>6. Not-timeline adjustments -

1. Heinie - traditional nickname for "Heinrich", the German "Henry", "Enrique", etc.

As "Hans" is the traditional nickname for "Johann" (John).

I'm not sure exactly what the encylopedias and bbref use as the sources of "real" names, whether it's birth certificates or what. (It probably varies case by case.) I was a little bit surprised when I ran that list of players called "Heinie" that they were all officially "Henry" rather than "Heinrich". It's likely that the official records are more anglicized than the names the families used at home.

So the exact same process that turned "Henry Clement Peitz" into "Heinie Peitz" should also have turned "John Peter Wagner" into "Hans Wagner." Which it did. But where did that weird spelling "Honus" come from?

Johannes to

Chris, can you provide us mere mortals with the nutshell version of exactly what the WARP2 adjustments include/do? The glossary at BP is most unhelpful: "adjusted for level of difficulty." Certainly the adjustments seem to be HUGE (and I ignore them, instead merely adj WARP for season length) as it seems most of us do. Can you tell us more?

Marc, unfortunately in this as in all other matters I am as merely mortal as the next guy. I have no idea what kinds of difficulty adjustments WARP2 lie behind its cryptic description. Some other mere mortal who owns the BP 2002 (iirc?) that introduced the system might be able to offer us more detailed information from the book.

McGinnity: pitched in the relatively weak 1901 AL. Otherwise in the stronger NL.

Actually, if you believe BP's DERA "adjustments", Iron Man Joe flowed against the tide, moving from the weak AL of 1901 to the weak NL of 1902-5. The AL did an excellent job of raiding a large number of quality players from the NL during 1901-03.

Here is an All-Star team of players that moved from the NL of 1901 to the AL of 1902.

Ca McGuire/Schreckengost

I agree with this, but after adjusting for defense (the "DERA"), the career numbers are nearly identical (3.78 for McGinnity and 3.83 for Caruthers). This adjustment shows that properly adjusted for defense, Caruthers and McGinnity were about equal as pitchers -- Caruthers doesn't blow him out of the water.

Sure, as a raw number. Work that into a standard Linear Weights calculation and you'll start to see the difference because of the additional 600 IP from McGinnity.

McGinnity's NRA and DERA are about equal, meaning he didn't get a boost in his real ERA by pitching in front of a notable defense. Caruthers' DERA is about 10% worse than his NRA, indicating that his real ERA is probably better than it would have been in front of an ordinary defense. If you adjust their real ERAs for this effect, and compare to league averages (weighted by IP) -- essentially, the Linear Weights calculation -- McGinnity is quite a bit better.

Preliminary ballot for 1914. The biggest change from previous ballots is that I've

TomH, interesting point about the large number of big OPS+ seasons recently. The decline in OPS+ seasons over time is thought to "prove" that the play of the average and/or replacement players was getting better. Is the average or replacement player now getting worse? Is the pool finally shrinking, as a practical matter, because of competition for the best athletes among the various different sports--football, football, buckets, skateboarding, I don't know..? Have performance enhancements upset the delicate balance of hitting and pitching? Or is it just a small sample?

By the way Matt B, while you and I are arguing (friendly) over Caruthers and McGinnity, Devin has ranked our ballots as most similar. :)

I liked the approaches I used to compare Kelley to Thompson, so I'm going to do the same for Duffy and Kelley, with a more extensive WS-based comparison as well. (Warning: this is long!)

Duffy has 11 seasons as a regular, 1889 to 1899, so I'll start by comparing those to Kelley's 11 best seasons, 1893 to 1903.

OPS+ over 11 years

Kelley -- 143.1 over 6125 PA in 1360 games

Jim, best wishes for a safe arrival over the weekend!

Are you sure that imminent birth isn't influencing your judgment in favor of new guys :-) ?

I think Duffy's 1891 is a key to unlocking his resting place on ballots.

Kelley and Duffy look close to a toss-up here, but if one considers that Duffy is the superior fielder and that WS undervalues fielding, Duffy might be favored.

...and don't forget that WARP overvalues fielding.

Are you sure that imminent birth isn't influencing your judgment in favor of new guys :-) ?

lol

Can someone please explain 1891 AA to me?

It looks like at least half of the stars who jumped to the PL in 1890 and made it the superior league (Brouthers, Duffy, Van Haltren, King Kelly, Lave Cross, Larkin, Charlie Buffinton, Weyhing, Bill Daley, Darby O'Brien, Ben Sanders) went to the AA in 1891 rather than the NL. Add in the horde of rookies and non-PL'ers to make the jump or started there (Abner Dalrymple, Patsy Donovan, Clark Griffith, Paul Hines, Hughie Jennings, Denny Lyons, Wilbert Robinson, Jack Stivetts) and it looks on paper that the AA should have been dominant that year.

I'm not denying that the 1891 discount is wrong, but it just doesn't pass the smell test for me.

With so many stars, what accounts for the huge difference in quality?

(Memo to self: Cut down on hyperbole this month.)

EricC #119

Thanks, Tom!

