BBTF's Hall of Merit Discussion :: Pennants Added (updated for 1914 ballot)

Hall of Merit
— A Look at Baseball's All-Time Best

Tuesday, July 01, 2003

Pennants Added (updated for 1914 ballot)

Using Michael Wolverton’s formulas, I’ve calculated the Pennants Added for all of the candidates on this year’s ballot, as well as already elected HoMers.

I used both WARP3 and Win Shares (when possible) for hitters, just WARP3 for pitchers because Win Shares just have too many issues right now.

The Win Share rankings have some changes, now that I’ve added in Chris Cobb’s NA Win Shares for the players he’s calculated.

With WARP3, I made one modification, using a straight line adjust for the length of the schedule (to 162 games), instead of their formula of (162/scheduled games)^(2/3). This puts players from the 1870s and early 80s on a more equal footing.

The WARP1 numbers are adjusted for season length, but not league difficulty. At some point this week, I’ll run a correlation using birthdate with pennants added. One way to tell the degree of the ‘difficulty adjustment’ (at least the age component) will be to check and see whether WARP3 or WARP1 correlates well with birthdate (another way would be the mid-point of the players career, adjusted for schedule length, but that would take awhile to figure). A weaker correlation would show more balance with regard to the year players were born, which is good in my opinion.

The numbers will be in the discussion link . . .

HoMers in bold on all lists. New eligibles in italics.

Position players, ranked by pennants added, using WARP3 (includes pitching).

RK  Player          Pennants   adjWARP3
 1. Cap Anson         1.57     177.0
 2. Roger Connor      1.23     136.0
 3. Ross Barnes       1.23     111.9 (does not include play before age 21)
 4. Jim O'Rourke      1.19     138.1
 5. Ed Delahanty      1.13     120.7
 6. Jack Glasscock    1.05     116.1
 7. Paul Hines        1.05     114.2
 8. Dan Brouthers     1.03     112.8
 9. John Ward         1.03     113.5
10. Deacon White      1.00     112.3
11. George Wright      .92      94.9 (does not include play before age 24)
12. Bid McPhee         .92     108.0
13. Hardy Richardson   .88      97.1
14. Billy Hamilton     .87      97.4
15. Jesse Burkett      .86      98.0
16. Charlie Bennett    .85      94.2
17. Sam Thompson       .83      92.6
18. Buck Ewing         .82      93.7
19. Joe Kelley         .80      90.4
20. George Gore        .80      89.3
21. Mike Griffin       .77      87.7
22. King Kelly         .76      87.7
23. Ezra Sutton        .76      88.2
24. Fred Dunlap        .75      81.7
25. Jimmy Collins      .75      86.2
26. Jimmy Ryan         .75      89.1
27. Hugh Duffy         .74      85.6
28. Ed Williamson      .72      80.5
29. Fielder Jones      .72      83.9
30. George Van Haltren .70      83.6
31. Harry Stovey       .69      80.9
32. Lave Cross         .69      84.3
33. Mike Tiernan       .69      78.3
34. Cupid Childs       .68      77.5
35. Tom York           .66      77.5
36. Charley Jones      .65      70.2
37. Jack Beckley       .64      79.7
38. Herman Long        .64      76.7
39. Kip Selbach        .64      74.3
40. Hughie Jennings    .61      65.5
41. John McGraw        .58      65.9
42. Cal McVey          .56      62.4 (does not include play before age 20 or after age 28)
43. Ed McKean          .56      66.6
44. Billy Nash         .56      66.2 
45. Lip Pike           .54      59.1 (does not include play before age 26)
46. Joe Start          .54      64.6 (does not include play before age 28)
47. Bobby Lowe         .54      66.2
48. Pete Browning      .52      61.1
49. Steve Brodie       .51      61.6
50. Dummy Hoy          .51      61.5
51. Tom Daly           .50      62.1
52. Chick Stahl        .50      58.7
53. Elmer Smith        .47      55.2
54. Tommy Corcoran     .46      59.1
55. Kid Gleason        .46      56.5
57. Denny Lyons        .46      54.4
58. Chief Zimmer       .45      55.8
59. Bill Joyce         .43      50.2
60. Jack Clements      .41      50.7
61. Levi Meyerle       .41      46.0 (does not include play before age 25)
62. John Anderson      .40      49.7
63. Duke Farrell       .40      49.7
64. Patsy Donovan      .38      48.4
65. Bill Lange         .38      44.2
66. Deacon McGuire     .38      47.0
67. Socks Seybold      .38      43.5
68. Tip O'Neill        .35      39.9
69. Arlie Latham       .34      41.3
70. Dave Orr           .32      37.7
71. Sam Mertes         .29      35.7
72. Jack Doyle         .26      34.0
73. Emmet Heidrick     .26      31.6
74. Piano Legs Hickman .25      31.3
75. Tommy Tucker       .24      31.1
76. Tommy McCarthy     .22      24.9

