The new Baseball Prospectus translations are very interesting. For example, Jake Beckley translates to over 3000 hits with 440 HR's for his career...

yest wrote:

Beckley translates to over 3000 hits with 440 HR's for his career...

I'm a supporter of Jake Beckley, but note that BP has translated stats to those of a very high-offense league. Eddie Murray shows up with 644 translated HR.

McGinnity and Griffith: a study of wins above team

This is the first part of a study that I began to test the validity of the WARP evaluation of Rube Waddell. 2002 WARP saw him as superior to Clark Griffith and Joe McGinnity, despite the fact that he had a worse won-loss record, in fewer innings, than either of the others. It was also inspired by the electorate's strong preference from McGinnity over Griffith, despite very similar career records and WARP's evaluation of Griffith as the more valuable pitcher. What I have done so far is to calculate three figures for Griffith & McGinnity (Waddell is still in process):

1) Wins above team, career and seasonal. Calculated by subtracting pitcher record from team record to find winning percentage of team without pitcher. Then multiply that wp by the pitcher's decisions to see what his record would be if he were equivalent to an average pitcher on the team. This can be an interesting number, but if a pitcher's teammates are terrible or excellent, that can heavily influence this number. Since McGinnity was paired with Mathewson, I was worried that this number would underrate McGinnity, so I calculated

2) Wins above team, with team record normalized to a league-average defense. For each season, I calculated avg. runs per team, park-adjusted it for McGinnity's and Griffith's teams, and used it with team runs scored to calculate a team winning percentage via the Pythagorean method. I used this wp to generate a team won-lost record and a won-lost record for the pitchers, if they had that wp for the same number of decisions each season. This eliminates the Mathewson effect, but it left open the possibility that teams with excellent defenses would boost pitcher-records against an average defense. I didn't see a convenient way to remove that element from the records, but I thought I could at least identify the extent to which it might be at work by calculating

3) Team defensive efficiency vs. league defensive efficiency for each season of the pitchers' careers. Defensive efficiency = outs on balls in play / balls in play. Since I don't have complete batter facing pitcher data, I calculated outs on bip as (IP*3) - K, and bip as (IP*3) - K + hits - HR. This leaves out the effect of fielding errors on efficiency, and it is not independent of pitching value, since pitchers influence defensive efficiency by their defensive play as well as their pitching, but the majority of the efficiency is achieved by the fielders.

Here are the results for each pitcher

McGinnity

Chris, thanks, very interesting. It's easy enough to elect Kid Nichols, but it's a lot harder (for me at least) to rate and rank the pending "pitcher glut" than our long-standing OF glut--i.e. harder to split hairs among similar pitchers than similar position players. To some degree that is because you don't tend to see in position players quite the disparity of career vs. peak. Position players with high peaks, with few exceptions, manage to accumulate respectable career totals, too, while with pitchers you seem to find that the yin-peak is paired with the yang-career.

Having eye-balled (just) Willis, Waddell and Joss, they represent almost the perfect storm in that respect. The correct rank order of the three seems obvious on career, and obvious on peak. But once you try to combine career and peak it turns to soup in a hurry.

And I've had Griffith (but not Iron Man Joe) on my ballot, but down in the double digits. Your analysis certainly justifies my sense that they are somewhat comparable, though it does not specifically provide any reason to know whether they should be 1-2 or 15-16. One might have wished for a smoking gun that would clearly separate them, but the reality just doesn't allow for that and it's good to validate that reality.

My sense is that Waddell will not fare any better and so there will be three more pitchers vying for 15th place at best. But I certainly am looking forward to that analysis.

PRELIMINARY BALLOT

I look more for wins above AVERAGE as opposed to above REPLACEMENT LEVEL when considering a player's greatness, , along with heavily weighting C, SS, and 3B defense.

1. GEORGE DAVIS, SS ? Like Luke Appling ? VERY good offensively, average defensively.

2. CHARLIE BENNETT, C -Comp is Roy Campanella. Until at least Roger Bresnahan, only Ewing was a better Catcher. Look how far ahead he is of Clements, McGuire, Zimmer, Farrell, Carroll, Milligan & Peitz.