Your info on Kelley's WS rate per PA makes me wonder about win shares and rates. It would seem to me that, since win shares are derived from games played, that for position players, at least, the rate ought to be given in terms of games only, for two reasons. First, the process of calclating batting win shares normalizes for higher offensive contexts, and thus normalizes rates of plate appearances. Calculating WS per plate appearance, then, would tend to favor players from low-offense eras. Second, since fielding WS are not tied to PA at all, they should not be expressed in a PA-based rate.

All that seems clear enough to me concerning position players, though it's possible my reasoning is flawed somewhere in there. Now, two questions:

1) Are the game and the season also the appropriate units for win-share rates for pitchers, as a way of combining defensive innings and offensive opportunities?

2) How is it fairest to deal with players who pitch and play other defensive positions during the season? Bob Caruthers is the poster-child for this issue. Indeed, he's the poster-child for the difficulties of assessing most everything: pitcher and good-hitting position player, pre-1893 pitcher, weaker league, shorter schedule. I'd like to get a better handle on what to do with him, (and I'm not forgetting about Jim Whitney and Tony Mullane as pitcher/position player combos) so I'm looking for models of how best to do it.

except one, had a corresponding change in performance of players who stayed in the league (presumably because the players added in expansion/lost in contraction were weaker than the established players). The exception is the NL expansion in 1879. I've concluded that the NL before 1879 did not have all the top baseball talent.

Except the NL before 1879 had the vast majority of the mega stars. This is what we should be looking at.

I have no doubt that there were more good players in the 1880s and 1890s compared to pre-1879. However, those players are not the ones we would be looking at as HoMers anyway. They in no way change the fact that McVey, the Wright brothers, Pearce, Barnes, etc. were the cream of the crop of their era (it does affect the level of their dominance, of course).

My basic point is if the NL (or the best team s from the 1860s) were to have had the best possible players pre-1879, the top guys (IMO) would still have been the same guys that were recognized as the best anyway. Therefore, they should not be excluded from discussion.

In #102, Ron gave us these numbers:

<i>Current HOM: If McVey Elected

P - 23.9% P: - 23.2%

The difference is likely that C (Deacond White and Cal McVey) and 3B (Ezra Sutton) have many of their PAs taken up by NA players. Going by season would bump them up some, I'd say.

Pitcher win shares are virtually ignored by this group. However, it seems to my rudimentary mathematical brain that some sort of adjustment could be made to fix era discrepancies, similar to the adjustments made to 1870's short seasons.

As a test, I ran the following:

1. Take the raw BJ WS from the book, include pitching, fielding and hitting.

<i>The 28 HoMers were regulars in 408 total seasons, that's 14.6 seasons for the average HoMer.

The positions break down like this:

P - 20.2%

Ron, thanks for your model and your musings!

So essentially what you are doing is looking at how many win shares the pitchers in question would have earned had they earned WS at the same rate that they did in their careers, but had only been able to start 35 games a year during each season of their career? That yields some numbers that look like they fit with our scale, but I'm doubtful that they provide a fair reflection of the pitchers' value.

Some thoughts in response to problems you identify:

1. Raw WS, how do we work with the Caruthers-type players and their batting at other positions?

Yeah. I'm thinking this through myself, and the one thing that seems clear is that some estimate of the batting WS Caruthers earned while playing other positions needs to be subtracted from his total before his rate is estimated, adjusted separately, and then added in at the end to create a career total.

2. Raw WS, are they somehow biased towards pre four man staff pitchers? This is the reason I ran Tannehill, because McGinnity scored so low.

Yes, they are, though the calculation you run eliminates most of the manifestation of that bias. Basically, the WS system assumes that the percentage of defensive value attributable to pitching remains constant throughout baseball history. But it hasn't. Early pitchers were able to pitch more because it was somewhat less physically demanding (_much_ less prior to the advent of overhand pitching), but they had correspondingly less impact on the team's defense. So raw pitching WS overstimate the value of early pitchers, though they are more biased towards early pitchers who pitched for good teams. When you cap the pitchers' usage at 35 games per season, you eliminate the excess. However, the result that you obtain is a scaled down version of a result that is arrived at in raw WS by a misestimation of the pitchers' rate value. Therefore, there's too much noise in the system for me to trust its results.

That said, McGinnity scores exceptionally low, I expect, in part because he did a lot more relief pitching than any of the others, so his ratio of innings to games, and therefore of WS to games, is lower than for the other pitchers. You could eliminate this problem by dividing his IP by 9 and using that number in place of his games total. That addresses the next question, also, I think:

3. Do we use G, or GS? If GS, what to do about relief appearances?

4. My numbers just focus on career, someone with a more powerful spreadsheet could use this formula to actually calculate each season's data, then add them up, rather than to do a multiplier of seasons.

Hmm. It could, but I'm not sure that would increase the accuracy of the results.

re: McGinnity finishing far below Griffith in Ron's calculation.

Both regularly pitched more than 35 games per year, but McGinnity was generally near the top in innings pitched, while Griffith was not.

In context, it might be more realistic to limit McGinnity to 35 games, but knock Griffith down to 30 or so to accurately reflect their respective values.

