It is obvious from the post content in these threads that there are a lot of savvy folks in here with a deep understanding of baseball, its history, its facets and execution of each ... which can also vary by degrees of the good, the bad and the indifferent.
I am in the middle of wrapping up a research work that has been several years in the making. Its focus is purely on defense. It covers every position except pitchers [for no special reason ... just the way I approached it]. It dawns on me that METRICS are presented as self-contained universes, in which mathematical models are devised, sorted, sifted, regressed and presented as EVALUATIVE of performance against whatever standard the metric maker has selected. Defense is subject to all sorts of nuances.
So, the question[s] are [include] these:
1. Do you want a metric that has a broad historic perspective? [My time frame is 1901-Present].
2. Do you want a metric that allows [to whatever degree] player comparisons from different generations of position play? [OR, do you prefer CURRENT evaluative presentations [snapshots]?
3. Do you prefer a metric that attempts to interpret basic data inputs that are uniformly available across the entire time frame being considered? [OR, do you prefer more complex, in-depth mathematical interpretations in which uniformity of available data may be restricted by available research efforts [back to 1950 ... 1970 ... 1988]?
4. How importance are matters of EQUIVALENCE to you? [Focusing on defense here, some/all of the following]:
a. Pitching staff K's recorded as compared to League, both Leagues, between-among teams
b. Park factors, like playable foul territory or HR dimensions;
d. Pitcher fly ball, ground ball tendencies;
I raise the question of EQUIVALENCE especially with regard to pitching K's recorded. Yes, at the team level, a disparity of 500 K's by two competing staffs has a big impact on batted balls in play and fielder opportunity. By repeated application, I have found a K to be "worth" about .11 defense runs; so in the example cited, 500 * .11 = 55 defense runs [of fielding opportunity].
That K impact is distributed among positions; so that each position gets a piece.
I will try to word this precisely, to avoid any sense of bias:
1. IF a metric focuses on POSITION [individuality] and execution [as a rate of executing input bytes PO,A,DP,E,PB,etc.] do you want RATING to relate EXACTLY what happened on the field of play? IF the metric is constructed to adjust for evolving standards of expectation [equipment and design, playing surfaces, ball structure and dynamics] does that enhance the basic input product, or not?
2. Do you want ALL data input calculations adjusted for pitching K's and OTHER factors that "level" the mathematically ideal model in which INTERPRETATION of EQUIVALENCY considerations are your sole bottom-line concern?
Of course, any additional ideas/challenges are invited as well.