Evolving MLB Defense: 1901-1929, at 2B

[leewileyfan]

Some posters have encouraged me to dust off old notes and my defense metric to post some player evaluations. I decided to start at 1901 because that's where my manuscript begins and lead of with 2B because a number of recent threads here are debating players at that position across several generations of play.

By brief introduction: the metric is position-challenge and opportunity based. I started out, taking months to map out various plays and strategies that prevailed in 1901; the equipment used; the ball quality and in-game duration; ball liveliness and batting strategies, and base running.

I started with the premise that each position has unique turf, challenges, error costs and play opportunities. An error at each position will have costs in runs and runner advancement unique to the position.

To evaluate players from 1901-2011 against peers and against players long deceased, the metric inputs had to be interactive with me as years passed and the game changed. When rating started to go beyond expectation, the model demanded adjustment. The model has been consisent overall, but has been flexible and responsive to changes in position dynamics.

Evaluations are measured to a TEMPLATE of perfection, reached by no player. The rating can then be converted to something uniform and familiar, which is why the RATINGS resemble fielding percentage. When something changes to improve defense overall, ratings rise and the template is adjusted. This allows for comparisons across generations; and as a kind of Devil's Advocate test, I can shift players out of their generations to see how well they hold up in the evolving game.

At 2B, starting in 1901, these players best spell out Phase I of the early 1900s, 1901 - 1911, when the first adjustment was made to ball liveliness. These players are best measured in the basic 1901-1912 Template and here are their RATINGS [average at this time is about .931]: their seasons are in brackets;

Nap Lajoie, .954 [1901-1912]
Hobe Ferris, .939; [1901-1909]
Johnny Evers, .931; [1903-1915]
Muller Huggins, .940; [1904-1915]
Eddie Collins, .940; [1908-1926]
Del Pratt, .945; [1912-1924]

EDIT: Had meant to include, in the first group, Bill Wambsganss of CLE [1914-1925], whose career was predominant;y before the 1921 and after live ball era. His rating = .947 against .931 average, inserted among the players above.

At .954 vs. average of .931, Lajoie [career] averaged +4.51 DR > average over the 13 seasons, or +58.6 DR.

The second group have career spans that are predominantly after the slight jolt to the ball and which enter a decent part of the live ball era during and after 1921. One significant change in the game was the increase in the DP's executed and the importance of that execution to the game. Also, between 1909 and 1922 [and ongoing] there were design upgrades in glove pockets and webbings. Fielding got better and the templates rose.

EDIT ADDED: Moreover, with the fatal beaning of Ray Chapman in 1920, the 1921 and subsequent seasons, in addition to the live ball, have the prohibition of formerly legal pitchers, with some veteran spitballers godfathered in until their career ended.

Freddie Maguire [1928-1931]; .974
Rogers Hornsby [1916-1931]; .953
Hughie Critz [1925-1934]; .977
Frankie Frisch [1920-1935]; .966
Bucky Harris [1920-1927]; .946

The average MLB rating between 1921 and 1929 varied between .940 and .961, settling around .950. Over a 154 game schedule, Hughie Critz would be around +9.4 DR above average, or +94 DR > average over the decade.

Two players with extended careers dip back into the 'teens and extend through the 1920's. They have been measured separately in groups according to their career years. Rogers Hornsby around +1.0 defense runs; +16 DR, career.

Hornsby, transposed to the 1912 template would be at .970. Eddie Collins, transported to the 1929 template would be at .921. Nap Lajoie, uprooted from 1901 to 1929 would rate .937 with all the game changes. I'm convinced he was a towering defensive force as the game moved into the 20th Century.

Some players have become established favorites of a number of defense metrics. Others seemingly are wrapped in shrouds they can't shake off. If this approach causes a few surprises, then I say OK - let it ride.

This, I hope is an opener. Comments and debates and player suggestions are all welcome.

[leewileyfan]

To clarify the evaluation metric, the following are the inputs used [with reasons why] and weightings for those inputs [and why].

1. Since my primary two interests are to create a credible defense metric AND to have that metric useful in comparing players from different generations of play, I use ONLY those data that are readily available [and pretty well vetted] for the whole period [1901 to Present] for each position. Those are Put Outs, Assists, Double Plays, Errors, Passed Balls, Wild Pitches.

2. Put Outs [PO] are considered routine for infielders and catchers. In the outfield, the PO is the bread and butter of range, and mistakes at these distances from home plate and the bases are costly.

3. Assists are the bread and butter for C, 1B, 2B, 3B, and SS and I assign degrees of difficulty for each position. Likewise, errors are weighted by potential run damage by position.

4. Inputs for PO, A, DP, E, PB, WP are added, with the NET producing a raw number that is converted into a rating. The RATING intentionally resembles fielding percentage, for ease of communication.

Since this starts out at 2B, here are the VALUES for various events, as presented for 1901:

PO = .10
A = 1.167
DP = 1.20
E = 1.347
Template = 5.150 [Template is model of excellence, which arithmetically converts to 1.000]. At the start, NO player at 2B reaches this net sum in raw data. The TEMPLATE becomes interactive as the years pass, with ratings rising due to some change in the game, rules, glove technology, bat-ball design, etc.

The dilemma, for me, is the basis for comparison. I referred to AVERAGE in the opening post to this thread; but I am much more comfortable with PLAYER X, as a subjective context that is suggested by the data input process. PLAYER X serves two evaluative-comp purposes:

1. It frees up the metric from getting bogged down in seasonal-positional averages, which can vary between leagues and from season to season. When you input data, both tedious and informative at the same time, you get a picture of a BASE of performance, nowhere near the stars but only moderately below the bulk of what intuitively appears to be "average."

2. Setting up the comparison basis in PLAYER X, we get the luxury of a model player who plays the entire schedule at his position. Whether it's GAMES or INNINGS, PLAYER X is always there. He is better than replacement level. In a .260 batting league, he's the guy who hits .245 and in the field [at 2B], when the stars are rated at .970, and "average" seems to vary between .940 and .945, PLAYER X will be at .925, about -7 or 8 DR in a season [-16 DR compared to the "star."].