Or, more specifically, pro-rate IP per season into 36ths.

Give the IP leader for the season 36 starts, and give everyone else x/36ths depending on how many innings they threw that year.

Give the IP leader for the season 36 starts, and give everyone else x/36ths depending on how many innings they threw that year.

IP leader is a little too unstable for this purpose - you don't want other players' fortunes to rise and fall with the career of an outlier like Ed Walsh. But you could use 5th best or 10th best or something like that and it would work.

Except the NL before 1879 had the vast majority of the mega stars. This is what we should be looking at.

Looks like I opened a large can of worms with my last post. I am actually a strong supporter of great pre-1880 players and regret if I gave any ammunition to those who want to disregard their accomplishments. My comments were directed toward how I fixed a subtle problem in my own rating scheme. I assume that there is a bell curve of baseball accomplishments for each season for the entire North American population , and that the players in the major leagues fill the far right tail of the distribution. When I fit to bell curves under these assumptions, there are more extreme outliers in the pre-1879 period than there ought to be. There are a number of possible explanations, but a simple one is that the top talent did not fill the top leagues. Evidence for this hypothesis is the effect of the National League raiding the International Association in the winter of 1878-1879. Even though the NL got larger, the performance of the established players did not get better on average, suggesting that the players coming in from the IA were just as good as the established NL players. For those who think that the mega stars were all in the NL anyway, I point out that HoMers Hardy Richardson and Pud Galvin were in the IA in 1878. In any case, I fixed my extreme outlier problem by arbitrarily assuming that

Paul Wendt- Here are my NL league factors for each year. I've decided against giving comparative league factors for other leagues, because the relative values between leagues are much less reliable. Many caveats apply to these numbers. First, they are based on the limited set of players in my database, rather than a full set. Second, the before and after same player performance comparisons used to find these numbers suffer from the exclusion effect: it automatically excludes cases in which a player was too bad in a season to be a regular. Third, the ages of the players were not taken into account. Fourth, it was assumed that there is a single adjustment factor that can be applied equally to all players. Finally, the numbers are only intended to show relative strengths in consecutive seasons, and must not be used to compare league strengths in different eras. The factor was arbitrarily set to 100 for 1881 and 1899.

The numbers given are adjustment factors for win shares. For example, the NL league factor is 95 in 1902 and 100 in 1903. Thus, an NL player with 20 WS in 1902 will have 19 AWS, and be equivalent in performance to an NL player with 19 WS in 1903. A higher league factor indicates a stronger league.

NL league factors 1876-1908:

1876: 91 1877: 99 1878: 96 1879: 97 1880: 95

Thanks, Eric.

Dick Cramer used data on Batter Win Average (his own measure) for all player-league-seasons with 20 or more Plate Appearances, and estimated average batting skill for each league-season. He found a slight decrease in quality, 1876-77. [BRJ 1980; reprinted in Thorn & Palmer The Hidden Game of Baseball, especially p132.]

For those who think that the mega stars were all in the NL anyway, I point out that HoMers Hardy Richardson and Pud Galvin were in the IA in 1878.

I said the NL had the vast majority of the best players, not that they had them all. King Kelly was another one that played in the IA, but, like Richardson, was essentially in the same situation that a young player in the AAA is today. In other words, they were "mega stars" in the same way as Mantle, Schmidt, or A-Rod were in the minor leagues. Up-and-coming, but not really the best at the time.

I concede Galvin is the exception during the 1870s.

Michael Schell used data on Batting Average for all player-league-seasons with 200 or more At Bats, and estimated average batting skill for each league-season by a pure measure of dispersion, the atbat-weighted standard deviation of BA (or of ballpark-adjusted BA).

I said the NL had the vast majority of the best players, not that they had them all.

Sorry that I twisted your words. Perhaps comparing King Kelly and Hardy Richardson's tenures in the IA with more recent players' stints in AAA is the best analogy. I'm not sure that it's a perfect analogy. I just reread Bill James' essay in the NBJHBA on the old minor leagues, and how they were independent leagues with players chasing pennants, not farm leagues with players trying to make the majors. Almost certainly, some "major-league-level" talent was not in the major leagues. While (with the exception of the Negro Leagues) I can't foresee voting a player into the HoM who never played in the major leagues, I think that it is valid to consider minor league play in borderline cases. What I realized is that there's a flip side of the argument, too. If Pud Galvin's 1878 IA accomplishments help his case for the HoM, then the accomplishments of the 1878 NL players are slightly less than they seem (emphasis on slightly), because they didn't have to face Pud Galvin.

Here's my premise: Rate 8 contemporary OFs by OPS+ (adding Keeler since he's eligible next year). I'll list all seasons with at least 300 AB, but put an asterisk if fewer than 400 AB, or if in AA.

Here's my premise: Rate 8 contemporary OFs by OPS+

How do the top yearly OPS+ ratings of the 1890s glut stack up against those of the HoM OFers Burkett, Delahanty, and Hamilton, and of Elmer Smith?

By most measures (OPS+, total win shares), Griffin ends up near the bottom of the glut, which makes him seem like a long shot. For some reason, he does very well in WARP measures.