3. BILL DAHLEN, SS ? Modern comp is Joe Sewell, only was EVEN MORE VALUABLE defensively.

4. HUGHIE JENNINGS, SS ? Best comp may be Lou Boudreau. Great fielder and great hitter for a SS. Only drawback is shorter than normal HOM career, but I don?t see how not having 5 years of Jeff Blauser career performance added on makes him NOT a HOM?er . MVP type years 1895, 1896 & 1897.

5. CUPID CHILDS, 2B - Hitting value almost identical to Hardy Richardson, AND played close to 13,000 innings at 2B. Comp is somewhere between Charlie Gehringer to Stan Hack. Now that McPhee and Richardson are in, Childs should be in also. Only MVP type year was 1890 in weak AA.

6. JOHN MCGRAW, 3B ? Comp is? no one, as there hasn?t really been an infielder who was this good offensively but played so little. Still has to rank as one of THE best 3Bmen of the 19th century. Would be #1 or #2 on this ballot if he had played a little bit more.

7. HUGH DUFFY, CF ? Strong comp with Kirby Puckett. Note quite the hitter that Mike Griffin was, but played longer. One MVP Year - 1894.

8. PETE BROWNING, CF/LF - Hits like Joe Jackson, fields like Greg Luzinski playing CF. Still has one of the highest Win Shares/Year for the 19th century. Possible MVP in 1882, 83, 85 & 90 - I count that quite a bit in moving him up the OF glut pack.

9. JIMMY COLLINS, 3B ? Ken Boyer offensively, right with Williamson and Cross defensively.

10. JAKE BECKLEY, 1B ? Comp is Tony Perez or Sam Crawford. Never had an MVP-type year, as Black Ink score is only 1!! BUT, his Gray Ink is 165, so he?s the poster boy of being very good for a very long time to be HOM worthy.

11. JIMMY RYAN, CF/RF ? Comp is Jimmy Wynn. Good hitter, and average fielder who played quite a bit of RF. MVP type year in 1888.

12. GEORGE VAN HALTREN, CF ? Almost identical to Ryan offensively, but loses out due to fielding and no MVP type year.

13. JOE KELLEY, LF/CF ? Best comp may be Bob Johnson. Falls right in to the middle of the outfield glut. Never really had an ?MVP? year, yet peak was better than Harry Stovey?s.

14. HARRY STOVEY, LF/1B - Comp is Albert Belle, only with baserunning ability. MVP type years in 1883, 1888 & 1889. Decided Stovey should be ahead of Griffin, Thompson & Tiernan this time.

15. TONY MULLANE, P ? Think I?ve missed the boat on Mullane previously. Only 4 Win Shares behind Galvin (399 vs. 403) in 1,400 fewer innings (but still a substantial 4,500 innings pitched). Peak almost identical to Pud?s (183 WS vs. 182 over 5 years). ERA+ of 118 vs. Pud?s 109.

LEFT OFF THE BALLOT:

Out of curiosity, why does everyone rate Thompson so low, despite his staggering RBI totals.

Well, you might say Thompson's rating is a point of some contention, as you'll see if you read through, say, the 1914 ballot discussion thread.

Whether voters rate Thompson low or high, few place a great deal of weight on the rbi totals. Thompson was an outstanding hitter, but his tremendous rbi totals are substantially influenced by context. He hit behind Billy Hamilton and Ed Delahanty, two of the best hitters for on-base-percentage of the era, in a era, esp. 1893-1896, strongly favoring hitters, so the most staggering thing about his RBI totals is the number of opportunities he had. If you look at offensive statistics that are adjusted for context, like OPS+, equivalent average, batting win shares, or batting runs in the WARP system, while Thompson looks outstanding, he's not "staggering."