In a debate over 2B player defense here [started with Carew-Morgan and landed on Collins-Hornsby] there were several comments coming down on either side. Since this thread starts at 1901 and runs [for openers] through 1929, a Collins-Hornsby defensive comparison seems a suitable exercise for showing the metric.

From 1901 through 1920, the 2B Template is 5.150; and the Player X rating is .915. In 1921 the Template adjusts to 5.7; and Player X rates at .925.

Eddie Collins is covered at 2B from 1908 [a part-time, break in year] through 1914 as a member of the A's; from 1915 through 1926 as a member of the White Sox. His 1925 and 1926 seasons are years of declining playing time, but at or above 100 games. Collins' ratings are split between the early Template and Player X models [1908-1920]; and the live ball Templates and Player X model [1921-1926].

Aside from 1908, Collins play up until 1925 was full-time, with an abbreviated 1918 season [WW I] in which he was enrolled as a corporal in the USMC in what appears to have been reserve status, with little or no loss of playing time.

Here are Collins' career ratings, by season [Player X = .915 [1901-1920]; .925 [1921-1926]

1908 .898 [part-time play; break in season]
1909 .920
1910 .943
1911 .920
1912 .926
1913 .939
1914 .948
1915 .948
1916 .931
1917 .915
1918 .950
1919 .942
1920 .960
1921 .960
1922 .931
1923 .940
1924 .939
1925 .939 [regular with reduced playing time]
1926 .937 [same as above]
Collins also played in 1927, returning to the A's, with playing time about equal to 1908. Not calculated in this profile.

Hornsby's career spans from 1915; but he was not a 2B until 1920. From 1915 through 1919 he was essentially a SS or 3B, his move to 2B occurring in 1920. He is rated under the early model for 1920 only and the live ball model from 1921 through 1929. Like Collins, he had an abbreviated 1918 season. He was drafted for WW I, but under a deferment, was assigned to work in the navy yards. It appears that no significant playing time was lost. His 1923 seasons was abbreviated by injuries. He played 2B with three teams, identified after ratings below.

His career defense ratings:

1920 .974 STL [1920-1926]
1921 .944
1922 .936
1923 .941 [reduced play; knee injury ... then re-injured]
1924 .978
1925 .941
1926 .941
1927 .983 [NYG]
1928 .951 [BOS]
1929 .972 [CHI]

[leewileyfan]

Post 3 was edited and erroneously posted as a DUPLICATION of Post 2.

Post #2 is edited, complete and correct.

[bluesky5]

Would averaging out these numbers give a better idea who the better second baseman rather than yearly averages which tend to rise and fall due to a variety of factors outside the players natural ability?

[leewileyfan]

bluesky5:

I have personally run career totals on players, which I believe does what you suggest. As an example, Hornsby's career rating, which specifically totals all the inputs across his career, comes to .953. Then, relating that to a Player X, over the years he played, would provide a more stable basis for comparison than "average" which can be volatile.

In the Hornsby career example, the Player X model is .915 ONLY for the 1920 season but is hiked to .925 for 1921 through 1929. Exactly weighted, Hornsby would be .953 as compared to the Player X composite at .924. This coverts to +10.06 DR over the 10 seasons = +106 DR better than Player X.

In the book manuscript, I go to great lengths clarifying that Player X is a solid cut above "Replacement Player." Player X, in my view, may be last year's veteran "regular" at 2B, now having lost a step or two and a bit of former range and quick release. He may be the guy you trade for or the guy you get on waivers. Over a full schedule, he'll cost you 6 or 7 defense runs a season.

With that clarification, Hornsby is solid on defense, maybe with up-and-down seasons, but a bit above average. Average may vary between .940 and .955 seasonally, being about .949-.951 overall, with Horsnby at .953.

Conversely, the Eddie Collins model is weighted 13 seasons at Player X [.915] and 6 seasons at .925. His career rating [.921] is marginally above the Player X composite [.918]; so I have him very nearly average to slightly below average [career]. Average is likely something under .930 overall.

Thanks for the question. I hope the response was in line with you suggestion. If not, just let me know.

[BigRon]

To clarify the evaluation metric, the following are the inputs used [with reasons why] and weightings for those inputs [and why].

1. Since my primary two interests are to create a credible defense metric AND to have that metric useful in comparing players from different generations of play, I use ONLY those data that are readily available [and pretty well vetted] for the whole period [1901 to Present] for each position. Those are Put Outs, Assists, Double Plays, Errors, Passed Balls, Wild Pitches.

2. Put Outs [PO] are considered routine for infielders and catchers. In the outfield, the PO is the bread and butter of range, and mistakes at these distances from home plate and the bases are costly.

3. Assists are the bread and butter for C, 1B, 2B, 3B, and SS and I assign degrees of difficulty for each position. Likewise, errors are weighted by potential run damage by position.

4. Inputs for PO, A, DP, E, PB, WP are added, with the NET producing a raw number that is converted into a rating. The RATING intentionally resembles fielding percentage, for ease of communication.

Since this starts out at 2B, here are the VALUES for various events, as presented for 1901:

PO = .10
A = 1.167
DP = 1.20
E = 1.347
Template = 5.150 [Template is model of excellence, which arithmetically converts to 1.000]. At the start, NO player at 2B reaches this net sum in raw data. The TEMPLATE becomes interactive as the years pass, with ratings rising due to some change in the game, rules, glove technology, bat-ball design, etc.

The dilemma, for me, is the basis for comparison. I referred to AVERAGE in the opening post to this thread; but I am much more comfortable with PLAYER X, as a subjective context that is suggested by the data input process. PLAYER X serves two evaluative-comp purposes:

1. It frees up the metric from getting bogged down in seasonal-positional averages, which can vary between leagues and from season to season. When you input data, both tedious and informative at the same time, you get a picture of a BASE of performance, nowhere near the stars but only moderately below the bulk of what intuitively appears to be "average."