BURKETT 81 62 57 54 50 46 44 40 34 30 26 24 24 16 15

BURKETT 81 62 57 54 50 46 44 40 34 30 26 24 24 16 15

Problems that I see.

>5. Joe McGinnity - 269 WS, 465 G, 20.25 aWS/35, 10 S, 202.47 aWS

Billy Hamilton has to be considered part of the outfield glut as well.

So, including Kelley and Keeler, we are up to 12 outfielders we are considering, with Burkett, Delahanty, and Hamilton already in and 9 wanting to join them.

In my mind, Burkett and Delahanty were the superstars, and Hamilton was already the top of the glut.

The HoM is already 25% outfielders. I have no problem slotting in a Thompson or Kelley or Duffy as players with high absolute skills. I do have a problem, though, with slotting the ninth or tenth best outfielder of the decade above the best first baseman (Beckley) or the best catcher (Bennett) or the best AA players (Stovey, Caruthers) or the best black player (Grant) or the best infielder (Childs, Williamson, Collins).

Each additional 1890s outfielder we induct is another step down a slippery slope. Welch and Caruthers were sui generis, Griffin and Van Haltren were a dime a dozen.

using the same methods, adding Hamilton, Delahanty, Burkett...

Here are the 15 outfielders that are -- or soon will be -- eligible for election, ranked by WARP-1

Fred Clarke -- 136.3

MattB, the real answer to your question--where should these glut OFers rank on our ballots--is: How do they compare to the non-OFers who are available? The fact that they appear in your post makes them "special." The guys who are not "special" didn't appear in your post at all--Fielder Jones and a hundred more. If we have to have 15 names on our ballot, I'd rather have one of these than a SS who is not as good as they are!

PS. Where do Stovey and Bennett fit into that list?

Marc,

When ranking players, I look at a number of things: absolute numbers, where they rank against their peers, value over replacement, value over average, etc.

I think it is entirely possible -- and likely -- that some eras will have a number of great players at one position, and none at another. I also think it is possible -- and likely -- that various stats aimed at evaluating players that no one who made the stat was alive to see will mis-judge numerous players.

When trying to decide how to rate players, I sort of create a mental bell-curve. If the player is at the far end of the bell-curve, away from the pack, he gets highly rated. But what about a player who is only 2nd or 3rd best at his position? I compare him to his peers.

When I see a big glut, I have to weigh what is more likely -- (a) a dozen outfielders were among the top 20 players of the decade, or (b) statistics are systematically over-valuing outfielders. I think that the sheer number of outfielders who are dominating the tops of the lists is strong evidence that (b) is at least a strong factor here.

Stovey would have been in the middle of the list (108.2) and Bennett just off the bottom (89.9). But they both tower over their peers. The best SS of the 1970s may be only one Win Share better than the 2nd best and 2 Win Shares better than the 10th best. That player doesn't get my vote for being best. Similarly, the fifth best that towers over the 6th best gets my vote, because that is evidence that there were 5 greats in the era.

What I keep coming back to is: If what Bennett did wasn't that special and should be placed below Kelley on my ballot, why aren't there are as many catchers with stats similar to his as there are outfielders with stats similar to Kelley's?

So why is Tiernan great on OPS+ and forgettable on WARP-1?

I agree with MattB that infielders have been receiving less support than they should. Evidence suggests that either the fielding value of infielders is being underrated or that playing conditions tended to shorten the careers of infielders, or both. Greatness needs to be measured in part by position, which establishes the average value of a player at that position and influences the maximum value a player can achieve.

However, I must disagree firmly with MattB on the following point:

<i>

Howie,

I think you are likely to find most of your answer in the "Games Played" category, rather than the Defensive numbers. (I'm sure defense plays some role, though, and between a dozen offensive clones, defense could make the difference . . .)

But OPS+ doesn't care how long you were around.

WARP (like WS) is essentially a counting stat.

Tiernan had 6716 plate appearance.

Compare 6832 for Griffin, 7827 for Duffy, 8210 to Kelley, 8979 for Van Haltren . . .

Only Sam Thompson (6502) had fewer among top candidates.

"Well, no. The difference between the top and bottom of this list, even accepting WARP's numbers, which I doubt intensely with respect to outfield defense, is quite large. Fred Clarke's career value is 1.5 times Mike Tiernan's, for goodness sakes! Jesse Burkett's is 1.1 times Hugh Duffy's."

I actually agree with this. Delahanty Burkett and Hamilton were all near the top of my ballot too.

When I said they were "all the same," I was mentally excluding the already inducted and non-yet-eligible, which excludes 6 of the top 7 names on the list. I put all 15 names together so that people who think Kelley and Ryan and immediately think "Top 2 Outfielders on the Ballot" could change that view to see them, rather, and the 6th and 8th best outfielder of their generation.

My ballot this year will include Kelley (best career) near the middle of my ballot and Thompson (best peak) near the bottom. Beyond those two, I can't see anything that makes any of the rest stand out.

