Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Hall of Merit > Discussion
Hall of Merit
— A Look at Baseball's All-Time Best

Thursday, April 02, 2015

Most Meritorious Player: 1904 Discussion

The Giants refuse to play the World Series this season and the AL pennant race makes them wish they hadn’t.

Player			SH WS		BBR WAR
Napoleon Lajoie		40.9		8.6
Honus Wagner		41.0		8.2
Bill Bradley		28.1		6.6
Elmer Flick		30.3		6.5
Jimmy Collins		27.3		5.3
Danny Murphy		25.5		5.1
George Davis		27.9		7.2
Bobby Wallace		23.1		5.7
Willie Keeler		26.7		5.0
Mike Grady		15.7		4.0
Tommy Leach		24.1		3.7
Frank Chance		29.0		5.9
Cy Seymour		25.3		4.6
Art Devlin		24.0		4.8
Freddy Parent		29.4		6.2
Sam Mertes		26.8		4.1
Dan McGann		22.5		4.4
Bill Dahlen		23.4		5.6
Harry Davis		20.2		4.5
Danny Green		26.9		4.2
Jake Beckley		23.0		4.2
Roy Thomas		26.3		4.6
Roger Bresnahan		22.3		3.5
Chick Stahl		30.8		4.5
Kid Elberfeld		17.6		5.4
Jesse Burkett		23.5		3.9
Dave Brain		17.5		3.1
Buck Freeman		24.9		3.1
Kip Selbach		23.9		3.5
Lee Tannehill		17.4		4.3
Joe Sugden		13.2		2.6

Pitcher
Jack Chesbro		53.5		11.0
Joe McGinnity		43.3		9.6
Rube Foster		??		??
Rube Waddell		33.5		9.7
Eddie Plank		29.0		9.0
Cy Young		34.7		9.8
Kid Nichols		26.6		7.5
Noodles Hahn		24.6		6.5
Christy Mathewson	34.4		6.1
Jake Weimer		27.7		5.8
Harry Howell		21.2		6.3
Sam Leever		22.4		5.4
Togie Pittinger		19.3		3.9
Jesse Tannehill		24.9		5.8
Patsy Flaherty		22.0		5.2
George Mullin		25.1		5.1
Bill Dinneen		25.8		5.2
DL from MN Posted: April 02, 2015 at 04:48 PM | 39 comment(s) Login to Bookmark
  Related News:

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. DL from MN Posted: April 02, 2015 at 05:07 PM (#4922578)
On Philadelphia Pete Hill hit .323/.344/.323/.666 but was outclassed by his teammate Sol White who hit .364/.400/.424/.824. Chappie Johnson hit .367/.406/.467/.873 and may be the best catcher in baseball. The best hitter on the team was Rube Foster who mashed .375/.400/.542/.942. I'm willing to believe Foster outclassed every other pitcher at this point in his career.

Grant Johnson had a very good season on the Cuban X Giants and Walter Ball had a good year pitching for them.

   2. Chris Fluit Posted: April 03, 2015 at 10:24 AM (#4922887)
1904 Prelim- AL Only

1. Napoleon Lajoie, 2B, Cleveland Naps: dominated the junior circuit with 203 OPS+ and 122 runs created
2. Jack Chesbro, P, New York Highlanders: Happy Jack's career year and 41 wins got him into the Hall of Fame but his 150 ERA+ and otherworldly 454 innings are only good enough for second place
3. Rube Waddell, P, Philadelphia Athletics: a league leading 165 ERA+ to go with 383 IP
4. Cy Young, P, Boston Americans: a top five ERA+ of 136 to go with 380 innings
5. Elmer Flick, RF, Cleveland Naps: 159 OPS+ and 94 RC are both second in the AL
6. Jack Powell, P, New York Highlanders: 390 innings are 2nd in the league; 112 ERA+ isn't bad
7. Eddie Plank, P, Philadelphia Athletics: a solid combination of rate (124 ERA+) and bulk (357 IP)
8. Bill Bradley, 3B, Cleveland Naps: a solid combination of offense (135 OPS+) and defense (+8 fielding runs)
9. Danny Murphy, 2B, Philadelphia Athletics: a big stick (134 OPS+) who isn't yet a liability at 2B (+3 fielding runs)
10. Freddy Parent, SS, Boston Americans: the best of the glove first candidates (+8 fielding at SS) because of what he adds with the bat (121 OPS+)

11. George Mullin, P, Detroit
12. Bill Dinneen, P, Boston
13. George Davis, SS, Chicago
14. Willie Keeler, RF, New York
15. Chick Stahl, CF, Boston
16. Frank Owen, P, Chicago
17. Bobby Wallace, SS, St. Louis
18. Harry Davis, 1B, Philadelphia
   3. DL from MN Posted: April 03, 2015 at 12:33 PM (#4922957)
1904 prelim

1) Napoleon Lajoie - huge separation from the pack in Dan R's WAR. Best hitter.
2) Honus Wagner - consistently high performance
3) Rube Foster - This is roughly Rube Waddell's pitching combined with George Mullin's hitting
4) Bill Bradley - Great fielder
5) Jack Chesbro - Happy Jack wasn't old, but he was a man
6) Joe McGinnity
7) Elmer Flick - best outfielder
8) Rube Waddell
9) Jimmy Collins - good fielding
10) Eddie Plank

11-15) Cy Young, Danny Murphy, George Davis, Bobby Wallace, Willie Keeler
16-20) Mike Grady, Tommy Leach, Frank Chance, Cy Seymour, Grant Johnson

Chappie Johnson might be a top 20 guy too. He might not be far behind Grady
   4. Chris Fluit Posted: April 03, 2015 at 02:52 PM (#4923114)
1904 Prelim- NL Only

1. Honus Wagner, SS, Pittsburgh Pirates: almost as dominant in the NL as Lajoie in the AL; OPS+ of 188 and Rc of 107
2. Joe McGinnity, P, New York Giants
3. Christy Mathewson, P, New York Giants: the two-headed pitching monster
4. Mike Grady, C, St. Louis Cardinals: catching bonus offsets lack of playing time; 167 OPS+ is second in NL
5. Noodles Hahn, P, Cincinnati Reds: 142 ERA+ in 297 IP
6. Jake Weimer, P, Chicago Cubs: 139 ERA+ in 307 IP
7. Jack Taylor, P, St. Louis Cardinals: 120 ERA+ in 352 IP
8. Cy Seymour, CF, Cincinnati Reds: the best combination of offense (135 OPS+) and defense (+8 fielding runs)
9. Frank Chance, 1B, Chicago Cubs: better rate stats than Seymour (150 OPS+) but fewer games played at less demanding defensive position hurts Frank just enough to drop him behind Cy
10. Kid Nichols, P, St. Louis Cardinals: the old guy's still got it

11. Jake Beckley, 1B, St. Louis
12. Roy Thomas, CF, Philadelphia
13. Oscar Jones, P, Brooklyn
14. Patsy Flaherty, P, Pittsburgh
15. Bill Dahlen, SS, New York
16. Roger Bresnahan, CF, New York
17. Mike Donlin, RF, NY/Cincinnati
18. Art Devlin, 3B, New York
   5. Chris Fluit Posted: April 03, 2015 at 03:57 PM (#4923161)
DL, you might be overstating Foster for this season. Every resource I've looked at suggested that Foster's pitching took a dip in 1904 before returning to form in '05-'06. He was still a very good pitcher, but not at Waddell's level for this year.
   6. DL from MN Posted: April 03, 2015 at 04:08 PM (#4923176)
Thanks. If I drop his pitching back toward Plank and Young his hitting still keeps him at 6th
   7. Chris Fluit Posted: April 03, 2015 at 04:12 PM (#4923179)
1904 Prelim- Combined

1. Napoleon Lajoie, 2B, Cleveland Naps: dominated the junior circuit with 203 OPS+ and 122 runs created
2. Honus Wagner, SS, Pittsburgh Pirates: almost as dominant in the NL as Lajoie in the AL; OPS+ of 188 and Rc of 107
3. Joe McGinnity, P, New York Giants: 168 ERA+ and 408 IP to narrowly edge out Chesbro as the top pitcher
4. Jack Chesbro, P, New York Highlanders: 150 ERA+ and otherworldly 454 innings are only good enough for second place among pitchers
5. Rube Waddell, P, Philadelphia Athletics: an AL leading 165 ERA+ to go with 383 IP
6. Cy Young, P, Boston Americans: a top five ERA+ of 136 to go with 380 innings
7. Christy Mathewson, P, New York Giants: the two-headed pitching monster
8. Mike Grady, C, St. Louis Cardinals: catching bonus offsets lack of playing time; 167 OPS+ is second in NL
9. Rube Foster, P, Philadelphia Giants: 218 OPS+ but only a 159 ERA+ in a bit of a down year on the mound
10. Elmer Flick, RF, Cleveland Naps: 159 OPS+ and 94 RC are both second in the AL