Voters who rate Thompson highly tend to value him for his outstanding hitting. Voters who rate him lower (as I do) have concluded that, in comparison with the other top eligible players, his short career, mediocre defensive value, and brittleness (he missed a lot of games in several seasons) lead to his being far from the most meritorious player available for induction into the Hall of Merit. The win shares analysis of his value tends to support this lower ranking; the WARP analysis of his value supports a middling to high evaluation of Thompson, relative to the other players now eligible.

Paul Wendt:

Was the Washington Olympic club from 1867 through 1870 considered a top-line club? Thanks!

One more tidbit from the NBJHBA:

In the Ozzie Smith comment, James introduces these two lists, without further explanation:

The top five shortstops of all time, rated as offensive<i> players, are:

Just a few quick observations about raw WARP-2003 numbers for players we have considered/are considering in 1915:

Top 10 all time by raw Warp 3:

1. Cap Anson - 142.7

<i>WS W3 Rookie Name-Pos (Died)

prelim....

1. GEORGE DAVIS

1915 Preliminary Ballot

Dahlis/Dalen on top. Everybody moves down; no other new candidates are close to the ballot. I haven't fully assimilated the new WARP evaluations, but since I base my pitcher-analysis on WARP data, I've started with that. Before the new WARP came out, I was working to assess the wins-above-team, wins-above-an-average-defense, and the quality of the fielding behind each of the main pitching candidates. The preliminary results of that analysis, and the new WARP numbers, have resulted in some changes in pitcher rankings.

A few notes on the numbers. Win shares are adjusted for fielding, season-length, and league quality. Pitchers' win shares are derived from WARP ratings, not official BJWS. In ranking infielders against outfielders and pitchers, I give them a 10% bonus to career value.

1) George Davis (n/e) 481 CWS. Total peak 89. Peak rate, 93-97 = 33.08. Among top 5 position players twice, at/above avg. WS in 14 seasons.

KJOK:

Someone asked for Jim McCormick's Wins Above Team. Total Baseball reports it as 56.6. However, TB reports McGinnity as 32.7 (which is about 6 wins higher than Chris and some others have calculated by hand) and Griffith as 46.1.

Chris:

See if you think there are parallels between your defense-adjusted wins above team analysis for McGinnity and Griffith and the conclusions I reached about their Linear Weights performances,taking into account BP's NRA and DERA numbers which purport to evaluate the strength of the defense behind the pitchers.

Basically, I calculate LWTS for each season, using the park factors in the back of the All-Time Sourcebook. LWTS is by its nature weighted by IP.

I then adjust each season's LWTS by DERA divided by NRA. I don't apply the DERA/NRA to the final LWTS total; it is imbedded in the formula so it only applies to the portion of the LWTS formula that compares League ERA to Pitcher ERA. I call the result the DERA LWTS.

CAREER:

McGinnity has 192.7 LWTS (with 1899 as his best season) and 183.6 DERA LWTS (with the 1899 and 1904 seasons essentially equal). This would indicate that for his career, on balance (since LWTS takes into account IP), McGinnity's numbers are inflated slightly by playing in front of above-average defense. This works out to 16.8 LWTS per 300 IP and 16.0 DERA LWTS per 300 IP. (Note: McGinnity's bad 1907 season messes up his career totals, because it subtracts 27 LWTS).

Griffith has 231.3 LWTS (with 1898 as his best season, by far -- TB shows it as the 49th best pitching season ever) and 231.0 DERA LWTS --essentially the same as LWTS. This reflects an essentially average defense behind Griffith over his career. A LWTS (or DERA LWTS) total will of course be higher the longer a player plays. However, Griffith's numbers work out to 21.0 LWTS per 300 IP and 20.9 DERA LWTS per 300 IP.

PEAK:

McGinnity's best three years total 135.7 LWTS and 113.3 DERA LWTS.

John Murphy asked:

10. JAKE BECKLEY, 1B ? Comp is Tony Perez or Sam Crawford.

multiple choice?

I'll take the kid from Wahoo for $1,000 and a HoM slot for himself to boot. :-)

Bill Ryczek, When Johnny Appendix B, lists the Olympics from 1868. He names three players acquired by the club in 1868, who "joined New Yorker Davy Force" from 1867 to make the team "surprising strong"

As always, thanks! Force happens to be the guy that I was focusing on, BTW. His HoM credentials look better to me with that information.