Pitchers, ranked by Pennants added, using WARP3 (includes hitting and fielding).

RK  Player          Pennants   adjWARP3
 1. Pud Galvin         .94      94.1
 2. Kid Nichols        .88      98.2
 3. Amos Rusie         .87      88.6
 4. Jim McCormick      .81      85.1
 5. Jim Whitney        .81      78.3
 6. Hoss Radbourn      .80      83.9
 7. Tim Keefe          .71      79.6
 8. John Clarkson      .68      72.0
 9. Mickey Welch       .65      70.5
10. Clark Griffith     .57      65.7
11. Bob Caruthers      .47      53.6
12. Joe McGinnity      .47      54.2
13. Ted Breitenstein   .46      51.9
14. Al Orth            .45      55.0
15. Jesse Tannehill    .44      52.5
16. Noodles Hahn       .43      48.1
18. Al Spalding        .39      45.3 (does not include play before age 20)
19. Bill Hutchison     .35      39.9
21. Tommy Bond         .35      35.2
26. Tony Mullane       .28      31.9
30. Bobby Mathews      .08       3.7
I'm not going to carry anyone over that isn't in the top 15 that did not receive a
vote the previous year.

Position players, ranked by pennants added, using Win Shares.

RK  Player          Pennants  aWSrepl  aWS
 1. Cap Anson         1.69    570.9   742.9 (NA based on Chris Cobb's estimates)
 2. Jim O'Rourke      1.33    454.1   595.8 (NA based on Chris Cobb's estimates)
 3. Deacon White      1.23    406.2   495.0 (NA based on Chris Cobb's estimates)
 4. Dan Brouthers     1.15    382.9   479.4
 5. Roger Connor      1.14    380.5   492.4
 6. Paul Hines        1.11    377.3   500.3 (NA based on Chris Cobb's estimates)
 7. Jesse Burkett     1.00    340.6   444.6
 8. Ezra Sutton        .98    332.1   438.4 (NA based on Chris Cobb's estimates)
 9. Ross Barnes        .98    285.2   341.7
10. Ed Delahanty       .96    320.3   412.5
11. King Kelly         .95    321.8   414.0
12. Billy Hamilton     .93    310.9   393.3
13. George Wright      .86    268.4   333.4 (NA based on Chris Cobb's estimates)
14. George Gore        .85    286.2   369.6
15. Harry Stovey       .80    273.9   363.0
16. Joe Kelley         .75    262.4   349.8
17. Tom York           .75    262.9   317.3 (NA based on Chris Cobb's estimates)
18. Hugh Duffy         .75    257.7   342.5
19. Cal McVey          .74    237.8   297.1
20. Jack Glasscock     .74    260.4   358.7
21. Bid McPhee         .73    267.8   377.3
22. Jimmy Ryan         .73    260.3   360.9
23. Hardy Richardson   .72    251.8   335.8
24. Pete Browning      .72    241.5   307.5
25. Lip Pike           .72    223.6   275.7 (NA based on Chris Cobb's estimates)
26. George Van Haltren .71    254.7   352.7
27. Buck Ewing         .70    249.4   323.3
28. Jake Beckley       .68    251.2   362.8
29. Joe Start          .67    239.3   338.4 (NA based on Chris Cobb's estimates)
30. Fielder Jones      .66    235.7   319.1
31. Charley Jones      .64    217.9   283.2 (NA based on Chris Cobb's estimates)
32. Jimmy Collins      .64    225.2   304.7
33. Mike Tiernan       .64    222.2   296.2
34. Sam Thompson       .61    215.4   289.6
35. Herman Long        .61    217.2   308.4
36. Mike Griffin       .60    214.9   289.5
37. Cupid Childs       .59    205.2   277.8
38. Dummy Hoy          .57    207.3   297.6
39. Hughie Jennings    .57    193.4   251.2
40. Ed Williamson      .57    202.2   278.1
41. Lave Cross         .57    214.1   319.4
42. Fred Dunlap        .56    188.8   250.1
43. Billy Nash         .55    200.8   281.1
44. Tip O'Neill        .54    184.4   237.8
45. John McGraw        .54    188.1   240.7
46. Arlie Latham       .52    188.6   276.2
47. Ed McKean          .50    181.1   264.7
48. Denny Lyons        .48    169.3   224.3
49. Tom Daly           .48    178.5   249.7
50. Chick Stahl        .48    170.9   231.6
51. Charlie Bennett    .48    172.7   234.3
52. Kip Selbach        .45    165.1   241.6
53. Elmer Smith        .45    158.7   217.0 (does not including pitching)
54. Dave Orr           .43    146.6   187.7
55. John Anderson      .42    157.2   230.8
56. Sam Mertes         .39    142.7   195.6
57. Tommy McCarthy     .38    137.9   201.5
58. Deacon McGuire     .38    149.0   227.4
61. Duke Farrell       .37    142.8   210.3
64. Jimmy Slagle       .36    132.4   193.6 
65. Dan McGann         .36    135.4   196.8
70. Socks Seybold      .34    121.8   166.4
71. Jack Clements      .34    127.7   182.5
72. Chief Zimmer       .32    123.1   178.0
75. Kid Gleason        .19     73.8   161.5 (does not including pitching)
*note, the WS above subtract 6.5 per full season (based on league avg AB+BB per 162 G
in each year).