Harry Buckner also had a great year in the Negro Leagues, pitching and hitting at better rates than Foster though not playing as much (possibly because he was 31). He slots in ahead of Bradley for me at #16 overall.
   8. bjhanke Posted: April 06, 2015 at 03:08 AM (#4924291)
Well, this is disorienting. Jack Chesbro, being used as if it were 1884 instead of 1904, has ten more Win Shares than anyone else on the header, and 1.2 more WAR. And yet, no ballot so far has him #1, and DL has him #5. Before I start making up a list, I'd like to know why he just doesn't run away from everyone this year. That's what I assumed was going to happen. - Brock Hanke
   9. DL from MN Posted: April 06, 2015 at 10:35 AM (#4924389)
According to Dan R, Chesbro does not have as much value above his positional average as Wagner or Lajoie.
   10. Chris Fluit Posted: April 06, 2015 at 11:39 AM (#4924436)
Brock, like you, I figured that Chesbro would be an easy #1 after a quick scan of the WS and WAR lists provided in the thread header. So I was surprised when Chesbro didn't come out on top. There were a couple of factors. One, his innings may be mind-blowing but his rate stats are only really good instead of incredible. He's 4th in ERA and 3rd in ERA+. That allowed other pitchers to close the gap and Joe McGinnity, who had 400 innings himself, to leapfrog him. Two, the gap between Chesbro and the other pitchers (Waddell, Joss) wasn't nearly as big as the gap between Lajoie and Wagner and the other hitters. They absolutely destroyed their leagues. Chesbro is still the #1 pitcher in the AL but, for me, he got beat out by two position players and a pitcher from the other league.
   11. Michael J. Binkley's anxiety closet Posted: April 06, 2015 at 01:31 PM (#4924540)
1904 Preliminary ballot:

I'll be the 1st one to have Happy Jack on top, just edging out Nap.

1. Rube Foster (57.61)
2. Nap Lajoie (55.97)
3. Rube Waddell (50.36)
4. Honus Wagner (50.10)
5. Joe McGinnity(47.13)
6. Cy Young (44.49)
7. Rube Foster (38.43)
8. Eddie Plank (38.05)
9. Elmer Flick (36.58)
10. George Davis (34.44)
   12. DL from MN Posted: April 06, 2015 at 02:09 PM (#4924607)
You may want to edit #11
   13. Michael J. Binkley's anxiety closet Posted: April 06, 2015 at 02:55 PM (#4924668)
Okay - redo. It won't let me edit #11 now. I was copying and pasting, switching around names from my 1903 ballot and I never changed #1. Thanks for the heads up, DL.

1904 Preliminary ballot:

I'll be the 1st one to have Happy Jack on top, just edging out Nap.

1. Jack Chesbro (57.61)
2. Nap Lajoie (55.97)
3. Rube Waddell (50.36)
4. Honus Wagner (50.10)
5. Joe McGinnity(47.13)
6. Cy Young (44.49)
7. Rube Foster (38.43)
8. Eddie Plank (38.05)
9. Elmer Flick (36.58)
10. George Davis (34.44)
   14. lieiam Posted: April 07, 2015 at 10:36 PM (#4926028)
FYI- I know this is the wrong thread, but system not letting me post on the 1903 ballot thread.
Yeah, I missed again (2 years now after, I think, voting in all the previous MMP elections).
I'm still reading along by haven't managed to make the time to vote recently... Partly due to not having a good grasp of the Negro Leagues as well as any relevant minor league credit.
   15. bjhanke Posted: April 08, 2015 at 08:32 AM (#4926130)
Iieam - If it helps, putting any negro leaguer on the ballot other than Rube Foster is a wild guess. He was clearly the dominant player in negro baseball at the time, and MLEs are dicey, no matter how well done (Chris Cobb), because the quality of competition was so bad and so varied. And minor league credit really isn't much of an issue except for Rube Waddell, who is already in the majors. The main problem, as with the negro players, is that it is VERY hard to figure out the actual quality of minor league competition. Waddell came up in mid-season, and had plenty of major league play to justify just figuring that he was about that good in the minors, too. That doesn't happen very often. - Brock Hanke
   16. DL from MN Posted: April 08, 2015 at 10:05 AM (#4926192)
lieiam - Post a ballot wherever or you can send to me via e-mail and I'll get it posted. Balloting closes today for 1903.
   17. OCF Posted: April 09, 2015 at 02:47 AM (#4926971)
Without doing anything new, just drawing on my old files of RA+ equivalent records for pitchers, together with (in parentheses) a "Fibonacci win point" number for each such equivalent record.

Joe McGinnity: 33-12 (45)
Jack Chesbro: 35-16 (42)
Cy Young: 28-14 (32)
Rube Waddell: 27-15 (29)
Christy Mathewson: 26-15 (28)
Eddie Plank: 24-16 (23)
Jack Powell: 24-19 (18)
Sam Leever: 17-11 (17)
Addie Joss: 14-7 (17)
Mordecai Brown: 14-9 (13)
Doc White: 13-12 (9)
Al Orth: 11-12 (4)
Vic Willis: 18-21 (4)

A listing of single-year FWP scores of 40 or higher (no earlier than 1893):

Johnson, 1913: 32-6 (53)
Rusie, 1894: 36-14 (48)
Johnson, 1912: 32-9 (47)
McGinnity, 1904: 33-12 (45)
Chesbro, 1904: 35-16 (42)
Rusie, 1893: 36-18 (42)
Alexander, 1915: 31-11 (42)
Young, 1901: 30-11 (40) (asterisk for startup league)
Coombs, 1910: 29-10 (40)
Mathewson, 1908: 31-13 (40)

Note: for my "home team," the Long Beach State "Dirtbags," the highly effective surprise newcomer in this year's weekend rotation is named Chris Mathewson. Cool.
   18. John (You Can Call Me Grandma) Murphy Posted: April 10, 2015 at 08:04 PM (#4928566)
Prelim:

1) Jack Chesbro
2) Nap Lajoie
3) Honus Wagner
4) Joe McGinnity
5) Rube Foster
6) Christy Mathewson
7) Cy Young
8) Rube Waddell
9) Elmer Flick
10) Eddie Plank
   19. John (You Can Call Me Grandma) Murphy Posted: April 10, 2015 at 08:05 PM (#4928568)
I'll be the 1st one to have Happy Jack on top, just edging out Nap.

Exactly the same case for me.
   20. MrC. Posted: April 19, 2015 at 07:45 PM (#4934802)
NL All Star team (AL to follow)

C Mike Grady Had only 363 PA, was a below average fielder, but his outstanding hitting set him well above any other catcher in the league
1B Frank Chance Solid hitter, very good fielder
2B Johnny Evers only 22 years old. slightly below average hitter outstanding fielder
3B Tommy Leach Average hitter, outstanding fielder
SS. Honus Wagner best hitter in the league; however, fielding was below average
OF Cy Seymour
OF Roy Thomas
OF Sam Mertes

Honus Wagner and possibly Frank Chance were the only two position players in the NL who should make the ballot

Pitchers
Christy Mathewson Mathewson's hitting advanced him above McGinnity as the top pitcher
Joe MCGinnity
Kid Nichols
Noodles Hahn

All four pitchers could find a spot on the ballot.
   21. RicketyCat Posted: May 06, 2015 at 03:11 PM (#4949063)
Moving discussion of my method to an actual discussion thread.

There is no UberStat. No one number can give you the entire picture. All should be measured, weighed, calculated, mutilated, spindled, dried, etc. No player is the best every year. No one should be afraid to say that some highly regarded player wasn't as good as another simply because of the name in any given year. The best players are those that, over the entirety of a career they were better than at least half of the players of any given year. Those that were judged better than the highly regarded play once or twice are probably not going to have the kind of career that garners lifetime recognition (HoF, HoM). You can't upgrade a player in any given year just because you know that later they will be given such an honor.