2. Setting up the comparison basis in PLAYER X, we get the luxury of a model player who plays the entire schedule at his position. Whether it's GAMES or INNINGS, PLAYER X is always there. He is better than replacement level. In a .260 batting league, he's the guy who hits .245 and in the field [at 2B], when the stars are rated at .970, and "average" seems to vary between .940 and .945, PLAYER X will be at .925, about -7 or 8 DR in a season [-16 DR compared to the "star."].

In a debate over 2B player defense here [started with Carew-Morgan and landed on Collins-Hornsby] there were several comments coming down on either side. Since this thread starts at 1901 and runs [for openers] through 1929, a Collins-Hornsby defensive comparison seems a suitable exercise for showing the metric.

From 1901 through 1920, the 2B Template is 5.150; and the Player X rating is .915. In 1921 the Template adjusts to 5.7; and Player X rates at .925.

Eddie Collins is covered at 2B from 1908 [a part-time, break in year] through 1914 as a member of the A's; from 1915 through 1926 as a member of the White Sox. His 1925 and 1926 seasons are years of declining playing time, but at or above 100 games. Collins' ratings are split between the early Template and Player X models [1908-1920]; and the live ball Templates and Player X model [1921-1926].

Aside from 1908, Collins play up until 1925 was full-time, with an abbreviated 1918 season [WW I] in which he was enrolled as a corporal in the USMC in what appears to have been reserve status, with little or no loss of playing time.

Here are Collins' career ratings, by season [Player X = .915 [1901-1920]; .925 [1921-1926]

1908 .898
1909 .920
1910 .943
1911 .920
1912 .926
1913 .939
1914 .948
1915 .948
1916 .931
1917 .915
1918 .950
1919 .942
1920 .960
1921 .960
1922 .931
1923 .940
1924 .939
1925 .939
1926 .937

Hornsby's career spans from 1915; but he was not a 2B until 1920. From 1915 through 1919 he was essentially a SS or 3B, his move to 2B occurring in 1920. He is rated under the early model for 1920 only and the live ball model from 1921 through 1929. Like Collins, he had an abbreviated 1918 season. He was drafted for WW I, but under a deferment, was assigned to work in the navy yards. It appears that no significant playing time was lost. His 1923 seasons was abbreviated by injuries. He played 2B with three teams, identified after ratings below.

His career defense ratings:

1920 .974 STL [1920-1926]
1921 .944
1922 .936
1923 .941
1924 .978
1925 .941
1926 .941
1927 .983 [NYG]
1928 .951 [BOS]
1929 .972 [CHI]

Lee, you're probably getting tired of hearing this from me, but my reluctance to accept your findings continues. Nowhere in your description/template do I see any reference to groundball/flyball tendencies of pitching staffs; predominant righthandedness/lefthandedness of staffs; strikeout propensity of staffs; walk propensity of staffs. All of these, and more, can and do significantly impact the inputs you use for your calculations. Now, it is hard work to get this info- even harder in the old days. Nonetheless, if you wish to advance the state of defensive evaluation I believe it is necessary to deal with these issues.

Your comments?

[leewileyfan]

Lee, you're probably getting tired of hearing this from me, but my reluctance to accept your findings continues. Nowhere in your description/template do I see any reference to groundball/flyball tendencies of pitching staffs; predominant righthandedness/lefthandedness of staffs; strikeout propensity of staffs; walk propensity of staffs. All of these, and more, can and do significantly impact the inputs you use for your calculations. Now, it is hard work to get this info- even harder in the old days. Nonetheless, if you wish to advance the state of defensive evaluation I believe it is necessary to deal with these issues.

It only becomes tiresome when the tone of the questions or challenges assumes the attitude that I am wrong in my basic premise. That creates an atmosphere where one can become like a firefly caught in a kid's jar with the perforated cap for breathing, lighting up a summer evening. If that were the case, I'd just say that I'm no firefly; and summer hasn't come yet - not even here in FL.

If it's open questions and answers, constructive and sharing, I have no problem at all with questions or challenges. They are fun. They also give me feedback so that maybe I will try to get this book published. Heck, I've put enough hours into it. In any event, the whole thread is intended to share, perhaps to learn a thing or two, clarify a thing or two ... enjoy defense and its study.

:Your comments?

As for your reluctance to accept my findings ... well, that is a not a problem for me at all. IF, after I have exhausted all reasonable efforts to clarify my points, you still disagree and find the whole enterprise wanting some depth you demand, then I say thanks for the interest, even if passing and dismissive.

I simply disagree with your recurring premises although I can appreciate where you're coming from. Perhaps it is my perspective on the entire project:

1. My MLB defense clock starts with the 1901 season. I am interested in player backgrounds, their clubs, their ballparks, their opponents, the ball, the gloves, the bats, the park dimensions, the flannel, the hot sun, the sun fields ... but most of all I am dealing with DEFENSE and what these players did at each position, given the opportunities and challenges at each position [and each position IS UNIQUE.

2. I see "handedness" as an integral part of the game, a very interesting phenomenon that can affecr strategies and positional options. I can recall the old saw that NO catcher an be a lefty thrower and that a 1B SHOULD preferably be a lefty. Year before PLATOONING came into play, we KNEW that left-right battles between pitchers and hitters somehow lessened the pitcher dominance over the batter.

3. I have read, studied in fact, several books, essays, white papers of all sorts on the metric measuring all aspects of the game, especially those dealing with defense. I have delved into the work of Michael Humphreys, The Hidden Game of Baseball, PECOTA, MGL, Tango, Dewan and I find much range and disagreement on findings. All I have done is decide to find [MAYBE] a better, simpler approach that does NOT attempt to address all the variables that you consider so essential. [I want to SEE the forest without either bumping into or tripping over trees and shrubs].