Position players, ranked by pennants added, using WARP1 (includes pitching, does not
account for difficulty of league).

RK  Player           Pennants   adjWARP1
 1. Cap Anson          2.90     290.8
 2. John Ward          1.99     191.6 (includes pitching)
 3. Jim O'Rourke       1.98     209.9
 4. Ross Barnes        1.97     159.6 (does not include play before age 21)
 5. Roger Connor       1.92     193.0
 6. Jack Glasscock     1.83     181.5
 7. George Wright      1.83     160.3 (does not include play before age 24)
 8. Deacon White       1.76     178.1
 9. Bid McPhee         1.70     179.4
10. Paul Hines         1.69     166.4
11. Dan Brouthers      1.63     164.4
12. Ezra Sutton        1.60     164.2
13. Ed Delahanty       1.55     155.9
14. Harry Stovey       1.46     147.8
15. Hardy Richardson   1.34     137.5
16. Joe Start          1.34     142.7 (does not include play before age 28)
17. Tom York           1.34     140.2
18. King Kelly         1.29     136.0
19. Buck Ewing         1.28     134.9
20. Charlie Bennett    1.26     131.4
21. Fred Dunlap        1.24     119.3
22. Charley Jones      1.24     119.5
23. Jesse Burkett      1.22     132.6
24. Ed Williamson      1.22     124.4
25. Sam Thompson       1.21     126.2
26. Billy Hamilton     1.18     125.5
27. Herman Long        1.17     127.6
28. Mike Griffin       1.17     123.4
29. George Gore        1.16     121.1
30. Hugh Duffy         1.14     121.8
31. Jimmy Ryan         1.11     124.9
32. Joe Kelley         1.11     124.2
33. Lave Cross         1.10     127.5
34. Jake Beckley       1.10     127.3
35. George Van Haltren 1.09     122.2
36. Pete Browning      1.08     110.9
37. Cal McVey          1.07     105.6 (does not include play before age 20 or after age 28)
38. Jimmy Collins      1.05     114.1
39. Ed McKean          1.04     111.8
40. Mike Tiernan        .99     106.7
41. Tommy Corcoran      .98     114.3
42. Cupid Childs        .98     103.8
43. Hughie Jennings     .98      97.3
44. Lip Pike            .95      91.1 (does not include play before age 26)
45. Billy Nash          .94     104.5
46. Fielder Jones       .92     103.8
47. Kip Selbach         .91     100.0
48. Kid Gleason         .86      98.4
49. Arlie Latham        .85      94.7
50. Denny Lyons         .84      89.5
51. Elmer Smith         .83      89.8 (includes pitching)
52. Bobby Lowe          .80      94.3
53. Tip O'Neill         .79      81.7
54. Dummy Hoy           .79      89.6
55. Steve Brodie        .78      88.2
56. Levi Meyerle        .78      76.5 (does not include play before age 25)
57. John McGraw         .78      85.1
58. Dave Orr            .77      77.3
59. Chief Zimmer        .74      87.2
60. Deacon McGuire      .72      85.5
61. Jack Clements       .71      82.4
62. Tom Daly            .71      83.8
63. Duke Farrell        .69      81.5
64. Chick Stahl         .67      75.1
65. Patsy Donovan       .64      76.7
66. Dan McGann          .62      72.1
67. John Anderson       .62      72.2
68. Bill Joyce          .58      67.2
69. Tommy McCarthy      .57      63.7
73. Socks Seybold       .51      56.6
77. Jimmy Slagle        .46      53.9
78. Charlie Hickman     .36      43.5