Although WAR purports itself to be the UberStat, it cannot begin to explain reality. It presupposes that there is some fictional replacement player against whose fictional stats all others can be measured against. Since such a player does not, in actuality, exist all measurements against such a player are flawed. You can't base a philosophy on a fallacy. If the calculation were to incorporate an actual replacement player from that season then, and only then, could WAR be a contender. But how the heck do you find such a player?

In my method there is a mix of raw and calculated stats. I'll only discuss Batters for now as there is some cause for concern over my placing McGann over Wagner.
Raw - AB, R, H, 2B, 3B, HR, RBI, BB, CS (not really available for most years until 1914 so most everyone has a 0), SO, PO, A, E. Yep, I use raw, counting stats.

Calc - PB/G (catchers only with no detriment to those with no caught games), BAvg., OBP, SLG, RC/G (RC divided by games not Inn), FAvg. (THE bugaboo for most everyone trying to understand fielding), FOC, TEff (these two take some explanation)

15.5 batting and base-running, 5.5 fielding/6.5 with PB (with caught games); approx. 74%/26% or 70%/30% with caught games. 13 Raw, 8 calc (9 with caught games; approx. 62%/38% or 59%/41% with caught games. All have equal weight. Importance is a value judgement.

FOC - I evaluate a person's season on the position most played during that season. Those with multiple positions can benefit or find a detriment to their overall ordering by being effective at more than one position. Chances divided by games give the person's average chances per game. Sum of these averages at the position for the season divided by number of people evaluated at that position equals average chances per game per position (this introduces only a small problem due to the average of average fallacy, but the problem is below the threshold of the calculation (est. .0004 differences could be a tad higher, but still small)). Do this for each position being evaluated that season. Divide the result by the largest result to get a ratio (Rp). FOC is then calculated for each person as ((PO+A-E)*Rp)/(3*Games played). A result of 3 means that for the position they are being evaluated for they were adequate. Depending on how that fits in with the ranks and quartiles it could also mean they were great, but more on that later.

TEff - Total effectiveness. If it is true that you can't win if you don't score and you can't stop another team from winning if you can't catch and throw, then both must be of equal importance even if it is initially 1v1 then 1v9. TEff measures the number of STDev above or below average score for that season the person's RC/G and FOC scores are. For those worried about problems with Skew and Kurtosis, for seasonal totals RC/G has a 50% rate of non-normalcy and FOC has a 75% rate of normalcy from 1871 to 1905 using chi-testing. (I haven't completed my project and this is as far as I have gotten for testing. During election cycles (1904-1911 so far) both are evaluating to normal 100%)

Take all those stats and rank them appropriately. All of them. Every player that even came close to the field. Exception: Pitchers that have no PA do not get evaluated as a batter. Ranks tell you only half the story. Quartile those bad boys. Now you're getting somewhere. Assign a multiplier to each stat that falls within a certain quartile (4,3,2,1) and count the number of times that player's stat fall into that quartile. (Total ranks - adjustment)/(Total of Score cal. + adjustment) = ARS (Adjusted Rank Score).

Adjustment: it may be normal for a certain position to play fewer games. Maybe it's a personality problem inherent in the people that play the position, or the position is prone to injury, or they were drunk or a bad boy the night before, or the manager was just tired of listening to them and told, to, "Sit down and shut up!". For whatever reason the proper way to adjust for this is to find the ratio of the averages again but base it on games played.

Please notice there is no adjustment for parks, season, league, handicapped parking, or general jerkiness.

Ex. 1904 NL the average number of games played that was highest was CF at 106.667. for SS it was 104.846 and for 1B it was 96.833. This yields an and adjustment .017 for SS and .102 for 1B. This can amount to a good improvement for someone who has few games played with the reverses also true.

For determining MMP (or MVP if you prefer) the equation I use is 1 - (ARS/PLAYERStot)-((ARS-ARSmin)/(PLAYERStot/10)). I could use 1 - PercentDiff and be done as it would not change the relative positioning.

Now, as far as McGann and Wagner

Q1 Q2 Q3 Q4 Score Rank ARS MMP
18 1 2 1 80 606 7.564 95.39%
17 1 1 3 76 661 8.695 89.04%

(hmmm...some day I'll have to relearn how to do tables in bbcode)

The rank and quartile of SO and Errors are what hurt Wagner; the Errors more so than the SO as they effect r&q for E, FAvg., FOC, and TEff whereas SO effect r&q of only SO, RC/G and TEff. FOC for both were M:3.688 (Q1) and W:3.222 (Q2). Since both SO and Errors effect TEff, Wagner would seem to be in a bad position against McGann in that stat, but Wagner actually had a better TEff score (2.910 vs. 2.290 both in the first quartile (fairly bad fielding all around that year, I suppose)) because of the better batting. Neither gained great benefit from the adjustment. When Pitchers are added to the mix this puts McGann second and Wagner sixth.

There is another component I use that doesn't enter into the MMP (MVP) determination. It consists of the placement (1-10) within Position, Team (based on the percentage calc.), MMP (MVP), ROY (if first year player), traditional Triple (BAvg., RBI, HR), and new Triple (RC/G, FOC, TEff) with points assigned from 1 to .1 (for 1st to 10th). It puts Wagner ahead of McGann 6.0 to 5.6. Both were the top player on their respective teams. That Season.

So, yes, fielding is a major component of my method, and batting, while having more stats to compare, does not outweigh it. I hope this explains my reasoning.

Oh, yeah, to find the fictional replacement player: before adding in Acc, quartile games using Qgames-(Qgames/10), quartile the ARS scores and the player with the highest ARS among those in the ARS fourth quartile with games in the third is your guy. 1904 NL is John McPherson Pb (Games 15, ARS 55.053, -171.19%). Although not a replacement player as such (being a fairly regular backup pitcher), he fits the criteria of the replacement in batting. (If you're curious, the replacement among pitchers is John Brackenridge (Games 7, ARS 20.530, -160.16%))
   22. DL from MN Posted: May 06, 2015 at 03:51 PM (#4949130)
It looks like you have a really complicated garbage metric. There are good stats for offensive production. Taking counting stats and arranging them into quartiles obscures and quantizes the actual data. You turned variable data into attribute data which no statistician would do if they had the choice.
   23. RicketyCat Posted: May 06, 2015 at 06:27 PM (#4949286)
Not exactly sure what you are getting at. The counts are discrete the calculated are continuous. I'm not quartiling discrete and continuous data together. I'm quartiling each stat individually to find its relative strength within the data.

Now, if I had "garbage metrics" then the end results would also be garbage. During development, I tested election results against HoM election results. Let's compare it to the actual results of the vote of 1898.

1.  Deacon White       exact placement matchmatches 13 votes
2.  Paul Hines         exact placement match
matches 13 votes
3.  Joe Start          four places higher than vote results
matches 2 highest votes
4.  Ross Barnes        exact placement match
matches 6 votes
5.  Old Hoss Radbourne exact placement match
matches 3 votes
6.  John Reilly        not included by the voting body
7.  Tommy Bond         seventeen places higher than vote results
8 place higher than the 3.5 votes he did receive at 15
8.  Bobby Mathews      twenty places higher than vote results
6 places higher than the 1 vote he received
9.  Al Spalding        exact placement match
matches one vote
10. Charley Jones      six places higher than the vote results
1 lower than 2 highest votes
11. Pud Galvin         one place higher than vote results
lower then 5 votes he received
12. Jim McCormick      seven places higher than vote results
lower than 4 votesequal to 1 vote
13. Cal McVey          exact placement match
14. Hardy Richardson   six places lower than vote results
lower than all but one vote
15. Charlie Buffinton  not included by the voting body 


7 pitchers, 8 position players with two being 1B. With the exception of two all were named by the electing body. I don't think that that would produce the worst consensus score. Now, if my metrics were garbage then as they passed into the election process they would produce garbage. Six of 15 exact matches to an unpredictable voting body (although values were already known by the time I tested) without using data from before 1871 does not equal garbage out.
   24. DL from MN Posted: May 07, 2015 at 08:10 AM (#4949692)
Dan McGann ahead of Honus Wagner is garbage out.
   25. RicketyCat Posted: May 07, 2015 at 02:32 PM (#4950078)
I tried to rise above and have a discussion, but...

Well, I now no longer wonder why you can't get more than 7 voters into this part of the project. You're responses are basic teenager argumentation. If someone doesn't agree with you, specifically, the equivalent of the word, "whatever," is all you have. There is no discussion possible with a teenage mindset such as that. You have fun with the rest.
   26. DL from MN Posted: May 07, 2015 at 04:47 PM (#4950260)
Okay, I'll go into more detail with my critique

In my method there is a mix of raw and calculated stats. I'll only discuss Batters for now as there is some cause for concern over my placing McGann over Wagner.
Raw - AB, R, H, 2B, 3B, HR, RBI, BB, CS (not really available for most years until 1914 so most everyone has a 0), SO, PO, A, E. Yep, I use raw, counting stats.