4. Over 111 years I honestly believe that handedness, ground ball/fly ball, park and other factors are far LESS important than what is there [and has been there all alon] in a good box score: PO, A, DP, E, PB, WP]. That is simply my opinion; but I believe my metric proves the point.

5. Defensively speaking, PARK dimensions will inhibit C, 1B, 2B, 3B, SS to the degree that pop fouls are "playable," in that "outs" may be lost as a consequence. At those positions, PO are more routine than ASSISTS, which constitute their bread and butter. NOTHING LOST in the infield. In the OF, PARK may well be a factor being the difference between a web gem and a HR. Even with that, I believe the metric holds up for LF, CF, RF [although I am open to innovative ideas that weave PARK into OFers and the HR factor.

6. Over 111 seasons, flamethrowers, spitters, knucklers, palm balls and eephus pitches have eerily produced rates of fly-ball/ground-ball that conform to a big picture predictability. NO, I do not lose sleep of that factor, either.

7. Walk propensities of pitching staffs may show a trend; but, to my knowledge, they are not contagious. It's part of the game and may or may no lead to DP's of HR's with extra runners on. My metric is not daunted by this either.

8. Pitching staff K's: AHA!!! Agreement. I was going to get around to that. My ratings are presented without adjustment for that factor. However, I also have a formula for EQUIVALENCE that can adjust +/-DR's as stated by the raw formula output, by position, with very specific adjustments. Therefore, that is addressed in the final metric and should not be an issue.

We disagree on what is ESSENTIAL to developing and advanced and credible metric. I WANT my metric to pare data, as much as possible, to the bone. The K adjustment I fully understand and appreciate. The rest ... not so much at all.

I have tried to explain my perspective, focus and intention. If you are reluctant and dismissive at this point, then thanks for the comments. It is my intention to share the metric here, perhaps defend it to those who may be interested in the possibilities [or who see opportunities to improve its basics]. I am not participating in the thread to sell anyone ... or to justify my thought process.

Bottom Line: I have been [and continue to] explore defense from the perspective of POSITION uniqueness, in opportunity and execution; the drama once the batted ball sets defenders into action; and the proper weighting of challenges with successful defending and the penalties for failure. This, to me, is the essence of defense. The metric may seem complex in its diversion in approaching each position [almost] like a fiefdom; but once in place, it never really goes beyond arithmetical calculation. I hope it generates some interest, questions, and discussion ... into all positions, player comps and generations of play.

[leewileyfan]

BigRon has raised the question of pitching staff strikeouts being a factor in determining defense ratings. I agree, having been challenged on this point when I was initially wrapping up final touches to the metric about five or six years ago.

In a league in which the average pitching staff record 1100 strikeouts, there will be pitching staffs above and below the average; and these variations impact batted balls put in play for the defense to field. The initial challenge offered to me was EQUIVALENCE, the point being that, without adjusting for those defense OPPORTUNITIES, the metric would have a bias.

To address this, I recalculated all positions over several seasons, modifying individual inputs [PO, A, DP, E, PB] to absorb the +/- DR at each position, player by player. The value of a "K" calculated to +/- .1065 DR, which then had to be equitably distributed over the eight positions. [I omitted pitchers in my study. I have no idea why I deliberately did this; but I was beginning with a position-player orientation]. In the final editing process, I may just add pitchers into the mix. I have left 1% of the distribution of "K" adjustments to do just that.

For the working model I employed these percentages in sharing the distribution of team +/- K rates compared to league:

C .04
1B .09
2B .14
3B .11
SS .17
LF .135
CF .17
RF .135

Since we are talking about 2B, we can return to the Collins-Hornsby comps as an example. If we use 1921, a season in which both played as contemporaries at 2B, we Hornsby [rating = .944] and Collins [rating = .960], having ONLY the season and the position played in common. They played in different leagues [Cardinals; White Sox]; so the metric's starting point is the basic defense metric rating.

In 1921, the average pitching staff K's recorded in the A.L. was 451. White Sox pitchers K'd 392 opponents; so there were 59 MORE batted balls in play than average. Adjustment for EQUIVALENCE, operating on smoothing out opportunity, calls for recording the difference: TEAM K - LG K, 392-451 = -59. Expressing the relation between team and league in this fashion insures that +/- is properly applied. [Position players behind low "K" staffs have MORE opportunity in the basic ratings than players behind high "K" staffs; thus the +/- results].

In 1921 Hornsby's Cardinal staff racked up 464 K's against a League average of 420. Thus, the Cardinal defense had 464-420 = +44 K's, meaning 44 FEWER defense opportunities in the field. These players must share +44 among them.

Collins
Basic Rating .960
K's -59
K Value = .1065
2B share .14
Adjustment, 2B = -59*.14*.1065 = -.88 DR
Basic Rating > Player X .960 > .925 = +12.15 DR
Adjusted for K's = 12.15 DR - .88 DR = +11.27 DR [This is for in-League adjustment].

Hornsby
Basic Rating .944
K's +44
K Value .1065
2B share .14
Adjustment, 2B = +44*.14*.1065 = +.66 DR
Basic Rating > Player X = .944 > .925 = +6.59 DR
Adjusted for K's = +6.59 DR + .66 DR = +7.25 DR [This is for in-League adjustment].

If we want to compare Collins and Hornsby one-on-one, adjusting for K's, we COULD just use actual K's by either staff, OR go through calculations by League, games, innings played, etc. I like to reduce things to LEAST [but reasonable] complexity.

The White Sox staff had 392 K's and the Cards had 464. That's 392-464 = -72 K's for Collins and ZERO adjustment for Hornsby, at the basic rating level.

Collins .960 ... -72*.14*.1065 *-72 = -1.07 DR
Original @ .960 = +12.15 DR -1.07 = +11.08 DR

... compared to Hornsby's basic rating [.944] which gave him +6.59 DR > Player X. Collins, 1921 = [11.08-6.59] 4.49 DR > Hornsby, 1921. [Playing Time adjustment follows].