The problem with the plate appearance argument, though, is that there's a bump of 20 games or so starting in 1898 - which happens to be the year that Tiernan declines. So Duffy racked up hundreds more PA by being a 1898-99 regular instead of 1887-88 like Tiernan. They SEEM contemporary, but the PA gap explains a lot of Tiernan's perceived 'inferiority.' He also didn't have part-time seasons, but is that a big loss?

Seems like Tiernan is partly getting screwed by finishing up just as the games-played column is expanding. Should he be? I don't think so. He did have a couple of "missed 20 games" seasons, but he was a lot more durable than Thompson and as durable as most. Duffy also benefited from playing for a 1990s Rockies-like team. Racks up tons of plate appearances, but the OPS+ adjustments take the wind out of his sails.

I really do expect this Tiernan phenomenon to be debunked. But so far, it hasn't happened....

Using baseball-reference implementation of OPS+:

CLARKE, 165 165 157 147 145; 141 140 136 132 131;

And I don't mean that they differ simply in the way WARP implies: Clarke should rank above everyone else on the list, and Flick below everyone on the list except Tiernan.

Clarke's OPS looks a lot like Burkett, only Burkett has a "monster" season while Clarke has several extra very good seasons. Puts them on a similar plane, which seems quite reasonable.

Paul, any opinion on Tiernan's seemingly-spectacular OPS+s? So far he seems to lose out only in PAs. Are we mentally adjusting for a McVey or Start's era, but NOT for Tiernan?

A few tidbits in response to Howie's comments on Mike Tiernan:

1) I think Mike Tiernan was a quite outstanding player for a short time. In my rankings, I place him ahead of Pete Browning, Sam Thompson, and Mike Griffin, all of whom also had short careers and missed significant time with injuries during those short careers. WS supports this placement. Tiernan is therefore behind Stovey, Kelley, Duffy, Pike, Van Haltren, Charlie Jones, Ryan, and Fielder Jones among eligible outfielders. All had considerably more career value than Tiernan, most had better peaks.

2) Tiernan, despite his gaudy OPS+ scores, is not a better player than Hugh Duffy. WARP and WS agree that Tiernan was an indifferent fielder; they agree that Duffy was an outstanding one. A difference of 10-20 points of OPS+ in a season can easily be offset in value by better defense, as both WARP and WS demonstrate. When one adds greater durability (see next tidbit) onto Duffy's side of the ledger, the two players draw farther apart.

3) The argument that Tiernan gets a raw deal in his career value because of schedule is simply false.

a) Tiernan and Duffy are the same age! Tiernan broke in at 20 in 1887, Duffy at 21 in 1888. Duffy's career is longer because he retained his skills longer. Tiernan got off to a faster start: he was a star at 20, Duffy didn't reach star status until he was 23. Tiernan played regularly through 1898, so Duffy only has the one long season on Tiernan, and Tiernan started playing as a regular a year and a half before Duffy. Tiernan's smaller number of games played is due to the steepness and earliness of his decline, and due to the large number of games he misssed while he was a regular.

b) Games missed during seasons as regular

Mike Tiernan commonly missed 10 or 20 games. I don't know how far he thereby stands out among the outfielders in the "glut" under discussion here.

Note. Suppose one knows by heart the number of games scheduled in each 1880s-1900s season. http://world.std.com/~pgw/19c/schedule.games.html

Keeler and Kelley practically missed 1888-1892, when the number of games scheduled was the same as 1899-1903, the first five seasons Tiernan practically missed.

Very well done, Chris. That's exactly what I was looking for.

My only quibble is this:

Andrew, I can see your observation in the abstract, but none of us ever saw these guys play.

Well, for a guy with 500 AB, one extra single will raise his OPS+ by about one point. Can a good outfielder save 10 singles over the course of a season relative to a bad outfielder?

The statisticians here should chime in on this and correct me if need be, but I don't think 20 points of OPS+ is statistically significant in any one season, nor is a 5 point difference over a career.

jimd,

Well, for a guy with 500 AB, one extra single will raise his OPS+ by about one point. Can a good outfielder save 10 singles over the course of a season relative to a bad outfielder?

Yes, but many outs that become hits in the outfield do not become singles.

I don't know about "ability", but in terms of value Duffy simply swept the floor in terms of put outs in the outfield. Tiernan averaged fewer than 1.5 put outs per game, while Duffy averaged more than 2.0. ("Yes, but Tiernan was a right fielder, who got fewer chances." "First, why was Tiernan in right? Because he wasn't as good a fielder. Next, compare their 1890 seasons. Duffy had more put outs in right field than Tiernan did in center. They were even in 1889, when they were both in right, and Duffy had 0.14 put outs more per game in 1888.")

Split the difference, and say that Duffy made 0.25 more outs per game. Over 120 games, that's 30 hits saved. Say that half were singles and half were doubles, Duffy's defense is worth about 30-40 points of OPS.

MattB,

Howie,

Don't know how to track that specifically, by a potential mitigating factor is that Tiernan played for New York, whose star pitcher, Amos Rusie, was always leading the league in strikeouts. Batters who strikeout are, of course, not going to hit flyballs to the outfielders.