RBI is garbage in, you're measuring contribution of teammates when trying to evaluate an individual player. Why would strikeouts double count versus any other out? That would work against power hitters in favor of guys who can't get the ball out of the infield. Why not save yourself trouble and use Total Bases instead of all the other counting stats? Putouts is clearly affected by position. Likewise Errors and Assists are affected by chances. Why no stolen bases but including caught stealing?

FOC - I evaluate a person's season on the position most played during that season. Those with multiple positions can benefit or find a detriment to their overall ordering by being effective at more than one position.


This is clearly garbage in. Why wouldn't you evaluate their defense versus games played at that position and evaluate hitting separately?

Chances divided by games give the person's average chances per game. Sum of these averages at the position for the season divided by number of people evaluated at that position equals average chances per game per position (this introduces only a small problem due to the average of average fallacy, but the problem is below the threshold of the calculation (est. .0004 differences could be a tad higher, but still small)). Do this for each position being evaluated that season. Divide the result by the largest result to get a ratio (Rp). FOC is then calculated for each person as ((PO+A-E)*Rp)/(3*Games played). A result of 3 means that for the position they are being evaluated for they were adequate. Depending on how that fits in with the ranks and quartiles it could also mean they were great, but more on that later.


What is this trying to accomplish?

TEff - Total effectiveness. If it is true that you can't win if you don't score and you can't stop another team from winning if you can't catch and throw, then both must be of equal importance even if it is initially 1v1 then 1v9. TEff measures the number of STDev above or below average score for that season the person's RC/G and FOC scores are. For those worried about problems with Skew and Kurtosis, for seasonal totals RC/G has a 50% rate of non-normalcy and FOC has a 75% rate of normalcy from 1871 to 1905 using chi-testing. (I haven't completed my project and this is as far as I have gotten for testing. During election cycles (1904-1911 so far) both are evaluating to normal 100%)


This looks like some sort of standard deviation adjustment but adjusting on different scales. Why not translate everything to runs and adjust based on runs?

Take all those stats and rank them appropriately. All of them. Every player that even came close to the field. Exception: Pitchers that have no PA do not get evaluated as a batter. Ranks tell you only half the story. Quartile those bad boys. Now you're getting somewhere. Assign a multiplier to each stat that falls within a certain quartile (4,3,2,1) and count the number of times that player's stat fall into that quartile. (Total ranks - adjustment)/(Total of Score cal. + adjustment) = ARS (Adjusted Rank Score).


This makes no sense to me. Why rank in quartiles when you're trying to determine who was the best player? Why does it matter which one ranked in the top quartile for put-outs?

Adjustment: it may be normal for a certain position to play fewer games. Maybe it's a personality problem inherent in the people that play the position, or the position is prone to injury, or they were drunk or a bad boy the night before, or the manager was just tired of listening to them and told, to, "Sit down and shut up!". For whatever reason the proper way to adjust for this is to find the ratio of the averages again but base it on games played.


Who do you use for averages? The top 10 players? There is a finite number of games, adding more players to the average will just bring down the number.

Please notice there is no adjustment for parks, season, league, handicapped parking, or general jerkiness.


I think failure to consider park effects is a mistake.

For determining MMP (or MVP if you prefer) the equation I use is 1 - (ARS/PLAYERStot)-((ARS-ARSmin)/(PLAYERStot/10)). I could use 1 - PercentDiff and be done as it would not change the relative positioning.


This is impenetrable word salad as far as I can tell. Is this some sort of best fit equation?

Now, as far as McGann and Wagner

Q1 Q2 Q3 Q4 Score Rank ARS MMP
18 1 2 1 80 606 7.564 95.39%
17 1 1 3 76 661 8.695 89.04%

(hmmm...some day I'll have to relearn how to do tables in bbcode)

The rank and quartile of SO and Errors are what hurt Wagner; the Errors more so than the SO as they effect r&q for E, FAvg., FOC, and TEff whereas SO effect r&q of only SO, RC/G and TEff. FOC for both were M:3.688 (Q1) and W:3.222 (Q2). Since both SO and Errors effect TEff, Wagner would seem to be in a bad position against McGann in that stat, but Wagner actually had a better TEff score (2.910 vs. 2.290 both in the first quartile (fairly bad fielding all around that year, I suppose)) because of the better batting. Neither gained great benefit from the adjustment. When Pitchers are added to the mix this puts McGann second and Wagner sixth.


So you're biased against shortstops because they have more chances and will have more errors than a first baseman. This is an anti-degree of difficulty adjustment.
   27. RicketyCat Posted: May 09, 2015 at 03:28 PM (#4951610)
I wasn't ignoring you and didn't run off to sulk. I had a medical issue that needed attending to. I also wanted to make sure I understood your position before answering back. I do have something written to answer your concerns, but it is lengthy. Also, some of my explanations assume you know Excel. It might even be easier if we can agree on a place for me to upload my 1904 evaluation file. Raw it's about 1M. Zipped would be smaller.

I would have sent a PM about this, but all that link seems to do is bring up my own profile page.
   28. DL from MN Posted: May 09, 2015 at 04:59 PM (#4951641)
No need to apologize. I don't expect real-time response. Lengthy responses are encouraged - ask Brock. There is a character limit so you may need to break it up.
   29. RicketyCat Posted: May 09, 2015 at 06:21 PM (#4951665)
First part (longish)

I've not kept some quotes as written simply so I can answer things more fully without unnecessary reiteration.

Two errors - one you caught one I caught. I did miss listing SB in there. Yes I do include it. Another error I made was mislabeling the quartiles. Quartiles should be listed as bottom to top 1 through 4. At some point I labelled them incorrectly in my file, never changed them and then forgot that fact when I wrote this up.

RBI is garbage in, you're measuring contribution of teammates when trying to evaluate an individual player. Why would strikeouts double count versus any other out? That would work against power hitters in favor of guys who can't get the ball out of the infield. Why not save yourself trouble and use Total Bases instead of all the other counting stats? Putouts is clearly affected by position. Likewise Errors and Assists are affected by chances.?

This makes no sense to me. Why rank in quartiles when you're trying to determine who was the best player? Why does it matter which one ranked in the top quartile for put-outs?


I include quartiles for the same reason Grades are given out in school. Since quartiles could actually move the curve and grant a D grade to the lowest score on a test which could be 73%. It's the same reason FAvg. is generally regarded as useless. I'd much rather toss it, but I include it because some still regard it as an indicator. Of what they can't be sure, but an indicator of something. In fact, quartiles puts anything under 91.6% in the D class For FAvg. in 1904. Which makes me wonder why you are up in arms about RBI and not FAvg.

RBI is not a measure of a teammate's ability. It is a count of how many times a person who hit has created an opportunity for a teammate. While it may be possible that that teammate would have had the same opportunity created by anyone else who had hit, it was that player who hit it that gets the credit. The teammate gets the credit for finishing the run which is appropriate to that teammate's ability.

SO do not double count. I'm not sure where you are getting this. SO do get figured into other stats which reflects the fact that a batter was unable to connect with the ball at all during that AB. Any other out is already counted in a positive fashion for the person who caught the ball that put the batter out and the batter is only charged with a hitless AB. This is already inherent in the basic stats.

Total bases is part of RC. Because I use RC/Games, keeping and evaluating the raw counts mitigates high RC/G for those with fewer games and fewer total components (H,2B,3B,HR) in the ARS calculation. A player might rank well in RC/G, but may not have enough counts to rank well in the components. This reasoning is also true for the rest of the "rate" stats.

I can't recall ever reading anything that correlates total chances to a consistent percentage of errors for any position. If you have a link I'd like to see it, please. It may be that I can use it to modify the calculation.

Yes, PO is absolutely affected by position. A can be affected by the ability of the thrower or the receiver. E is mostly affected by the thrower. Both A and E are skills based. To a smaller extent PO is also skills based as high-use positions, such as Catcher and 1B, must be able to field hard, slightly off target throws more often than other positions. Line-drive positions, 3B and SS and to a smaller extent 2B, must be able to stop balls in the air or pick up ground balls and reposition for throws (this is why they generally have roughly equal PO and A). The better they are able to handle those situations, the more PO they will accumulate.