Playing time for Collins [% of team at 2B] = .877 * 11.08 = +9.72 net DR. For Hornsby [.928*6.59] = +6.12 DR.

[leewileyfan]

Would averaging out these numbers give a better idea who the better second baseman rather than yearly averages which tend to rise and fall due to a variety of factors outside the players natural ability?

Bluesky5 [and anyone else who's interested]: Perhaps the best way to respond well to this question is to calculate, step-by-step, how a player rating is developed and how it is converted into a rating and then into defense runs. Since I kicked this off with early play [1901-1929], with an initial focus on 2B, I'll use Nap Lajoie's
defensive play at 2B in 1907 as the example.

Games: I have revisited expressing this a number of times, always looking for innings played at a position in a given year as far back as possible. With corrections and deletions and edits I've seen over the years, I am using "GAMES" [Baseball-Reference] from 1901 until GAMES STARTED is uniformly available [around 1920]. I use GAMES STARTED from that point through 1953, after which INNINGS PLAYED is uniformly available. That way, I do not belabor myself over partially available seasons in which "finer" definition is available, for multiple reasons:

-It presumes the data is absolutely correct, even if not available for all players in a season ... I will sacrifice a slice of "infallibility" for uniformity of data;

-If, for example, such data is available at a premium, I am unwilling to explore that option, since I cannot assume the average interested fan has the same access, or that a beginner has the access, or, quite possibly seeks a simpler means of entering study of MLB defense.

-I believe there are enough defense metrics in which metric designers, even with the most infinite data at their fingertips, elect to toss out clusters of plays which they deem irrelevant or creative of statistical "noise."

Back to Nap Lajoie, 1907:

Games = 128

Put-outs, in this metric [for 1B, 2B, 3B, SS] are treated as the most routine of chances in the field [pop flies, force plays, run-down tags, double-play force outs] and are given a .10 weighting, per event:

Lajoie PO = 314 * .10 = 31.4, which is then treated as relateive to GAMES [31.4/128 = .245

Assists, in this metric, are the infielders' bread-and-butter and more challenging plays to execute, weighted at 1.167 per event, relative to games:

Lajoie A = 461 * 1.167 = 537.99/128 = 4.203

Double-plays are seen as a valuable skill set, justified by the return on effort: 2 outs with one batter. The 2B share for the DP is set at 1.2 per event, again related to GAMES:

Lajoie DP = 86 * 1.2 = 103.2/128 = .806

We add these POSITIVE inputs for Lajoie [.245 + 4.203 + .806 = 5.254

Now, the offsetting element on the negative side: Errors.

Lajoie committed 25 errors. Errors at 2B are valued at -1.347, so 25 * (-1.347 * 25)/128 = -.263

The net input composite for Lajoie = 5.254-.263 = 4.9915 [rounding], a number which has no familiarity or relevance to anyone, which brings me to the TEMPLATE.

Each position has a TEMPLATE, a standard of excellence for a position that NO player at the position has achieved to date. [It's the Golden Apple, just out of reach]. Before I finalized a single initial season [1901], I devoted time to the inputs and weightings, but especially to setting up a TEMPLATE. EACH position has its own unique template and weightings, because each has its own unique demands, opportunities and challenges. The TEMPLATE also provides the basis for converting raw inputs into a rating format that has commonality among all divergent positions and has a familiar "look."

TEMPLATE and CONVERSION: The TEMPLATE at 2B in 1907 is 5.15. Lajoie's net sum is 4.9915. To make the conversion to a RATING, the formula [for 2B]:

((1) - (Position Template - Player Composite) * (.05))

OR,

((1) - (5.15 - 4.9915) * (.05)) = .992, Lajoies defense rating for 1907, which challenges the TEMPLATE. PLAYER X, that slightly below average 2B who is about 8 or so runs below the average full-time regular is .915. The differential value at 2B is .347, so, removing decimals:

992-915 = 77 * .347 = +26.7 defense runs above Player X or about 18-20 defense runs better than major league average regular in 1907.

[leewileyfan]

To wrap up the full metric presentation for Nap Lajoie [1907], there are two final adjustments:

1. Lajoie played 128 Games of the CLE schedule of 152 Games. Therefore, Lajoie's +26.7 DR > Player X becomes: 128/152 = .842 * 26.7 DR = +22.48 DR.

2. For those seeking complete EQUIVALENCE, the American League average pitching staff recorded 560 K's. The CLE staff recorded 513 K's. Therefore, 513 - 560 = -47 K's, to be distributed among all fielders. The 2B "share" is .14: .14 * -47 = =6.58 K's @ .1065 Run Value per K = -.70.

Thus, compared to League K average, Lajoie would be +22.48 - .70 = +21.78 DR > Player X. [About +12 to +14 > average full-time regular at 2B, 2007].

[leewileyfan]

If I had my druthers, I'd probably change the title of this thread to EVOLVING HISTORY OF MLB DEFENSE [1901-PRESENT]. The position specific ratings cover 112 seasons; and I am now finished basic edits through 1969, with 1970 on mostly being a matter of checking accuracy of INNINGS PLAYED at a position, each layer, each season.

The whole editing process recalls, for me, reminders of all the awards handed out today for very well paid players. Awards were not so conspicuous in 1901; so I figured I'd take a shot at posting Diamond Defensive Dandies [if such were ever conceived for 1901].