Pitchers, ranked by Pennants added, using WARP1 (includes hitting and fielding, does
not account for difficulty of league).

RK  Player          Pennants   adjWARP1
 1. Pud Galvin        1.87     169.2
 2. Al Spalding       1.76     143.7
 3. Tim Keefe         1.75     161.5
 4. Jim McCormick     1.71     150.8
 5. Bobby Mathews     1.60     150.3
 6. Charles Radbourn  1.57     141.4
 7. Tony Mullane      1.45     135.7
 8. Tommy Bond        1.43     126.5
 9. Jim Whitney       1.40     124.1
10. Kid Nichols       1.37     140.6
11. Amos Rusie        1.32     122.7
12. Mickey Welch      1.30     126.4
13. John Clarkson     1.27     120.4
14. Bob Caruthers     1.09     106.3
15. Clark Griffith     .82      91.2
17. Joe McGinnity      .76      82.3
22. Bill Hutchison     .66      67.9
23. Al Orth            .65      77.0
25. Jesse Tannehill    .61      70.7
26. Noodles Hahn       .61      63.8
I'm not going to carry anyone over that isn't in the top 15 that did not receive a
vote the previous year.

Average pennant winners, average teams and standard deviations based on NL teams from 1876-1913.

JoeD has the Imperial March Stuck in His Head Posted: July 01, 2003 at 04:45 AM | 74 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. MattB Posted: July 01, 2003 at 01:19 PM (#514927)

I assume that the 6.5 WS per full season represents the Win Shares a "replacement level player" would have earned.

2. karlmagnus Posted: July 01, 2003 at 01:19 PM (#514928)

I don't trust these numbers. Spalding in particular looks way out, as he was the leading pitcher, nearly the ONLY pitcher, on FIVE pennant winners. If all pennants are equal, then his pennants added should be about 2.0, not 0.4. These numbers are not giving us what they purport to give us, IMHO.

3. MattB Posted: July 01, 2003 at 01:26 PM (#514929)

Very interesting numbers. Just goes to add to my confusion of the correct ordering of Sutton, Glasscock, and Richardson. Using WARP-3 numbers gives you Glasscock, Richardson, Sutton. WS gives you Sutton, Glasscock, Richardson.

I am surprised that Mike Griffin does so well on both of the charts, and I will have to reconsider Whitney and McCormick, based on the pitching data.

Thanks for the data. Just got to figure out how to use it now!

Also, did you just throw your hand up trying to find pennants added for pitchers using Win Shares?

4. Carl Goetz Posted: July 01, 2003 at 02:32 PM (#514930)

'Spalding in particular looks way out, as he was the leading pitcher, nearly the ONLY pitcher, on FIVE pennant winners. If all pennants are equal, then his pennants added should be about 2.0, not 0.4.'