FOC was designed to handle this.

This is clearly garbage in. Why wouldn't you evaluate their defense versus games played at that position and evaluate hitting separately? What is this trying to accomplish?


There are 44 separate evaluations in 22 categories for each batter. One rank and one quartile per category. A rank and a letter grade (with a value 4,3,2,1). Where is it that you think that I'm combining Hitting/base-running and Fielding except in TEff? It's possible that you may consider the total ranks and score (grades with value) as combinatorial when I calculate ARS, but I don't know if this is what you meant. If it is, I hope the following explains it a bit better.

One of these categories is FOC. FOC answers several questions.

1) How many chances is average for each position this season?
1a) in my Excel file the formula in AK2 is =IF($G2="1B",($AB86+$AC86+$AD86)/$H86,0) where G is position, AB is PO, AC is A, AD is E, and H is games played, which means, "If the position of the player evaluated is 1B then PO+A+E/Games."
1b) cell AK197 is =IF(SUM(AK$2:AK$197)]0,SUM(AK$2:AK$197)/COUNTIF(AK$2:AK$197,"]0"),0) which is the equivalent of saying, "If the sum of this column is greater than 0, then sum the values greater than zero and divide that sum by the number of times the value was greater than zero."
1c) columns AL through AT are for each of the other positions (for 1904 that's 2B, 3B, C, CF, LF, P, PH, RF, SS other years may include PR and DH - DH, PH, PR are included but may have no count in fielding as they did not play other positions and are not penalized in TEff for having no fielding)

--for 1904 the averages for 1B and SS are 9.376 and 4.735--

2) What is the ratio of each average to the position that normally has the most chances?
2a) cell AK198 is =IF(AK198]0,MAX($AK$198:$AT$198)/AK198,0)
2b) cells AL through AT are for the other positions
3b) the ratio scales the average of a position with a lower average to match that of the position with the highest average making it a scalar value and not an adjustment modifier

--for 1904 the ratios for 1B and SS are 1.000 and 1.980 --

3) Facts: there are 27 outs in a game, there are nine positions on the field for 3 outs per player. Assumption: the distribution of 3 outs awarded to each position in a game is highly improbable. Given that for this season there is now a known average value for each position and a scalar value to apply to those positions with a lower average, what is the the fielding value of the player at this position during this season?
3a) Cell AU2 is =(($AB2+$AC2-AD2)*HLOOKUP($G2,$AK$1:$AT$199,$BX$1+3))/(3*H2) where BX is the count of players being evaluated so it would find the scalar value at row 199 in the column for that player's position
3b) Trans: ((PO+A-E)*scalar)/(3outs*Games played) or scaled performance/expected workload if it were possible to have all 9 positions get 3 outs apiece with a standard result that should approach 3

So you're biased against shortstops because they have more chances and will have more errors than a first baseman. This is an anti-degree of difficulty adjustment.


There are no adjustments except the scalar value. Again, I can't recall ever reading anything that correlates total chances to a consistent percentage of errors for any position.

The following are approximations to the 3rd decimal and apply only to these four stats when calculated in isolation from hitting.

McGann: ((1481+94-15)*1.000)/(3*141), 1560/432, 3.688 this gives him a rank of 21 and a quartile (grade) of A (for 4 points) for this stat. That would be 21/4=5.250 for this stat only.
Including PO, A, and E would give: PO rank 2 grade A, A rank 51 grade B, E rank 137 grade D, and FOC rank 21 grade A. Sum of ranks is 211 and score total is 12. So, ranks/total score is 211/12=17.583

Wagner: ((319+376-51)*1.980)/(3*132), 1275.12/396, 3.220 this gives him a rank of 56 and a quartile (grade) of B (for 3 points) for this stat. That would be 56/3=approx. 18.667 for this stat only. Difference from McGann 18.667-5.250=approx. 13.417
Including PO, A, and E would give: PO rank 32 grade A, A rank 12 grade A, E rank 185 grade D, and FOC rank 56 grade B. Sum of ranks is 285 and score total is 12. So, ranks/total score is 285/12=23.750. Difference from McGann 23.75-17.583=6.167

Improvement of difference: 13.417-6.167=approx. 7.25 ARS gained by inclusion of components.

So SS has fewer chances than a first baseman, but the scalar value evens it out. Again, not including the components (PO, A, E) in the ARS calculation would negate any mitigation or benefit the higher A value gives to Wagner.

As a comparison:
2nd best SS in 1904 (by my method) Bill Dahlen: ((316+494-61)*1.980)/(3*145), 1483.02/435, 3.410 this gives him a rank of 42 and a quartile (grade) of A (for 4 points) for this stat. That would be 42/4=10.5.
Including PO, A, and E would give: PO rank 33 grade A, A rank 2 grade A, E rank 193 grade D, and FOC rank 42 grade A. Sum of ranks is 270 and score total is 13. So, ranks/total score is 270/13=approx. 20.769.

Best LF in 1904 (by my method) Sam Mertes: ((248+19-12)*4.704)/(3*148), 1199.52/444, 2.702 this gives him a rank of 113 and a quartile (grade) of C (for 2 points) for this stat. That would be 113/2=56.5.
Including PO, A, and E would give: PO rank 49 grade A, A rank 121 grade C, E rank 124 grade C, and FOC rank 113 grade C. Sum of ranks is 270 and score total is 13. So, ranks/total score is 407/10=40.7.

Note: I cannot determine why the Errors themselves occurred. I don't drill down to play-by-play and that would be the only way to determine whether it was because of a wide throw or a missed catch that was credited incorrectly. Looking at the Fielding of the rest of the Pirates that year I'd guess it would be a combination of both, but as I said, I can't tell you for sure.

At this point I hope that this isolation example tells you why, in the final calculation, hitting and fielding are not separate for purposes of ARS calculation.
McGann's final ARS was 7.589 and Wagner's was 8.695. Wagner's hitting was impressive, but not enough to close the gap against McGann's better than average hitting.

If it makes you feel better, in my files Wagner won in 1903 and McGann wasn't close to 10th and may not have been close to 20th.
   30. RicketyCat Posted: May 09, 2015 at 06:31 PM (#4951669)
Second part (not as longish)

Regarding TEff
This looks like some sort of standard deviation adjustment but adjusting on different scales. Why not translate everything to runs and adjust based on runs?


Okay. I want to make sure I'm reading your statement correctly: you complain earlier that I'm combining hitting and fielding and then you want me to convert fielding into runs and thus combine them. What?

TEff is a tertiary calulated stat whose components are RC/G and FOC. There are no adjustments. It simply measures the number of standard deviations above or below the average of each component the player's score is. There are different values for average and standard deviation for each component, but that would be expected since they are the average and standard deviation of different data sets.

This is an example from my notes file.
Ross Barnes, 2B 1871 NA Boston Red Stockings (Braves)
RC/G = 1.227, average for the stat is .515, stdev = .345
Since his stat is higher than average, (1.227 - .515)/.345 = 2.064
FOC = 3.317, average for the stat is 2.152, stdev = 1.130
Since his stat is higher than average, (3.317 - 2.152)/1.130 = 1.031
2.064 + 1.031 = 3.095 (actual calculation is 2.936 with the difference due to rounding in the example)

So, with hitting and fielding combined, Barnes was judged to be about 3 deviations better than an average player in that year. This doesn't necessarily mean he was the best player in 1871 since this is 1 of 22 components in ARS, but it sure means he's likely to be top 5.


Adjustment: it may be normal for a certain position to play fewer games. Maybe it's a personality problem inherent in the people that play the position, or the position is prone to injury, or they were drunk or a bad boy the night before, or the manager was just tired of listening to them and told, to, "Sit down and shut up!". For whatever reason the proper way to adjust for this is to find the ratio of the averages again but base it on games played.


Who do you use for averages? The top 10 players? There is a finite number of games, adding more players to the average will just bring down the number.



Perception of usefulness is reflected in the number of games a person is allowed to play by management. Whether due to injury or the perceived lack of usefulness at a position other than their main, if a player plays fewer games then they can not be judged as good as someone who does; this is especially true when assessing for whole seasons. Game by game the non-qualifying player could be judged better than that other, but over the course of a season the usefulness toward wins for their team is less.

Therefore, I'm using games totals from 196 players. Who are the top 10? The top 10 games played players? Wait, you're not suggesting you already know who is good before the assessment begins, are you?