Catchers: Bill Bergen [CIN]; Billy Maloney [MLW]

1B: Fred Tenney [BOS]; Frank Isbell [CSX]

2B: Don Padden [STL] or Cupid Childs [CHI - less playing time]; Nap Lajoie [PHA]

3B: Tommy Leach [PGH]; Jimmy Collins [BSX - holdover from the 1890s Beaneaters, whose former teammate, Bobby Lowe, is playing across town at 2B]

SS: Bobby Wallace [STL]; Billy Clingman [WAS]

LF: Jimmy Sheckard [BRO]; Jimmy Barrett [DET]

CF: Billy Hamilton [BOS - wrapping up a long pre 1900 career]; Ollie Pickering [CSX]

RF: Elmer Flick [PHI]; Cy Seymour [BAL]

[leewileyfan]

Looking for feedback, I have been wrestling with the Player X concept [as a basis for defensive comparisons, by position, with SOMETHING]. I have very nearly convinced myself that Player X [from the perspective of what I want [and want to avoid] is perfect:

1. AVERAGE is ever changing, season-to-season, by degrees which are normally within a reasonable range. However, for whatever reason, there may me an outlier season at some position or another that makes anything looking to be "absolute" very questionable.

2. REPLACEMENT is a player concept I have never been able to wrap my head around, NOT because I perceive flaws in the concept itself, BUT RATHER that is is not consistent with HOW I wish to present defensive comparisons, with each position getting its fair and equitable comparisons.

3. Player X is several cuts above replacement level, a level which has very definite payroll/investment roots. The essence, as I understand it, is placing a statistical level of performance expectation that is equitable with any of several and divergent models: minimum wage player; minor league call-up, worst-or-next-to-worst MLB player at a position, etc. The concept generally converts to a W-L performance around .290-.310, or about a 47-115 or 50-112 record in a MLB season.

4. Player X is the guy you hope to get on waivers, or in a swap for "a player to be named later." He's either an aging pro hanging on [and with some skills, even if a bit eroded] or a comeback kid, hoping for his second wind. He may also be a raw kid with some flashes of promise, one the parent club doesn't have the time or inclination to "groom." A team of Player X will probably cost you 8-12 runs at his position, such that a team so comprised will be at 70-92, maybe as low as 65-97, never much below that. He is a phantom, yet he is real.

Just tonight, editing innings played in 1950, I came across an old sentimental favorite from my childhood, Jimmy Bloodworth, utility 2B with the Philadelphia "Whiz Kids." What drove me to follow his career so closely, I may never completely figure out. I saw him, with the Senators and the Tigers before the War; and he made some exciting plays ... I believe he cranked a homer to beat the Yankees ... a shot into the LF bullpen. Whatever. With Ted Williams as main idol and Cecil Travis close behind, Jimmy Bloodworth became my "Everyman."

Just for some background: Bloodworth served 2+ seasons in WW II, then returned, being shuffled around and ending up as a Dodger property around the time Jackie Robinson, another 2B, was breaking the color barrier. Jimmy won the International League MVP and went to the Reds, where he played full-time and won Comeback Player of the Year in 1949. Then the trade to the Phillies.

So, I guess, Jimmie Bloodworth is one model for Player X. He's in his 30's, been shufled around [and up and down], lost almost 3 years to service, and he's a utility man.

Player X at second base in 1950 is expected to be .925 in the field. Let's say League average is around .947 this year. Player X, full-time, will be [.925] about -8 defense runs.

Here's Jimmy Bloodworth's 2B line for 1950, as a sub:

Innings = 157 = 17.44 Games, with 53 PO, 55 A, 13 DP and 0 errors. His defense rating = .972, +14.55 DR above Player X expectations. His playing time is only 11.32% of the team's total schedule; so his contribution is + 1.65 DR seen from a full-time perspective.

The guy Jimmy subbed for, Mike Goliat had:

1221.1 innings = 135.7 games, with 345 PO, 393, A, 89 DP and 21 E. His rating = .937, or + 4.16 DR above Player X. With 88.1% playing time, Goliat's total net defensive value is +3.66 DR.

Player X is a CONCEPT; and many players who fill the role will outperform. Conversely, when I tabulated some regulars or noted position players, they surprise defensively coming in at or below Player X levels. Most will follow a pattern: growing into peak performance levels, then gradually declining.

I've posted previously that many published metrics I've seen seem to make indelible reputations of players, either good or bad, where I do find a fair share of contradictory surprises.

Player X provides a fairly stable "image" of expectation at a position each season. Just thinking out loud here - feedback most welcome.

[leewileyfan]

The final edit and review [1901-2011] is nearing completion, with inclusion of additional players in the original manuscript, through the 1961season. I hope to have 1954 updated and finalized later tonight. At the rate I'm going, 2011 will make it a "wrap" before the All-Star break.

Previously, I had submitted the book for possible publication, getting positive feedback on the initial Proposal and solicitation of the entire manuscript from a top NYC agent who was in the process of opening up his own "shop," going independent. He had it for eighteen months, did nothing to move it; so I was back at square one.

All positions [except PITCHER] are included in this study. I thank the 300 plus readers who have visited here over the last month, and especially those who have commented and/or raised questions.

I'm hoping to contact some folks with graphic design skills, illustration, and/or general print/electronic media layout and presentation ... to proceed with a publication effort.

It's a history of defensive play [1901-Present]. If you have questions, observations, player comps of the "who was better" [or worse] variety, please ask away. PLEASE, GUYS, I'd like some feedback, questions, comments, suggestions ... I may be aging, but I'm not fragile.

[brett]

To clarify the evaluation metric, the following are the inputs used [with reasons why] and weightings for those inputs [and why].

1. Since my primary two interests are to create a credible defense metric AND to have that metric useful in comparing players from different generations of play, I use ONLY those data that are readily available [and pretty well vetted] for the whole period [1901 to Present] for each position. Those are Put Outs, Assists, Double Plays, Errors, Passed Balls, Wild Pitches.

2. Put Outs [PO] are considered routine for infielders and catchers. In the outfield, the PO is the bread and butter of range, and mistakes at these distances from home plate and the bases are costly.

3. Assists are the bread and butter for C, 1B, 2B, 3B, and SS and I assign degrees of difficulty for each position. Likewise, errors are weighted by potential run damage by position.

4. Inputs for PO, A, DP, E, PB, WP are added, with the NET producing a raw number that is converted into a rating. The RATING intentionally resembles fielding percentage, for ease of communication.