Spalding seems about right to me. They won 5 pennants, but with that All-star team, in that time-period, I think a Replacement level pitcher would have won 4 and possibly all 5 pennants. There is no disputing that he was the best pitcher of the era, just a dispute over how important pitching really was to teams then.

5. karlmagnus Posted: July 01, 2003 at 03:24 PM (#514932)

Joe, the non-Spalding pitchers in the pennant years went 23-9 (.719) which tends to validate your claim -- though one second place team, Philadelphia 1875, managed to do better at .726. But the 1877 Chicago team, minus Spalding and White but with everybody else, went 26-33 (.441) down from 52-14 (.788) in 1876. That suggests that Spalding and White together added .347 to the team's WP. I wouldn't be shocked at a Spalding calculation that gave only 1.2 or so penannts added, but if he was only 0.4 where did the other 4.6 pennants come from? -- there simply weren't enough good players on the team to account for the difference.

6. MattB Posted: July 01, 2003 at 03:39 PM (#514933)

Looks like, from a Win Shares perspective, we've been doing pretty well (although we may need to do a Gray Davis-style recall vote on Buck Ewing.)

WARP3 doesn't like NA pitchers at all. I disagree with the extent, but I can see where they are coming from. Defense was definitely more important then.

On another note, WARP-3 seems to justify my high ranking of Charlie Bennett (0.86 pennants), but Win Shares does not (0.48 pennants). Can anyone think of any rationale for the big difference?

7. Jeff M Posted: July 02, 2003 at 12:45 AM (#514938)

If someone wants to make a downward adjustment for the AA (in Joe's WS/Pennants Added calcs), I think they'll need Joe's spreadsheet. They'll need to discount the WS before running the pennants added formula, based on their own evaluation of what discount applies.

For instance, if you think AA gets a 10% discount, I'm not sure it works to discount Pennants Added by 10%. It might be too small or might be too large a discount to apply to the derived Pennants Added. We should probably discount the WS by the appropriate amount (10% in this example) and then run the Pennants Added calc, or better yet, discount the RC, then run the WS, then run Pennants Added (though that seems like a heck of a lot of trouble).

Also, I will again caution folks about applying Davenport's AA discount to WS or any other non-WARP3 measure. We don't know how the discounts are derived and there doesn't seem to be a single discount figure for particular years. Also, remember that Davenport discounts every other league and every other season, so it is all relative to whatever his base year is. For instance, if Davenport discounts 1882 AA by 30% (and I'm not saying he does), then he's probably discounting 1882 NL by 10-15%. So you can't use his discount for one league without factoring the discounts for the other leagues.

Joe, what did you finally use as Avg Pennant Winners Win Pct., StDev Pennant Winners Win Pct. and StDev Average Team Win Pct? I know it's different each year by a little bit, so maybe give me a single year after the leagues were established, like maybe 1890 or something. Just curious.

Also, did anyone figure out the "Peaks" formula? I couldn't make the math work when Tom H provided his version of the formula.

8. Rusty Priske Posted: July 02, 2003 at 03:42 PM (#514939)

Where are these results posted?

9. Rusty Priske Posted: July 02, 2003 at 03:43 PM (#514940)

That's odd. As soon as I posted that question, they appeared.

10. OCF Posted: July 07, 2003 at 11:03 PM (#514944)

Where's Mike Tiernan on the WS version of the chart?

As a FOHS, I suppose I'll have to figure why the WS version of this is so much friendlier to Stovey than the WARP3 version.

11. jimd Posted: July 08, 2003 at 12:39 AM (#514948)

Joe, I think that's an excellent description of the WARP process, much better than the cryptic notes available in Davenport's glossary on the BP site.

However, I also think that what you're describing there is the WARP-1 calculation. If you read those cryptic notes carefully, he says that those pitching/fielding adjustments are built into WARP-1. WARP-2 takes the WARP-1 numbers and adjusts them for league quality (NL better than contemporaneous AA; NL of the 1880's better than NA of decade before; etc.). WARP-3 then adjusts WARP-2 for season length, sort of, using that funny "2/3's" formula.