As far as bringing down the average, that's exactly what I want. Under this system Catchers fail to reach such a minimum number games per season on a regular basis. With it, only three well-regarded Catchers has qualified as good from 1871 to 1905: White (short seasons, 6 times out of 8, 4 times best catcher),

McVey (short seasons, 2 times out of 4, 3 times best catcher), Ewing (2 times out of 8, 6 times best catcher). Charley Bennett and others may reach the minimum, but aren't as highly regarded. To help adjust ARS to give some benefit to Catchers (at least), some method needed to applied. Rather than plucking some

imaginary number out of the ether I developed this.

1904 NL adjustments

PosPlysGms.  G/Pl   Ratio        Rnk.
1B   12    1162  96.833  0.102 0.000 8
2B   14    1196  85.429  0.249 0.000 6
3B   14    1214  86.714  0.230 0.000 7
C    31    1519  49.000  1.177 0.000 3
CF   12    1280 106.667  0.000 0.000 10
DH    0       0   0.000  0.000 0.000 10
LF   16    1222  76.375  0.397 0.000 5
P    63    1484  23.556  3.528 0.000 2
PH    1       3   3.000 34.556 3.504 1
PR    0       0   0.000  0.000 0.000 10
RF   20    1350  67.500  0.580 0.000 4
SS   13    1363 104.846  0.017 0.000 9
                         9.861
                         STDev 


So, number of players at position, sum of games those players played in, average games for each player, the ratio of that position's average to the highest average. If the ratio is greater than the number of players then divide the calculated ratio by the standard deviation of those ratios. I don't know a mathematical justification for doing this other than to limit the adjustment for extremely small numbers of players in a position. I haven't gotten around to asking anyone if there is a valid algorithm that uses this particular technique and I admit that this seems a bit unorthodox. Then again, do you want the PH to be judged the best in the season due to such a huge adjustment?

The ranks are just there for my reference and don't enter into any other calculation. It just tells me which ratio is highest.


I think failure to consider park effects is a mistake.


I thought the idea was to measure relative skill. It may be true the home-field advantage exists, but it is more a psychological rather than a physical truism. No home team wins every home game. Any player unable to adjust to smaller or larger fields is unlikely to maintain high stats over the course of a season. While I agree the park effects would have an effect on the numbers, without full data on the park each game was played in that could be incorporated easily the calculation of this effect would be beyond the scope of my project. It would also be on a layer below the level of seasonal which is where my project ends


For determining MMP (or MVP if you prefer) the equation I use is 1 - (ARS/PLAYERStot)-((ARS-ARSmin)/(PLAYERStot/10)). I could use 1 - PercentDiff and be done as it would not change the relative positioning.


This is impenetrable word salad as far as I can tell. Is this some sort of best fit equation?


Each player has a percentage associated with them calculated as follows: 1 - (this player's calculated ARS divided by the the number of players evaluated) - (the difference between this player's ARS and the smallest ARS divided by one tenth of the number of players being evaluated). At the time I wrote that equation I was just trying to incorporate the size of the field building on the idea that 1st out of 10 is pretty good, but first out of 100 is dang impressive. The percentage difference equations are easily found around the web and would be more accurate to my needs, but when I tested it against my valuations I found that the relative positioning did not change except at the lowest ends.

I use the value of these percentages in determining the best in league ranking because the same thing is happening on the Pitcher side. In comparing the two, ARS becomes inadequate and needs conversion into a number that illustrates the same assessed skill vs. field of players as the Batting side.
   31. DL from MN Posted: May 11, 2015 at 11:56 AM (#4952706)
Wagner's hitting was impressive, but not enough to close the gap against McGann's better than average hitting.


This doesn't make sense. Wagner has 200 points of OPS on McGann in about 40 fewer plate appearances. If Wagner went 0 for his next 60 he still tops McGann at the plate in average, OBP and slugging.
   32. RicketyCat Posted: May 11, 2015 at 02:59 PM (#4952973)
If I were to use OPS or OPS+ it would still be only 1 of 23 or 24 other categories. I don't, so I'm not sure why you're even bringing it up.

Of course, this is why WS and WAR can only be used as indicators. Both would have you believe that because of his hitting ability, Wagner, alone and with no help from teammates, was responsible for 2 less then half of his team's wins that year. That's just not possible. Nap's performance is measured the same way with the same sort of impossibility at 3 less than half.

It is also why I don't use PA. It indicates the number of times a person was up to bat. If someone else gets caught during his appearance, or he flys-out, or sacrifices, or any number of other things, it is counted there and the batter loses an AB. Using it as a counter for anything other than to determine whether a pitcher qualifies for batting status or if the person was used as a PH or PR is just plain silly. AB is a better determiner as it indicates the person batting made something happen. Good or bad, at least they weren't just standing there winking at the pitcher.

Here's the full breakdown of ranks for both McGann and Wagner.

...           AB  R  H   2B 3B  HR  RBI BB  CS  SB  SO  PO   A   E   AVG   OBP   SLG    RC/G  Favg  FOC   TEff  
McGann
.       517 81 148 22 6   6   71  36  0   42  29  1481 94  15  0.286 0.354 0.387  0.601 0.991 3.688 2.290  
        Rank  21  11 9   13 32  3   7   31  1   4   136 2    51  137 22    24    25     15    29    21    10    604  7.646
        Grade 4   4  4   4  4   4   4   4   4   4   2   4    3   2   4     4     4      4     4     4     4     79 
Wagner        490 97 171 44 14  4   75  59  0   53  44  319  376 51  0.349 0.423 0.520  0.878 0.932 3.220 2.910  
        Rank  30  2  3   1  2   11  4   4   1   1   169 32   12  185 5     8     4      1     123   56    5     659  8.787
        Grade 4   4  4   4  4   4   4   4   4   4   1   4    4   1   4     4     4      4     2     3     4     75 


Just looking at the calculated stats someone might exclaim, "My god, Wagner kicked is backside all around the track!" And, yeah, he was better, but with mitigation from the components you find that he wasn't as much better than McGann as those calculated stats would seem to indicate. Before you point at 2B and say, "He was twice a good there!" I need to point out, again, it is 1 of 21 points of comparison.

The simple answer to your dilemma is that you rely too heavily on popular opinion and two over-rated indicators. The sun does not pop out fully formed from James' forehead nor does it set in his backside. I do not completely discount or discredit those indicators, but I believe that they incorrectly measure total skill because they rely on a non-existent entity as their first premise and nearly ignore anything to do with fielding.

You ignore almost all of what I carefully outlined, ignore the fact that I don't rely on any one count or calculated stat, and try to argue against me by pointing at a stat I don't even use. At this point, I can't make it any simpler for you. Can we just shake hands, spit on each others' shoes and call it a day?
   33. DL from MN Posted: May 11, 2015 at 04:34 PM (#4953079)
Not sure if you're trolling or not so I'll answer assuming you're just misguided here. OPS can be used because it correlates better than those other 24 categories individually do to run scoring. It is an effective shorthand. Players who post a higher OPS in the same number of plate appearances are better hitters. The only significiant flaw in OPS is it underrates OBP. In this case Wagner has a higher batting average, on base percentage and slugging percentage even if you give him an 0-fer to match McGann's plate appearances. He is without a doubt a much more valuable hitter. You SHOULD use OPS instead of looking at all the individual counting categories and believing they're all equally important. They certainly are NOT.

PA is more valuable than AB because walks have value. A hitter that drew a walk certainly did "make something happen".

The simple answer to your dilemma is that you rely too heavily on popular opinion and two over-rated indicators. The sun does not pop out fully formed from James' forehead nor does it set in his backside.


This is just trolling and isn't worth a response.

Yes, I ignore most of what you carefully outlined because quite a bit of it is either completely meaningless or incorrectly weighted when determining relative value. Thanks for opening up the inner workings of your method. I stand by my statement that you're using a garbage metric that does a bad job of determining value. There are dozens of freely available metrics that do a much better job. Feel free to vote using your garbage metric but it won't have any influence over my ballot.
   34. RicketyCat Posted: May 11, 2015 at 06:22 PM (#4953167)
You made it an argument, of sorts, by waving your hands and declaring it garbage. You end it the same way. I wasn't trying to change your mind. I didn't even need to go into so much detail as the only requirement is that I give some reasoning for the choices. As far as inner workings, this is the surface detail. If you can point out the flaws without pointing to single stats, using philosophical fallacies, or rhetorical nonsense, I'll be happy to listen. The following are just points and I neither require nor request an answer.

For some reason you think I'm Jonesey or someone equally reprehensible. I haven't trolled once. The most insulting thing I said was the equivalent of, "James is human." You don't like that. Oh well.