I understand the assist focus for infielders, even for first basemen, but I have seen estimates that first basemen digging out throws is one of the biggest separators between first basemen AND that it doesn't correlate to first baseman assists-a guy can be great at coming up with throws but not necessarily good at range and throwing.
Also of course first baseman digs would have a big impact on his infielders assist rates.

[leewileyfan]

I agree with each observation. However, I am convinced that the pattern of 1B assists, along with errors committed and participation rates in completing DPs gives us a solid clue as to who deserve to be top-rated for being able to "dig" errant throws.

I believe that a close scrutiny of Ferris Fain at 1B is an excellent model for this. Those of us who have seen him in action PLUS those who have read media accounts of his play are consistently reminded of his aggressive, diving and digging hustle around the bag. He certainly made his fair share of errors; but the value was in runs he "saved" which were significant.

Another is Vic Power.

[bluesky5]

I understand the assist focus for infielders, even for first basemen, but I have seen estimates that first basemen digging out throws is one of the biggest separators between first basemen AND that it doesn't correlate to first baseman assists-a guy can be great at coming up with throws but not necessarily good at range and throwing.
Also of course first baseman digs would have a big impact on his infielders assist rates.

There is nearly no way to quantify that statistically without spotters at the field. I don't understand why MLB doesn't get actively involved with advanced statistic analysis. If anyone would be interested you'd think it would be them. It's a potential revenue stream. After all baseball's numbers are viewed with more reverence than that of any other sport.

[leewileyfan]

There is nearly no way to quantify that statistically without spotters at the field. I don't understand why MLB doesn't get actively involved with advanced statistic analysis. If anyone would be interested you'd think it would be them. It's a potential revenue stream. After all baseball's numbers are viewed with more reverence than that of any other sport.

A sportswriter of some note is 1920 wrote,"If you want to know the true value of an infielder, just look at those assists." [His name and newsprint citation are buried here somewhere in my notes, and when I come to that point in my editing, I'll cite it here].

Wise words from a keen observer. I'm sure he wasn't merely asking for a literal review of box scores, looking only in the "A" Column; but rather asking his readers to observe actual play, to see for themselves why the "A" column was so meaningful. [Kinda like "hiding in plain sight."]

[SHOELESSJOE3]

Some good stuff here Lee.
What I'm wondering, when comparing fielding stats is there any adjustment (which would be difficult) or consideration for fielding stats before and after 1920. Of course no way to know the number of throwing errors or muffed catches.

The reason I ask is because as you probably already know, a hugh improvement in glove design took place in 1920.
Card pitcher Bill Doak approaching Rawlings and selling his idea, a webbed mitt that came into the game in 1920.

Not only a web lace between the thumb and first finger but other improvements. A built up heel so after some use a pocket would be formed in the palm area from balls being caught in the palm area. Also some gloves had a pre-lubricated pocket.

Gloves before 1920 were stone age, 1920 was a whole different world for mitts. I took a quick glance at one stat only, fielding percentage, team and some individual postitions and only a very small number out of the top 25 were before 1920.
One look at some of the mitts before 1920, not very good, no web, flat and little flexibility, one hand catches had to be very difficult unless hit right at the fielder.

[leewileyfan]

Yes, Joe, CHANGE lies at the heart of the metric. When I first started doing data entry for 1901 players, I had hoped that the numbers would give me feedback, like the sudden or pronounced decline or rise in player ratings. Without trying to imbed that feedback, I figured I'd be in for a good deal of eye strain.

I was very grateful to see such feedback moments at several points along the road between 1901 and 2011, one very obvious one being defense trends between 1917 and 1924. [There had been one or two minor ones between 1901 and 1912]. Every time that happened, there was a very easy to find change in the game or its equipment. The intoduction of the "web" became a prominent feature in the metric.

[leewileyfan]

In a thread entitled Total Zone Runs, Tyrus....Cobb asked what we thought of that metric and if there was some credible metric for evaluating players across generations. He specifically cited 1B and the TZ position that Albert Pujols is the greatest defensive 1B of all time. I posted over 30 nominees from1901-Present, separated into playing generations, just listing thosw whom I thought might warrant consideration as "competition" for Pujols.

Last night I decided to prune that list by running defensive ratings for each 1B I had listed, to find how my metric measures all of these and Albert Pujols.

For 1B whose careers were almost entirely pre 1921 and the live ball and the introduction of the "web" into MLB mitts and gloves, I tweaked the value inputs slightly for PO, A and DPs [not a whole lot, but a bit], while leaving the E penalty alone. Therefter, since the original later glove innovations, like the basic "Trapper" design were introduced by the early 1930's, I essentially left all subsequent input models alone.

If any playing generation is discounted by these methods, I believe it would be the group that played between 1924 and 1962 [but I don't believ whatever bias would amount to very much].

This exercise is also, I hope, demonstrably responsive to those who have asked if my metric is sensitive to changes in the game.

Group 1: The Deadball Era, Pre 1921 ball change and web introduction to glove design.

Chick Gandil .967
Ed Konetchty .949
Stuffy McInnis .949
Jiggs Donahue .946 ... Gandil moves on vs. next group ...

Early Live ball Era:

George "Highpockets" Kelly .970
Jack Burns .955 ... Kelly and Gandil move on


P.S. ADD: Bill Terrry, inadvertently omitted = .971
Live Ball Pre-Post War; Pre-Expansion:

Frank McCormick .970
George McQuinn .961
Elbie Fletcher .965
Ferris Fain .988
Gil Hodges .975
Stan Musial .968
Vic Power .991

Fain, Power and Hodges move on, with Kelly, Gandil and Terry

Expansion Era:

Ernie Banks .967
Joe Pepitone .971
Donn Clendennon .980
Bob Robertson .984 [only 544 games]
George Scott .958
John Mayberry .964
Lee May .960
Keith Hernandez 1.101
Rod Carew .978
Chris Chambliss .972
Eddie Murray .973
Bill Buckner .983
Don Mattingly .977

Keith Hernandez, a TEMPLATE-buster, certainly moves on ... and Buckner [nothwithstanding that indelible WS film clip] was a darn good 1B.