I made a previous post with various links to Davenport's glossary that I believe help make my case. I think the PRAR is most pertinent; it is part of the "Advanced Pitching Statistics" which are copied verbatim to calculate WARP-1 in the "Advanced Batting Statistics" (Sample player card) Also interesting are these additional links, Difficulty Adjustment and Pitching/Fielding Breakdown (they are also somewhat contradictory as to whether the breakdown is 50/50, 70/30, or 75/25).

12. jimd Posted: July 08, 2003 at 12:39 AM (#514949)

A separate but related question here. Does anyone know if Davenport documents the slope of the 1947-2002 trend-line (see Difficulty Adjustment)? And if so, where? Or at least what the slope of the trend-line is?

13. MattB Posted: July 28, 2003 at 12:54 PM (#514954)

I assume that the bottom chart (pitchers) should be WARP-1 rather than WARP-3.

I'd still like to be able to see how pitchers' Win Shares come out, even though we know there are problems with them.

14. jimd Posted: July 28, 2003 at 07:48 PM (#514955)

Joe, thanks much for the WARP-1 calculations. Bobby Mathews is a surprise; I suspect Tommy Bond is an omission.

It looks like the "weakest" players selected by these metrics are: Buck Ewing (Win Shares), George Gore (WARP-1) and John Clarkson (WARP-3). None were hotly debated, their reputations remained intact.

15. Carl Goetz Posted: July 29, 2003 at 07:22 PM (#514957)

Joe,

16. MattB Posted: September 04, 2003 at 05:08 PM (#514960)

Not to be greedy, but I think that the numbers posted here have been very useful (although I use them as much for my own purposes as for the use that Joe intends), and I am currently loathe to post even a preliminary ballot until I see where some of the new guys fall.

17. jimd Posted: September 04, 2003 at 07:44 PM (#514961)

Joe, if you ever get some free time (it's in short supply for me too lately), it'd be nice to see Tommy Bond's WARP-1 numbers. Thanks.

18. John (You Can Call Me Grandma) Murphy Posted: September 08, 2003 at 05:18 PM (#514965)

I have been using 6.5 WS per season (full 162 game season) as my replacement level. Should it be higher? Or Lower? Any thoughts?

You don't have different replacement levels for each position? Better question: does WARP? The level for a catcher shouldn't be the same for a first baseman, IMO.

19. MattB Posted: September 08, 2003 at 06:22 PM (#514966)

Assuming that, in general, "replacement player" means the level of play below which no player (on average) will ever be able to play in the majors, then I would say that replacement level should be based on the worst team ever.

The worst record for a modern team is the 1962 Mets (pending the end of the Tigers' season), and they went 40-120.

Assuming 120 Win Shares to go around, the average '62 Met earned 13 1/3 Win Shares per 162 (actually 160) games. (120/9).

Since Win Shares are calculated over the course of a season, the appropriate replacement should be the worst season ever. 6 1/2 is Cleevand Spiders level, which is too low (although it may not seem so in 1909). Cleveland was bad on purpose, as all of its best players were given to St. Louis, and many of those remaining were well below replacment level (the starting rotation was a combined -0.8 WARP).

20. MattB Posted: September 08, 2003 at 06:39 PM (#514968)

Oops, forgot to divide that number by about 2/3, since about 1/3 goes to the pitcher.

Make that a replacement level of around 9 WS. (Of course, I'm for putting more weight on an "average" baseline, but that's just me.)

21. Chris Cobb Posted: September 08, 2003 at 09:12 PM (#514970)

Jimmy Ryan appears to be missing from the aWS pennants added list.

22. DanG Posted: September 15, 2003 at 01:05 PM (#514973)

I don't see John Ward on the aWS list. Is this an oversight?

23. Chris Cobb Posted: September 24, 2003 at 03:40 AM (#514975)

Joe,

Does Van Haltren's Pennants Added according to WS include his pitching?

24. tangotiger Posted: October 03, 2003 at 03:06 PM (#514977)

Tom, I know that Prospectus says their replacement level player is a replacement level hitter plus an average fielder, but Tangotiger disputes this in his interpretation. I definitely agree with your thoughts what dropping 2 WARP per season would do to the raw WARP rankings.