Duffy played most of the 1890s with Boston, who's star pitcher was Kid Nichols. Nichols was no slouch either, though, near the lead in strikeouts every year. Jack Stivetts was the #2 pitcher, and no slouch either.

To take a random year (1895), Rusie & Co. struck out 409 (first in the league), while Nichols and friends K'd 370 (3rd in league). That's 39 outs that won't go to Tiernan. This appears to be a consistent advantage for Duffy. But it'd be a mistake to say that Duffy was "helped" by the fewer strikeouts (Boston was always near the top in strikeouts), just that Tiernan was hurt a little more.

Don't know how to get GB/FB stats, so I looked at Nichols' and Rusie's put outs and assists (thinking maybe that the numbers would tell us how many 1-3's and 3-1's they induced -- these would generally be ground balls.)

Rusie had 2.28 (PO+A)/G. Nichols had 2.16 per game.

That puts Rusie and Nichols in the same ballpark, generally. (Rusie, of course, might have just been a better fielder.)

Overall, I think that any difference that may turn up will be swamped by the simple fact that rightfielders are just less important than centerfielders and rightfielders.

Good stuff, Matt. And I think you get the point - it's dangerous for us to ASSUME that Duffy's better 'fielding stats' are purely the result of his wonderful fielding. We "park-adjust" our hitters, but fielding, not as much. That worries me.

I don't know what to think about "fielding ability." I mean, assume Tiernan was a horrible fielder but a great hitter. Today, he'd be a Designated Hitter. In the 1890s, that was not an option. So let us assume, instead, he played rightfield for a team led by right-handed ground-ball, strikeout pitchers. As a result, the ball was almost NEVER hit to right field while Tiernan was standing there. Who cares how "good" he was. Treat him like a DH, I say.

Note that, late in his career (1897 and 1898), Tiernan played more than a handful of games in RIGHT FIELD. Why would a poor defender move from right to left? Well, in those years, for the first time, the team had Cy Seymour and Ed Donehy (lefties both) starting a bunch of games. I don't think it's a stretch to think that Tiernan was moved to left when there was a southpaw on the mound.

Er, those ALL CAPS words should be LEFT FIELD.

> Those I see hurt most by deadball play

Even aside from the large question of, so what? is the question of whether you wrote what you meant to write. You have 2 categories of players who would have been "hurt" by deadball play and no category of player who might have done better?

I agree with Marc's sentiments. I honestly don't care if a player from the 19th century could play the lively ball game. I also don't care if today's hitters would look like fools attempting a fair/foul bunt or a 21st century hurler attempting to throw underhand. All that matters to me is where each player ranked during his time.

Tom's question is an interesting exercise, BTW.

I like the premise a lot, redsox1912. All measurements are worth describing further for weaknesses and strengths.

However, I don't see that chart as making Mantle look BETTER than Mays or Aaron. Look at the number of "over 160, over 170," etc. seasons.

Was Mantle that much better than Mays or Aaron?

If you apply any discount for "quality of opposition", Mantle loses some luster (though far from all) due to the NL superiority during the late 50's/early 60's.

I listed some older players so John Murphy could play along too.

I'm trying to figure out why I was singled out in your post, redsox1912. While I think OPS+ is a good tool, I don't think it's perfect and don't use it when ranking the candidates. Were you referring to something specific I said from a previous post?

BTW, you made some good points in your post. However, MLB should only give asterisks for Bonds' records (if proven he used steroids) when asterisks are presented for any pre-1947 record established when the color line was in place (to be fair).

Howie, how can you possibly conclude:

>However, I don't see that chart as making Mantle look BETTER than Mays or Aaron. Look at the number of

I'm trying to figure out why I was singled out in your post, redsox1912.

If I'm not mistaken, redsox1912 was not commenting on your player rankings. Just on your AGE. :-)

Whoa, Marc.

What's good about OPS+ is that it looks at HITTING. The WARP stuff, from what I can tell, mingles really good, solid hitting numbers with squishier fielding stats.

Sure, but lots of statistical measures look only at hitting, and they do so in a statistically more meaningful way. Linear Weights batting, Runs Created/27, Batting Runs, etc. All of those must be compared to league averages.

And, as Joe has pointed out, OPS isn't statistically relevant across all eras. I believe he advocated giving twice as much weight to "O" in the early years of baseball.

MattB wrote re John Murphy:

First, Howie, I see that you are right about 6-1. But:

>Anything over 180 is so rare that it needs special attention, IMO.

Do I need to be skeptical about this Bonds fella?

Thanks, Jeff. I'm looking for the cracks in the OPS+ armor, actually, and I realize there are some.

lol on Bonds, Marc.

lol on Bonds, Marc.

If I'm not mistaken, redsox1912 was not commenting on your player rankings. Just on your AGE. :-)

Remember Matt, I'm not that much older than you. :-)

I'm working on a long post on outfield defense, but since it's not ready, I thought I'd bring a couple of small tidbits highlighting the importance of context in interpreting statistics.