You keep saying weights. Where do you see weights? The only weights used at all is in the calculation for RC and those are universally accepted (whether I can currently find justification for those coefficients or not, they aren't mine). Maybe it's the lack of a weighting scheme that bothers you. Where are the tables that justify a weighting system? Am I supposed to grab some imaginary number as a coefficient to make the data say what I want it to say because I know this guy was better than that guy? Is that how data analysis works for you?

You say meaningless, and with your argumentation it reads more like you get it and don't really have an argument against it, so here's this straw man (OPS) that I'm going to wave around. If you think it's meaningless then point out why you think it's meaningless. Your argument points out the flaw in OPS then uses the flaw to justify why it's better.

Yup, walks have value and are used in RC. Where, exactly, do PAs fit in there? Unless it was an intentional walk, it has little to do with the skills of the batter. Of course you might be thinking of the niceties of foot placement, crouch, stance, and angle of attack which all limit the potential strike window and that would be a consideration in skill, but where are those in the stat tables? Nowhere? Well, goodness gracious me.

If you truly believe that this is the result of something I whipped up last week without consideration of other methodologies and without some logical justification, then you are sadly mistaken.
   35. DL from MN Posted: May 11, 2015 at 09:47 PM (#4953346)
Unless it was an intentional walk, it has little to do with the skills of the batter.


Swinging at strikes and not at balls has EVERYTHING to do with the skills of the batter. This is a ridiculous statement and you know it. If your assertion was true then pitchers would just throw pitches six inches off the plate every time. Batters would strike out every time and the only time they would get a hit is when the pitcher made a mistake and missed his spot. Intellectual dishonesty like this has to be intentional.
   36. Chris Fluit Posted: May 12, 2015 at 05:08 PM (#4954147)
RC, I'm sympathetic to your position. When I first joined the HoM, I tried voting off of an ink-based system, basically giving players credit for top ten finishes in key categories. I had a lot of back and forth arguments with other voters- and they weren't always kind. But I eventually realized that my ink-based system was replicating the mistakes of the HOF. I was voting for specialists rather than well-rounded players, when the well-rounded players were often superior. I later switched to a system that leaned a lot more on advanced stats. I still disagree with other HoM voters and I don't rely strictly on WAR (or much at all). That's okay. There's room for a difference of opinion within the voting body. But I'm also glad that other voters pushed me and helped me see my blindspots and assumptions.

Let's look at your McGann/Wagner breakdown in post #32. Honus Wagner beats Dan McGann in 11 offensive categories: runs, hits, doubles, triples, runs batted in, walks, stolen bases, batting average, on-base percentage, slugging percentage and runs created per game. In some cases, Wagner's advantage is quite large. For example, Wagner has twice as many doubles as McGann, 44 to 22, and more than twice as many triples, 14 to 6. Yet your system treats them as equals. You give both players 4 points each for doubles and triples. According to your system, Wagner's extra 22 doubles and additional 8 triples have no value. Does that actually make sense? Do 22 doubles and 8 triples have ZERO value?

There's a similar discrepancy in rate stats. Wagner leads McGann in avg by 63 points, in obp by 69 points and in slg by 133 points. Yet, once again, your system treats them as equals, giving both players a full 12 points for the rate categories. Do you really think that McGann's .354 on-base percentage is equal in value to Wagner's .423? Or his .387 slugging is just as good as Wagner's .520? I'm sorry, but that doesn't make any sense.

On the other hand, McGann is better than Wagner in 3 offensive categories- at bats, home runs and strikeouts. Both players get full marks for at bats and home runs, and both players are docked for strikeouts. However, Wagner is docked more than McGann, getting 1 point to McGann's 2. But here's the rub. Wagner leads McGann in 11 of 14 offensive categories while McGann leads in only 3 yet somehow your system has McGann lead Wagner in offensive totals, 58 to 57. You're basically arguing that McGann is the better offensive player because he had 33 fewer strikeouts. That's it. That's his only advantage. Can you see why we're having so much trouble with your argument? You're saying that McGann's advantage in strikeouts is more valuable than Wagner's advantage in runs, hits, doubles, triples, runs batted in, walks, stolen bases, average, on-base percentage, slugging percentage and runs created per game. No offense, but that's kind of ridiculous.

How about the defensive side of things? You included 4 defensive stats- putouts, assists, errors and fielding percentage. I already see one problem. You have 14 offensive categories and 4 defensive ones, which suggests that offense is three and a half times more important than defense. I get that there aren't as many defensive stats as offensive ones, but you've introduced an imbalance in your methodology that automatically favors offensive players. A glove-first guy like Bobby Wallace doesn't stand a chance in your system.

The bigger problem is that you treat all defensive counting numbers as equal. They're not. They're incredibly dependent on position. This isn't about having a position quota. It's about understanding the context of defensive numbers. Imagine a game with 27 groundouts- 9 each to the third baseman, shortstop and second baseman. Each of those players would have 9 assists while the first baseman would have 27 putouts. Does that make the first baseman three times better at defense than his teammates? Of course not. It's a structural aspect of the game that the first baseman will get the most putouts while other infielders get more assists. If your method only included putouts and assists, it might even out (though it still wouldn't tell us much).

However, you also included errors and fielding percentage. A shortstop will have more errors than a first baseman because he plays a more difficult position. And a first baseman will have a better fielding percentage because he has higher percentage of easy plays. You really have to look at defensive numbers within a position. For example, Honus Wagner was 2nd in putouts and 3rd in assists among shortstops; Dan McGann was 5th in putouts and 3rd in assists among first basemen. Those numbers are still somewhat team dependent but they're a much fairer comparison. McGann led 1B in fielding percentage while Wagner was only 4th among shortstops. However, Wagner was first in range factor for SS (which measures the number of plays per game) while McGann was only 5th among first baseman.

Basically, they were both above average defensive players, but Wagner was further above average among his peers while also playing the more demanding position. Yet your system penalizes Wagner, giving him only 11 defensive points to McGann's 13. You say that you don't have a position quota. That's fine. But you do have a position bias. Your system is set up to reward first basemen and penalize other infielders (second, third and short). This shows up in your 1905 ballot in which McGann is treated as a better defensive player than Wagner. It also showed up in your sample HOM ballot as first basemen like John Reilly placed well above consensus.

RC, this is not about a doctrinal adherence to WAR. I don't care about WAR. This is about inherent flaws in your system that ignore significant offensive advantages and penalize players at tougher defensive positions. I would love to have you continue to be a part of our project. But you need to correct these systematic flaws or your future ballots will continue to be illogical.
   37. RicketyCat Posted: May 12, 2015 at 11:17 PM (#4954417)
Thank you, Chris. Those are actually the answers I was looking for and I appreciate your efforts for articulating the concerns rather better than others have.

The intention for my project was to show how, with pencil and paper, the average person could construct a PHoM that falls somewhere in between the thought processes of the BBWAA and the BBTF electorate. It is a way of bridging the gap between common thought and SABRmetric thought. By providing a way to evaluate an election process it would be more easily understood by the average fan as to why a certain person was selected over another without resorting to such complex mathematics that they would not be able to construct it without a computer. It's a way of inviting people in, to generate more interest in and garner more exposure to a project I've always thought was worthy of wider recognition.

It was never actually intended to show best in league, although I track it as an ancillary line. I by no means think that McGann was actually the better player over his career or even the better player than Wagner over the span of years that I have completed in which they played at the same time (1896-1905), although my system says he's better in just this year (1904). Best at each position is a different award in my mind. You can't really drop everyone who wasn't the best at their position from consideration. I didn't think that it was the purpose in this part of the project to put forth the best player over a career. I would think the Best Player would be what the HoM was about. There is no doubt in my mind, although I haven't reached that part of my PHoM project, that Wagner will be a shoe-in for PHoM and McGann will be hunting hard for votes.

Going forward, I will be keeping your explanations in mind and seeing if the bias you point to is greater than the bias of the HoF and and Award voting electorates. Right now, through the years that I've completed, there is a bias to 1B in ARS, but I can't tell if the bias is greater than the award electorate since no MVP selections were made until 1911 and I'm 5 years off from there. And many more years from when they were a regular thing. The current percentage for the actual MVP awards: 1B - 17.99%, RF - 12.70%, LF - 11.64%, SP - 11.11% and all others at less than 10% each. Mine are (1871 to 1905) biased 1B, LF - 18.52, 2B - 16.67, and all others less than 10% each (more SP are in second place than any other position and it's possible that moving to the percent difference calc. would change that, but I haven't encountered a year wherein that change made that happen). The bias seems to be there, but I won't know how strong that bias is until I can compare it to the whole.