Bill Ryczek, When Johnny Appendix B, lists the Olympics from 1868. He names three players acquired by the club in 1868, who "joined New Yorker Davy Force" from 1867 to make the team "surprising strong"

As always, thanks! Force happens to be the guy that I was focusing on, BTW. His HoM credentials look better to me with that information.

John, what more do you know of Force pre-1871? He's not close to my ballot right now, but I haven't been giving him any credit pre-1871. If he has a pre-1871 career similar to Lip Pike's, I'd want to give him another look.

Paul Wendt suggested a study of Bob Caruthers for wins above team and wins above average defense. I haven't yet run a study of the extent to which his defenses were above average as a check on the wins-above-average-defense, but here are the wins-above-team and wins-above-average-defense numbers, season by season:

Season -- IP -- W-L -- W over team -- W over team w/ avg. defense

Marc wrote of Davis and Dahlen:

KJOK:

redsox, Davis was a superior fielder to Dahlen???

John, what more do you know of Force pre-1871? He's not close to my ballot right now, but I haven't been giving him any credit pre-1871. If he has a pre-1871 career similar to Lip Pike's, I'd want to give him another look.

I have him now as borderline with the pre-1871 seasons. I don't think he's going to make my ballot, but his case was better than I thought it was originally.

Redsox,

You wrote: <i>If 3rd base was the more difficult position, then Davis' 527 career games at 3rd to Dahlen's 223

Two thoughts about the new WARP numbers: (1) They appear to have been completely overhauled, so if you've been using WARP in your analysis, you ought to at least consider the new WARP numbers and how they may affect your rankings and (2) I think it's best to treat the new "translations" feature as a toy right now, rather than a true comparative measure -- at least until we have a better idea of how they are derived (apart from the BA/OBA/SLG normalization).

For illustration purposes (and because there have been some comparisons on this thread b/w McGinnity and Griffith): The translations subtract 89 wins and 21 losses from McGinnity's record (reducing his win pct. from .634 to .565) and subtract 84 wins and 45 losses from Griffith's record (reducing his win pct. from .619 to .602). Not all of the old timers are adversely affected in the translations from a win pct. point of view. Galvin gains about 20 win pct. points -- though he loses 181 wins from his raw total. :O

Maybe a discussion thread on WARP 2003 -- how it works, how it compares to WARP 2002, what its value is for us -- would be useful, so those who want to chew over the system have a place to keep the comments on it together?

Most people here have been ranking Davis and Dahlen right next to each other, but there are a few exceptions, notably redsox2004. As near as I can tell, redsox2004 was reacting strongly to two percentages: fielding percentage at SS, and batting average. I have nothing much to add to Chris Cobb's post #156, other than that I do believe that Dahlen was an outstanding fielder. I also note that errors were dropping sharply throughout the 1890's, and Dahlen was a shortstop for several years before Davis was - that particular timeline makes a little bit of difference.

But one should not think of Dahlen as a weak hitter. Davis once led a league in RBI - and Dahlen once led a league in RBI. They had the same number of career triples, and Dahlen had more career HR - they both had some pop in their bats. Dahlen once led a league in road home runs. (OK, small sample.) Davis's offensive advantage on Dahlen lies entirely in batting average, and that's partly compensated for by Dahlen's walks and HBP.

Over their careers, Dahlen had about 240 more plate appearances than Davis (about 2% more). In those plate appearances, Dahlen had 194 more BB, 66 more HBP, 19 fewer sacrifices, 176 fewer singles, 38 fewer doubles, and 11 more HR.

All in all, it's a small offensive advantage for Davis, which increases a little once we account for park effects - but it's not a night and day difference. In their prime years, Dahlen and Davis both hit well enough to play outfield. Given that Dahlen had more defensive value just by virtue of playing more SS and that Dahlen was probably the better defensive shortstop, I don't see any way to put any space at all between them.