Sam Thompson's stolen base totals, while not a significant part of his case, are cited from time to time as evidence in support of him, most recently by yest. I observe:

1) In 1891, Thompson stole the most bases of his career: 29. This total tied him for 24th place in the National league

2) In his career, Thompson stole 229 bases. Among the 18 position players on last year's ballot who played enough of their careers after 1885 for them to have a significant stolen base record, plus the 3 top position player candidates this year (Kelley, Collins, F. Jones), Thompson ranks 19th in career stolen bases, ahead only of Deacon McGuire (117) and Jimmy Collins (192).

Chris, one more pin in Jimmy Collins' balloon.

Here's an attempt at carrying forward an account of the value of defense relative to the value of offense. Comments, criticisms, and corrections welcome! I assume that much of what I've written up is repeating what cutting-edge defensive sabermetrics has worked out, but given our historical project, it's worth our talking these matters through for ourselves.

jimd has provided us with a helpful estimate that for a player with 500 PA, 1 additional single in place of an out in the player's record would add 1 point of OPS+ for the player.

Using this as a rule of thumb, I would suggest that we can begin to grasp the relative value of defensive performance by looking at defensive efficiency, moving later from defensive efficiency to consider individual contributions to that efficiency.

Defensive efficiency is basically a defense's success at turning balls in play into outs. It's easy to calculate exactly if one has data on batters facing pitchers. I don't have that data for the nineteenth century (is it available anywhere?), but we can get a good estimate of defensive efficiency by the following formula: 3IP - K / 3IP - K + H - HR .

Let's look at 1895 as an example, since it's been mentioned earlier, focusing first on Boston and NY.

Boston created 3155 outs on 4463 balls in play for a DE of .707. They played in a strong hitters' park, which increased scoring by 7%. Some of that was from an increase in HR, I believe, so, for purposes of a conservative example, I park adjust their DE upward by only 2%, to .721

New York created 3032 outs on 4357 bip for a DE of .696. Their park was neutral, so that figure needs no park-adjustment.

Now a difference in defensive efficiency of 2.5% may not look like much, but when we're dealing with over 4000 bip, the effects are considerable.

If NY had managed a DE as high as Boston's, they would have needed only 4205 bip to create their 3032 outs, so we can say that the greater defensive efficiency of Boston saved 152 hits relative to the NY defense. The question of how to assign credit for hits saved to becomes exceedingly complicated, of course. But let's keep things simple for a moment. First, there's the division between pitchers and fielders. Bill James divides credit for DE evenly between pitchers and fielders. The Voros McCracken argument would assign all credit for DE to the fielders; Tom Tippett has shown that McCracken has overstated the case, but we still can assume that a majority of the credit for defensive efficiency belongs to the fielders. James's 50% division seems quite conservative with respect to the fielders' contribution, so for purposes of illustrative example, let's use it for now.

Of the 152 hits saved, then, credit for 76 belongs to the fielders. Divided evenly that would be about10 hits for each position. Since this is a comparison between a good defensive team and an average defensive team (Boston with a conservative park adjustment ranks 3rd in the league in DE, NY ranks 7th), considerably greater differentials are possible. The top team in the league .729 Cleveland, saved 325 hits over .667 Louisville: 20 hits per defensive position. The variance among the best and worst individual fielders would be, of course, much greater than the variance between the best and the worst teams.

This general analysis shows, I think, that fielding differences worth more than 20 points of OPS+ over the course of a season should be common when there is a 6% variance between the highest and the lowest team DE scores, as there is in 1895. So it's quite possible that one outfielder's superior defense could make up a 20-point difference in their OPS+ scores (assuming equal playing time and that other offensive elements are roughly equal as well). Two questions remain, as I see it. Can those individual differences can be accurately and reliably identified? Do the available fielding metrics appear to do so?

Some time in the next few days I'll take this analysis down to individual outfielders, looking at Hugh Duffy, George Van Haltren, Mike Griffin, and Mike Tiernan in particular.

Chris, great work as always. I wonder, though, about your focus on those four players. Without checking your ballot, maybe their the top four among the "OF glut" on your ballot?

OTOH, the tops in the consensus are (top 15 holdovers) Stovey, Duffy, Thompson, Ryan, Pike, VanHaltren and Browning, plus (newcomer) Kelley. And as one poster, noted, we don't really have an "OF glut," we have a "CF glut." So a focus on Duffy, Ryan, Pike (if data is available), Van Haltren and Browning would seem to be most useful to the most voters. Obviously, F. Jones and Griffin would be more than welcome.

But it seems that spending your time on Tiernan would only really be useful if you could provide his twin, Thompson, as well. And in LF Kelley, Stovey and C. Jones would be of interest.

Hey, sometimes beggars can be choosers!?

Lots o' responses to Marc:

Chris, great work as always. I wonder, though, about your focus on those four players. Without checking your ballot, maybe their the top four among the "OF glut" on your ballot?