The logic I'm coming from is, given 100 students who take 21 tests of equal value, who ranks best by rank and who ranks best by a moved-curve grading system. Within that 100 the ranks will be fairly straightforward: 1 to 100 with possibly some ties. The margins between the quartiles (for the grades) are where the differences lay. We can simply picture that 1 through 20 will be safely within the A category (usually). Some drift from 21 through 25 is possible and some drift between 15 to 20 (though probably less often) is also possible. In the above example we can see that the drift for AVG/OBP/SLG does, indeed, drift upward to the A level and while they may equal each other in rate the are not equal in rank. Since there is no pre-selection of players the data is of the whole set. This may account for the closeness that the numbers show. Or; they are closer to each other than either of them are close to an average player.

Re: Offensive stuffs - disregarding TEff for a moment, I have Wagner behind by 1 rate (grade) point 66 to 65 and Wagner ahead 254 to 354 in ranks. So without fielding or TEff Wagner wins with 254/65=3.908 to McGann's 354/66=5.364. That's a fairly wide margin.

I actually have 5.5 defensive categories (FOC and half of TEff), but I understand your point. I think the percentage that I came up with (somewhere near 23%) came up after I read http://sabr.org/research/measuring-defense-entering-zones-fielding-statistics. I can't quite seem to find the numbers on that article, but it is strongly associated in my mind to a quote in someone's article stating 23%. As I admitted, those numbers are dependent on position. It is why I create the scalar value. I will need to go back and actually calculate the percentage-error on those averages, but I also already admitted that there was an error rate and that it was small (I estimated it and admitted that it might be higher). The averages are based on all others at that position and the scalar value equalizes to the position with the highest average.

Further than this I can't really go, mainly because I haven't delved into play-by-play. The project really does stop on the yearly level. I'm sure that If I were to go into game-by-game or play-by-play I would be able to see a way to include some modifier to account for position difficulty. I am an old-school "penalize for errors" kind of guy, which is frowned upon by UZR, RRF, and many other fielding metrics. To me, errors point to off target throwing, but where would an SS throw to? I would guess 1B because to think they are off-target that often to 2B is (0.o) worthy. I guess I would like to use DA, but not having BIS data (nor would the average at-home fan) I cannot use it.

In your example of the 4 positions, my method would only apply if you knew what the other 3B, SS, 2B, 1B did in the other games that day. If they were the only team in the only league and this was the only game, then this method would apply. The example is a bit hyperbolic, I haven't tested it against one so let's examine to see if hyperbole breaks it together. I'll assume no subs on the pitcher, after all, it's a perfect game, yeah? Even so, they are playing ghosts! ;)

For FAvg.: 1B - 27/27=1; 2B, SS, 3B - 3/3=1
FAvg.: all - 1,4

For FOC: Avg.: 1B - 27; 2B, SS, 3B - 3. Ratio: 1B - 1; 2B, SS, 3B - 9.
FOC: 1B - ((27+0-0)*1)/(3*1)=27/3=9 for 1,4 ; 2B, SS, 3B - ((0+3-0)*9)/(3*1)=27/3=9 for 1,4.

For TEff (field half) calc: stdev = 4.743 and average is 4. All - (9-4)/4.743=1.054 (-.843 for the other 5)
TEff (field half): all, 1,4.

(Something that hadn't come up: I expect errors, so 0 errors gets ranked as if they were the worst and rated as a D. This mainly because I haven't encountered anyone with significant field time that had 0 errors over a season other than the occasional pitcher. This may change if I ever encounter a player with a year where they were utterly phenomenal and did have a significant fielding season with zero errors.)

Total ranks,rates:

...    1B   2B   SS   3B   (5 others)
PO:    1,4  2,4  2,4  2,4  5,4
A
:     2,3  1,4  1,4  1,4  5,3
E
:     1,1  1,1  1,1  1,1  1,1
FAvg
.: 1,4  1,4  1,4  1,4  5,3
FOC
:   1,4  1,4  1,4  1,4  5,3
TEff
:  1,4  1,4  1,4  1,4  5,3
rank
/
rate:  7,20 7,21 7,21 7,21 26,17
FARS
:  .350 .333 .333 .333 1.529 


For ARS, lower is better.

Percentage differences show bias 4.97% over 1B and 128.46% over 5 others. Slightly off. My only defense on this is it isn't meant to show relative performance for any one position but relative usefulness to the team for that player at that position. It's possible that with an adjustment for percentage error on the ratios this will even out. I'll take a look.
   38. RicketyCat Posted: May 13, 2015 at 12:56 AM (#4954448)
Whoops, just realized I forgot to address Bobby Wallace. I have him as 2nd or 3rd best SS from 1897 to 1904 and best in 1905 and best on his team in 01 and and 04. The other half of my election cycle is Acc (accolades) he is averaging 2 Acc per year for a projected total of 50 and that puts him right in the range of election (which by 1924, his first election year, should have an average of 65 or so). I can't tell you his ARS yet as he isn't in the election evaluation range yet. It doesn't seem to have hurt him any and it seems his lifetime totals place him high. At a guess, he gets into my PHoM no later than 5th year in, but that is a guess.
   39. RicketyCat Posted: May 13, 2015 at 12:09 PM (#4954759)
Note to self: don't try to write online when you're tired. Made some egregious errors which probably only served to confuse. I'll try to correct.

100 students who take 21 tests of equal value


Should have been 100 students who take 21 tests of equal value to their final grade.

In the above example we can see that the drift for AVG/OBP/SLG does, indeed, drift upward to the A level and while they may equal each other in rate the are not equal in rank.


With 196 players evaluated in the league it means that: 196/4=49 should be in the A group with some sliding from 44 possible downward. There was no sliding from 20 to 25 as those are solidly within the A group anyway. The A value for those go down to .265/.321/.342.

The whole of the calculation from the example was screwed. Would you like me to post the correction to the ranks or correct for the First Baseman having 18 PO with no help? Never mind. I'll correct it both ways.

Correcting ranks

...    1B   2B   SS   3B   (5 others)
PO:    1,4  2,4  2,4  2,4  2,4
A
:     4,3  1,4  1,4  1,4  4,3
E
:     1,1  1,1  1,1  1,1  1,1
FAvg
.: 1,4  1,4  1,4  1,4  5,3
FOC
:   1,4  1,4  1,4  1,4  5,3
TEff
:  1,4  1,4  1,4  1,4  5,3
rank
/
rate:  9,20 7,21 7,21 7,21 22,17
FARS
:  .45 .333 .333 .333 1.294 


Percentage differences show advantage of 29.89% over 1B and 118.13% over 5 others.

Correcting to give other IF all Assists (9 each)

For FAvg.: 1B - 27/27=1; 2B, SS, 3B - 9/9=1
FAvg.: all - 1,4

For FOC: Avg.: 1B - 27; 2B, SS, 3B - 9. Ratio: 1B - 1; 2B, SS, 3B - 3.
FOC: 1B - ((27+0-0)*1)/(3*1)=27/3=9 for 1,4 ; 2B, SS, 3B - ((0+9-0)*3)/(3*1)=27/3=9 for 1,4.

For TEff (field half) calc: stdev = 4.743 and average is 4. All - (9-4)/4.743=1.054 (-.843 for the other 5)
TEff (field half): all, 1,4.

...    1B   2B   SS   3B   (5 others)
PO:    1,4  2,4  2,4  2,4  2,4
A
:     4,3  1,4  1,4  1,4  4,3
E
:     1,1  1,1  1,1  1,1  1,1
FAvg
.: 1,4  1,4  1,4  1,4  5,3
FOC
:   1,4  1,4  1,4  1,4  5,3
TEff
:  1,4  1,4  1,4  1,4  5,3
rank
/
rate:  9,20 7,21 7,21 7,21 22,17
FARS
:  .45 .333 .333 .333 1.294 


So changing the number of assists didn't actually change the outcome from the corrected version.

It might be beneficial to test against an actual perfect game as such a small sample size (9 players) makes PO look really weird, although it's working as designed. I'm not sure that 18 players will make it look much better in such an outlier kind of game, but I'm willing to take a look.

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Dynasty League Baseball

Support BBTF

donate

Thanks to
The Piehole of David Wells
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Syndicate

Page rendered in 0.6622 seconds
41 querie(s) executed