Modern Day Ball:

Joyner .983
Grace .962
Clark .967
Palmeiro .968
Bagwell .964
Galarraga ..959
Helton .997
Olerud .960
Pena .960
Teixeira .956
Pujols 1.004

Here we have three possible candidates: Pujols, certainly; Helton so close as to be a virtual tie; and Wally Joyner, a reasonable candidate.

In sum, for this quick study, my metric would list these as the best defensive 1B [1901-Present]:

1. Keith Hernandez
2. Albert Pujols
3. Todd Helton
4. Vic Power
5. Ferris Fain
6. Gil Hodges
7. Wally Joyner
8. George "Hichpockets" Kelly
9. Bill Terry
10.Frank McCormick
11. Donn Clendennon

[brett]

Is it fair to even rate first basemen from the liveball era (or third baseman) along with those from the deadball era. I know its relative, but first base attracted higher level defensive players back then. Some have even suggested that First base was more important than second base until the live ball era. First basemen actually were lesser hitters than second baseman I believe until the live ball era-I think their OPS+ was lower meaning that it was reserved for worse hitting, better fielding players.

[brett]

Here is Matt Souder's rankings, based on his PCA system:

1: Keith Hernandez - Not really exciting to see Hernandez at the top again...but here he is. I don't frankly think there's any argument that can change this pick. There's just no getting around it...Keith Hernandez was magnificent with the glove and held his peak for longer than anyone in history. Look at this line:


2: Pete O'Brien - Now before you overreact to this not being a name you're familiar with - you should know that O'Brien lost his major league job because he stopped hitting, not because he failed in the field. A tragic example of a man playing a non-skill position outstandingly well and losing half his career to noodle-bat syndrome. Check out this career:


3: Roger Connor - For a while there, he was second on my list, but with the invention of the GI method, I found a better balance between longevity and peak performance and Connor slipped below the short but sweet career of O'Brien. In any event, Connor is underappreciated here, and often overlooked by his contemporaries in favor of Hal Chase, the world famous thrower of games.

4: Todd Helton - Through 2005, Helton was actually only 6th all time, but there's little evidence that he's slowed down much on defense since 2005 and that bumps him up ahead of Tony Perez for the 4 slot. Even now, Helotn is often credited for helping Troy Tulowitsky get his error rate down.

5: Tony Perez - A major part of the Big Red Machine, Perez glued that infield of misfits together and, for example, cut Joe Morgan's error rate significantly in the first couple of years of his reign in Cincinnati before he started losing his skills.

6: Harry Stovey - Stovey was in the AA what Connor was in the NL. The league's dominant defensive first baseman. Of course, Stovey was also very speedy and legged out a lot of triples and home runs. I've seen descriptions of him that call him one of the most flexible and speedy players of the 1880s. It's only logical that he should have been a great fielder.

7: Joe Kuhel - Here's a guy who shouldn't have stuck aroud as long as he did if you're looking only at his bat...but who kept getting work because of his glove. That career OPS+ of 104 looks better than it was...he had huge swings up and down from year to year but never managed to be a significant run producer aside from a sprinkling of 4 seasons non-consecutively through his career.

8: John Olerud - As a Mariner fan, I get a smile seeing Olerud this high on my list. "O-LE! O-LE O-LE O-LE! OOOOOOOH-LEEEEE! O-LE!" Those were the good years....*sigh* Dude was a magician with the glove and one of the toughest outs in baseball. He's got a shot at the hall of fame if the voters are fair.

9: George Kelly - 109 OPS+ from a first baseman who made the hall of fame. What do YOU suppose happened?

10: Mark Grace - Any Cubs fans in the house? You know he belongs here.

11: Vic Power
12: Fred Tenney

[leewileyfan]

I am hoping my faith in my metric is justified by this type of exercise. IF the metric recognizes the different demands of a position [and is responsive to same], then it can tacke this stuff without embarrassment.

Highpockets Kelly doesn't fare too badly in this exercise.

As far as Kelly's admission into the MLB HoF with a 109 OPS+, I suppose those in the selection committee concentrated on his overall performance between 1919 and 1930, after his rough start, and during which his teams went to the WS for four consecutive years. I figure that period probably produced a collective OPS+ of 113-115 or so.

He was, by reputation, a kind of fourth outfielder [or dual function infielder] in that he excelled in the cutoff and relay functions at 1B. In any event, the +/- input data in my metric picked up concentrations that upped his defense above all 1B who had preceeded him [after 1900] up to his time.

[leewileyfan]

In response to another metric's finding for a number of 1B, I reviewed them all, with these results:

Roger Connor and Harry Stovey were both 19th Century players, whose careers ended before 1901. I don't go back any further than 1901, when I believe the modern game of MLB, as we know it, came into being. Prior to 1901, there had been so many differences in rules and dimensions that I consider it a different game. Of the others:

Pete O'Brien .979 I completely overlooked. He rates well; but would not make the top 10, via my metric, unless I rigidly adhered to raw numbers, with absolutely no wiggle room for earlier guys with very close ratings and much longer careers. Then, too, strict adherence to numbers would have Bob Robertson ahead of him, along with Donn Clendennon.

Joe Kuhel [.952] was a steady and dependable 1B; but several from his generation of play [both leagues] would surpass him defensively, including Frank McCormick, Elbie Fletcher, George McQuinn and Zeke Bonura, when Zeke felt like playing defense.

John Olerud .960 and Mark Grace .963, both clearly above average defensively, by my metric do not rise to the level of player I listed earlier as "tops" [1901-Present].

[PVNICK]

Lee I don't know how you could possibly deal with this, but IIRC Buckner had gimpy ankles from before Boston. Many of his assists were pitcher flips due to his slowness rather than plays made too far from the bag (or at least that's how Bill James put it back in a mid 80s Abstract).