignores both issues almost entirely (2B Fielding WS per season were what 1.0 lower around 1910 than the modern day? Almost no difference).

Also, sunnyday, if it takes you an hour per player to calculate OPS+ and WS above positional average, why not just use RCAP? I'm sure someone in the group can provide them to you for all the players in your consideration set (although I myself don't happen to have them).

If Doyle were to be elected to the HoM, which is of course not going to happen, his plaque could probably mention the following, all of which is factual.

"Acc. to WS Doyle was the best position player on the Giants in 1910-11-12-15 and he was #2 to Art Fletcher in 1913 and also #2 on the 1917 Cubs to Fred Merkle. The Giants won the pennant in 1911-12-13 and Doyle was elected MVP in 1912. He provided his teams from 1907-20 with a median 28 point advantage in OPS+ versus the median 2B in the league, even despite the fact that 2Bs at the time sported an all-time high EqA(?) during this period of time, and he also delivered about 9 WS more than the median 2B. DanR disagrees strongly that Doyle is meritorious."

And, could a team that had Larry Doyle as its best position player win a pennant? Well, maybe it was the fluke of the universe, but they did it twice.

As Doc says, I'm not saying he's wrong/I'm right. But you can look it up.

Not sure what your point is, sunnyday. The clear disagreement here is over how good Doyle was--if you dock him about 2 wins a year for his poor fielding and the difficulty of his position relative to its historical average (as my system does), he's nowhere close to deserving, whereas if you more or less ignore both of those factors (as WS does) he looks like a serious candidate. His rank relative to his teammates on the Giants is just a product of that underlying issue.

Of course.

My struggles with trying to get a real estimate of Bill Mazeroski's fielding led me to look at another 2b high in my comparison set, Larry Doyle.

Well, nine years later, Bill James decided to help you out! Turns out Mazeroski was a really good fielder, and Doyle was a really bad fielder. ;)

James' best defensive 2B ever, minimum 1000 games:

Rk First    Last         Yrs GG First Last Won Lost WPct 
1  Bill     Mazeroski    17  8  1956  1972 96  29  .771 
2  Placido  Polanco      14  2  1998  2011 31  12  .724 
3  Bobby    Grich        16  4  1970  1986 72  28  .721 
4  Brandon  Phillips     13  4  2002  2014 44  17  .719 
5  Jody     Reed         11  0  1987  1997 37  15  .716 
6  Dave     Cash         12  0  1969  1980 52  21  .712 
7  Jose     Lind          9  1  1987  1995 40  16  .709 
8  Frankie  Frisch       18  0  1919  1937 85  37  .697 
9  Nellie   Fox          19  3  1947  1965 90  40  .692 
10 Bobby    Doerr        14  0  1937  1951 80  36  .689 
11 Dick     Green        12  0  1963  1974 40  18  .688 
12 Ryne     Sandberg     16  9  1981  1997 70  32  .686 
13 Tom      Herr         13  0  1979  1991 54  25  .684 
14 Orlando  Hudson       11  4  2002  2012 40  18  .684 
15 Red      Schoendienst 18  0  1945  1962 70  33  .683 
16 Mark     Ellis        12  0  2002  2014 45  21  .683 
17 Fernando Vina         12  2  1993  2004 30  14  .682 
18 Joe      Gordon       11  0  1938  1950 65  31  .679 
19 Dustin   Pedroia       9  4  2006  2014 34  16  .678 
20 Glenn    Hubbard      12  0  1978  1989 50  24  .677 
21 Eddie    Stanky       11  0  1943  1953 47  23  .671 
22 Robby    Thompson     11  1  1986  1996 44  22  .669 
23 Tommy    Helms        13  2  1965  1977 40  21  .656 
24 Robinson Cano         10  2  2005  2014 43  23  .656 
25 Nap      Lajoie       17  0  1900  1916 93  50  .650 

And worst ever:

First    Last       Yrs First Last Won Lost WPct 
Rickie   Weeks      11  2003  2014 15  29  .340 
Juan     Samuel     14  1983  1998 24  37  .387 
Steve    Sax        14  1981  1994 40  53  .430 
Ron      Hunt       12  1963  1974 28  37  .435 
Larry    Doyle      14  1907  1920 56  71  .439 
Delino   DeShields  13  1990  2002 29  36  .447 
Bill     Wambsganss 13  1914  1926 42  48  .469 
Eric     Young      15  1992  2006 25  27  .476 
Dan      Uggla       9  2006  2014 27  29  .480 
Rogers   Hornsby    17  1916  1937 52  55  .483 
Tony     Bernazard  10  1979  1991 25  26  .485 
Glenn    Beckert    10  1965  1974 33  34  .493 
Ian      Kinsler     9  2006  2014 28  28  .497 
Ray      Durham     14  1995  2008 39  40  .498 
Tito     Fuentes    12  1965  1978 36  36  .499

Just for kicks, worst with 500-999 games played:

First    Last       Yrs First Last Won Lost WPct 
Alfonso  Soriano     9  2000  2009 11  25  .312 
Frankie  Gustine     8  1940  1949 13  24  .346 
Jerry    Browne     10  1986  1995 10  17  .361 
Ralph    Young       8  1915  1922 26  45  .370 
Fresco   Thompson    7  1925  1931 15  25  .372 
Jorge    Orta        8  1972  1984 14  23  .376 
Jim      Delahanty   9  1904  1915 16  25  .386 
Joey     Cora       11  1987  1998 15  24  .387 
Derrel   Thomas     15  1971  1985 13  18  .407 
Jose     Lopez       7  2005  2012 12  17  .417 
Julio    Franco      9  1985  1997 14  18  .432 
George   Grantham    8  1923  1933 25  33  .432 
Don      Gutteridge  6  1942  1947 16  22  .433 
Bill     Sweeney     6  1907  1914 17  22  .434 
Cass     Michaels   10  1945  1954 21  27  .435 
Nelson   Liriano    11  1987  1998 11  14  .436 
Gary     Sutherland 11  1968  1978 15  19  .437 
Frank    LaPorte    10  1905  1915 24  31  .441 
Billy    Goodman    14  1948  1962 15  18  .443 
Odell    Hale        9  1931  1941 14  17  .445 
Mike     Andrews     8  1966  1973 18  22  .448 
Chuck    Hiller      8  1961  1968 12  14  .448

The catcher and 1B lists are in the Ernie Lombardi and Vic Power threads.

Somehow, I missed that Bill also listed the best with 500-999 games, so I'll throw that in here too. #1 replicates what he found and mentioned in the New Historical Abstract.

First    Last       Yrs First Last Won Lost WPct 
Jackie   Robinson    9  1948  1956 31   7  .808 
Snuffy   Stirnweiss  9  1943  1951 37  11  .776 
Lonny    Frey       12  1935  1948 46  17  .733 
Jerry    Adair      13  1958  1970 29  12  .713 
Jose     Oquendo    10  1986  1995 24  10  .709 
Gil      McDougald  10  1951  1960 22   9  .701 
Jerry    Coleman     9  1949  1957 21   9  .700 
Jamey    Carroll    11  2003  2013 17   7  .696 
Scott    Fletcher   12  1981  1995 24  11  .687 
Burgess  Whitehead   9  1933  1946 32  14  .687

Doyle's offensive career had a Santo-esque vibe, in terms of packing in the production.

One had a 126 OPS+ in 14 years, the other a 125 OPS+ in 15. neither ever had fewer than 250 PA in a season. they hit the ground running, and didn't linger at the end (the worst OPS+ came from the guy who played the "extra" 15th season). Santo's offensive peak a bit better.

and yeah, fielding.

So, based on #63 & #65, Larry Doyle = Ron Santo's offense + Steve Sax's defense? That's an intriguing candidate, probably just off my ballot but definitely worth a closer look.