Paul Wendt:

I hate to be a pain in the butt, but (no pun intended :-) do you know what positions Force played for the Olympics? I know he started out as a catcher in NY. Thanks!

BTW, glad to have you on board, Sean! I missed your introductory post above.

Second base fielding appears to have had big changes with WARP 2003, along with lowered pitching values.

Besides McPhee, the following 2B are over 100 raw WARP 1 (equal to or better than Kid Nichols, the only pitcher over this number).

Fred Pfeffer - 108.9

JoeD,

Wasnt Tango talking about replacement level? If so, I dont think those have changed.

These big "revisions"in the number's just reinfoce my natural banker's suspicion of new-fangled metrics. I think we can measure who was the better hitter and (with difficulty) who was the better fielder, but the "magic number" approach that adds the two just doesn't work. For one thing it's not at all obvious that value is 50% hitting, 50% pitching/fielding; one could imagine a game in which differentials in hitting were all that mattered and pitching/fielding was irrelevant (cricket in the 1930s, on their perfect over-prepared pitches). One could also imagine a game in which pitching was all-important and hitting was irrelvant (cricket on a 1950's sticky or dusty wicket.) 50-50 is a GROSS assumption, and makes the reuslts thoroughly unreliable.

I was just going to respond to Ron's post and say what Jim Spencer said and then saw his post. And ditto karlmangus' post. Ron wrote:

> Besides McPhee, the following 2B are over 100 raw WARP 1 (equal to or better than Kid Nichols, the only

And while it pains me to say so, yest's preference for more traditional numbers is hard to dispute when you see something like this!

At any rate, there should be a detailed explanation as to why the changes were made. Without that, how can anyone know if the new Warp is better than the old version?

Very slight clarification to the raw WARP 2003 numbers. All of the 2B I referred to in # 163 are below Nichols' 117.5 raw WARP 1. I accidentally looked at Kid's WARP 2 number when referring to him (101.5).

I have a rudimentary spreadsheet with WARP-2003 numbers for all players we have elected and strong considerations. The number of 2B over 100 raw WARP 1 really struck me. In all, 7 2B are over this number, and Claude Ritchey is close (97.7).

Joe asked about 3B:

Lave Cross - 125.3

I will stick with my "OPS+ and some other offensive stuff, then add or subtract for defensive skill and position" philosophy for now.

I still am noting G Davis as a good fielder and Dahlen as an even better one, for instance. But I just can't 'blind faith' any of these early-era defensive numbers right now...

If fielding is so important relative to pitching in the '90s, what of the '70s? If McCormick and other pitchers of the '70s and early '80s have dropped down, who has moved up in their place? Well, I can go look it up myself. And of course, we've already elected G. Wright and Barnes and Sutton and McVey and J. (D.) White. But I wonder if any other IF from that era jump up?

Raw WARP 3 2003 Careers over 70 (over 65 and electees immediately follow)

P:

There are many notable differences between the baseball of the 1890's and baseball of the 1990's. One is the distribution of outs within the defense. (See an earlier post for the rest of the numbers.)

-------- SO's C-SO Infd Outf -- SO's C-SO Infd Outf

Viewed through a modern DIPS-oriented lens, this is a significant shift in value from pitching staff to infielder. To shift some of that value back to the pitcher may require a finding that top 1890's pitchers had more of an ability to influence Balls-In-Play outcomes than their modern counterparts appear to have, or that there was a wider variance in that ability. (That project would be beyond my limited statistical ability.)

Interestingly enough, one of the most provocative claims of the fielding analysis system under discussion right now on the Primer is that a pitcher's ability to induce pop-outs, which are close to automatic outs on balls in play, is one of the chief ways in which pitchers can and do influence batting average on balls in play. This claim has not been tested on evidence from the early game, as far as I know, but, since Bill James has shown that not only infield put-outs but infield pop-outs were much more common in the early game (they have decreased steadily from the 1870s to the present), if this claim proves true, it would reveal the mechanism by which some of the defensive value now assigned to infielders would flow back to the pitchers.