In fact, that's not my reason for focusing on these guys at all: neither Tiernan nor Griffin makes my ballot. 1895 came up in an earlier example concerning Tiernan vs. Duffy, and it's a convenient year to work with because Duffy, Van Haltren, and Griffin all play CF that year for virtually all of their teams' games. Van Haltren is interesting because he's Tiernan's teammate, which helps with comparisons between OF positions, and the OPS+ spreads among the three fall into the range whose significance we've been discussing. All I'm attempting is a sort of sample explanation of how fielding value works in hopes that others will either a) undertake the analysis of fielding value themselves more systematically or b) make more use of the available defensive metrics, which, I am persuaded, are more reliable than some believe. And it's good that they are, because fielding analysis is, frankly, tremendously complicated and time-consuming. Since I've done a lot of the foundational work for 1895, I could expand my focus to look at the other notable outfielders who are active that year -- Kelley, Ryan, Thompson, and the forgotten (and undervalued) Elmer Smith.

? Chris, one more pin in Jimmy Collins' balloon.

It's all about context, Marc. Context, context, context :-) Stolen base totals drop substantially with the new century, probably due more to changes in accounting practices than to changes in play. Thus, most of Collins' career takes place during an era of much lower overall stolen base rates. For instance, in 1903 Collins stole his highest number of bases: 23. With this number he placed 17th in the league (as opposed to Thompson's 24th with his high of 29). Moreover, Collins's total was 50% of the league best of 45, while Thompson barely managed a quarter of the 111 stolen bases accumulated by his esteemed teammate, Sliding Billy Hamilton. So while Collins's stolen bases are hardly the best line on his resume, his numbers need to be compared to those of other players who played the majority of their career 1900 - 1910 before we start sticking pins in anything.

Having the highest ranking for Ed Williamson of ballots so far posted (I think I can say that without engaging in damaging mid-ballot analysis), I can speak up for Sliding Jimmy with a clear conscience :-) .

for a player with 500 PA, 1 additional single in place of an out in the player's record would add 1 point of OPS+ for the player

Everyone seems to agree, so it must be too late for me to think.

the large study done by D Adams showed the AL being significantly weaker than the NL in 1901 and not catching up to about even until the later years of the decade.

Do you have publication data for the Dallas Adams study?

Chris Cobb (#205) on 1895 team defense:

It's been said: as one poster, noted, we don't really have an "OF glut," we have a "CF glut." So a focus on Duffy, Ryan, Pike (if data is available), Van Haltren and Browning

I have a small problem with this. Of these five, only Van Haltren was a CF for the bulk of his career.

VanHaltren played 69% of his career games in CF, his teams's regular in 11 seasons.

Paul Wendt wrote:

<i>? Chris Cobb (#205) on 1895 team defense:

My understanding is that in modern times, the range of values at any defensive position is about the same. Ie, a good defensive 1B saves as many runs/hits/whatever over a bad defensive 1B as a good defensive SS saves over a bad defensive SS. Which makes sense. At any position (except DH) you are willing to give up some defensive to improve on the bat. Once the fielding gets too bad, the player must either shift on the defensive spectrum putting him back in the acceptable range, or DH, or get benched.

I dont know if this applies to earlier eras or not though.

Share of fielding games played in CF, with notes on other positions.

More information about early, longtime regular CFs.

Thanks, Paul!

RobC wrote:

Yeah,

Patrick, thank you for reminding us of Thompson's standing in WARP3 and explaining how it is the starting point for your rankings. I had, frankly, forgotten about Thompson's standing there, as WARP3 is the last of the comprehensive metrics I look at, rather than the first, but it is a comprehensive metric, and its approach to combining offensive and defensive values needs to be taken seriously.

The coincidence of WS and WARP3 on Thompson's value is interesting, esp. because, if both systems are measuring value corrrectly, they shouldn't coincide! My hypothesis would be that Thompson comes out better in the WARP1 to WARP3 calculation than do his peers because more of his value lies in his hitting for power than in any other aspect of his game. WARP3 seems to see power hitters as diamonds -- they are forever -- while it sees players whose value is much more linked to other aspects of the game as silver -- they tarnish with time.

I'm not at all convinced that the WARP3 adjustments are correct, but given that we're having to make more comparisons between players from different eras (and that's a problem that will only worsen with time!), it's worth thinking through more fully what WARP3 does, and why.

Chris,

You might be right about the statistical outliers being in the tougher defensive positions, at least in the AL. But, in the NL, the statistical outliers in the other direction (really bad fielders) still get to play if they can hit enough. So the NL may still have the same range of values.

I looked at Put Outs per Game in the NL in 1895, divided by outfield position. 1895 seemed like a good year to pick, since just about all of our outfield glut members are represented.

In order to be included a player had to (a) play at least 95 games that year at a single outfield position, and (b) had to play at least 85 percent of his games in a single outfield position. This left 11 leftfielders, 9 centerfielders, and 7 rightfielders.

average -- best PO/G -- worst PO/G

Leftfielders

I tried 1896 as well, and got nearly identical numbers and ranges: this looks at 9LF, 9CF, and 8RF from that year who met the criteria above:

average -- best PO/G -- worst PO/G

Leftfielders

slightly off the beaten path..

extension to #215, including the five (marked *) other 9-year regular CFs who played in the Deadball Era and earlier. This is that subset of the 30 (up from 25) who did not play after 1899.

86%; 9if ; LangeB (ne)