I'd probably toss WARP 1 at this point, and focus more on WARP3.

I don't know. Wouldn't focusing on WARP3 only compound the problem? If someone isn't comfortable with where WARP1 comes from, it would be impossible to be comfortable with where WARP3 comes from, since it is WARP1 with some mysterious adjustments for league quality, adjusted for season length.

I don't know. Wouldn't focusing on WARP3 only compound the problem? If someone isn't comfortable with where WARP1 comes from, it would be impossible to be comfortable with where WARP3 comes from, since it is WARP1 with some mysterious adjustments for league quality, adjusted for season length.

Not exactly. WARP2 is adjusting not only for "league quality" but for things having to do with "opportunity," I believe. As WARP sees it, second basemen have a tremendous defensive value in the context of the 1880s game, but in an "all-time" context, much of that value evaporates. Fielding value is adjusted for all-time differently from hitting value or pitching value, and different elements of fielding value, I believe, are also adjusted at different rates. Therefore, it does not replicate the massive value of early infielders in its all-time assessment. So if you want an answer to the question -- how good would this guy be in the game of baseball as I know it now -- then WARP3 offers an answer.

I'm not saying that this answer is what I'm looking for in trying to understand a player's meritoriousness (it's not), but I think this is what WARP2-3 is designed to show, and it does offer one defensible way of bringing what are apparently outlandish defensive values into a more reasonable scope.

Based on the numbers Ron posted (thanks, Ron) I am comfortable in saying that the timeline is way too steep. I don't doubt that replacement and average level are much higher in the '90s than the '70s, and I understand that the kinds of numbers that G. Wright and Barnes et al put up reflect the poor competition. But in effect the timeline treats them (proportionally) like average or replacement level players themselves (in terms of the discount). There is no way to assert that the outliers (the greats) should be discounted the same as the rank and file.

This is in response to the '03 numbers. For all I know, the '02 numbers were just as objectionable in this respect.

I'm not saying that this answer is what I'm looking for in trying to understand a player's meritoriousness (it's not), but I think this is what WARP2-3 is designed to show, and it does offer one defensible way of bringing what are apparently outlandish defensive values into a more reasonable scope.

You are right. My comment was a little too narrow by mentioning only league quality. However, the underlying thought is that when WARP2 is adjusting WARP1 for league quality, defensive value and anything else, we don't know the methodology for getting there. So, if someone is uncomfortable with WARP1 b/c there's no way to figure out the formula, so should they be uncomfortable with WARP2 and WARP3 which add more unknown calculations.

Agreed that WARP2 and WARP3 are one defensible way of bringing what are apparently outlandish defensive values into a more reasonable scope. However, while I'm sure it is defensible, I wish someone would actually defend it with the methodology. But I certainly understand the proprietary nature of the formula.

I'm not really complaining. Just wishing we knew more. Also, I get nervous about people jumping on new metrics when there's no way for us to understand them.

Does anyone have new WARP3 numbers for Tip O'Neill and Harry Stovey?

Does anyone have new WARP3 numbers for Tip O'Neill and Harry Stovey?

Er... make that just O'Neill

Here you go, Brad:

http://www.baseballprospectus.com/cards/o=neiti01.shtml

Rough partial aid in calculating the strength of the AA - or arguably, a case that we are underrepresenting them. Take your pick.

HOMers thru 1915 in CAPS, those among the top 30 vote-getters in 1914 are listed. Minimum 10 G

Note how the HOM names are virtually unchanged for the entire decade:

1882

The single most outrageous year on Patrick's list is 1894, at .8 runs more than any other year. One interesting thing about 1894 is that while it generated it's share of "with the flow" extreme seasons (Duffy's .440/.502/.694, Hamilton's 192 runs), it also enabled a relatively extreme "against the flow" season: Amos Rusie pitching 444 innings at an ERA+ of 189. (It's an actual ERA of 2.78, but the context makes it extreme.) As for 1908 - we'll be talking about Ed Walsh soon enough. And (as a relatively extreme "against the flow" performance), Honus Wagner.