The Prospectus players WAR is, as best as I can tell, replacement level hitting + replacement level fielding.

Gary H at BP was quoted by a Primate as saying that they do replacement level hitting + average level fielding.

However, that's not what their site shows. If you click on the above link, you will get for Mike Schmidt:

25. tangotiger Posted: October 03, 2003 at 03:27 PM (#514978)

Go to the above link, and look at Tim Wallach's 1993. He was

26. tangotiger Posted: October 03, 2003 at 04:38 PM (#514980)

All you really have to do is take say players from 1993 to 2002 (or any period you want). Then, take all the regulars from that time period by position, and compute their:

27. tangotiger Posted: October 03, 2003 at 04:52 PM (#514981)

How about the 1985 Jose Oquendo? +38 runs above replacement hitting. +18 fielding runs above average. That's +56 runs above true replacement. Checks in with 9.3 WARP.

Duane Kuiper, 1978? +10 hitting runs above replacment, -37 fielding runs relative to average. +.5 WARP. Why? Only -6 fielding runs relative to replacement.

************

I have a biggeer issue with how the replacement level is set (annually, instead of by era, but that's another topic).

28. tangotiger Posted: October 03, 2003 at 05:06 PM (#514982)

In fact, what you will find is that with those positions where the replacement level is like 30 runs below average, that this will get doubled to 60 runs below average. This is shown in the rather shocking totals from the nonshocking seasons of Kuiper and Oquendo above.

I will guess that in the above Kuiper example that he is actually -60 runs relative to average 2B, overall, yet, he comes in at close to zero to replacement.

29. RobC Posted: October 03, 2003 at 05:18 PM (#514983)

The BP site is pretty clear (too me, I guess not to everyone else, including GH) that replacement is repl hitting + repl fielding + repl pitching. As there description points out, a team of this would give you the Spiders not the Mets (or Tigers). If you consider the Spiders to be a team of replacement level players (and not a team that is a mix of replacement and below replacement level) then there system is "right".

30. RobC Posted: October 03, 2003 at 05:26 PM (#514984)

Warning: if too/to/two/there/their/they're problems drive you crazy, do not read post #58.

Sorry about that. Uggh.

I have never screwed up lose/loose though.

31. Tangotiger Posted: October 03, 2003 at 06:49 PM (#514986)

I agree,it is clear what they are doing, and I said as much at the original Studies thread.

However, that replacement level (essentially 40 runs below average) is by far the lowest replacement level I have ever seen, and one which I would guess almost no one would agree with.

I'm pointing out that

32. Marc Posted: October 03, 2003 at 07:39 PM (#514987)

With all "do" respect ;-) can somebody give me an English language translation as to what this really means for WARP? Can the average person adjust it somehow to correct for these problems? Or is it just useless?

Thanks.

33. RobC Posted: October 03, 2003 at 07:55 PM (#514988)

Marc,

You would adjust it the same way Win Shares would be adjusted. WS has an even lower baseline. James just has the sense to not call it replacement level though. See the numerous posts discussing the WS adjustments. It looks like for a full time player, an adjustment to what some consider a more reasonable replacement level would be about 2 WARP. For less than full time, the adjustment would be proportionately less.

Personally, I happen to like a replacement level that scores the worst team of all time at ~0. It just seems right to me. YMMV.

Realistically, there is no such thing as "replacement level hitting" and "replacement level fielding". Replacement level is a total sum of the players value concept. I would prefer if BP gave us numbers that are Hitting above average postion, Fielding above average, and a replacement value for amount of playing time. Splitting the replacement value into hitting and fielding portions makes no sense (to me).

34. Marc Posted: October 03, 2003 at 08:06 PM (#514989)

RobC, thanks. And not to sound too stupid here, but if the replacement level is too low (?) and we think it should be higher, then we would adjust a player's WARP score downward, right? Or did you say the replacement is too high? Then we would adjust the WARP upward. As you can see, I am still confused.

Also, there is a mention that middle infielders might be overrated and others not. This suggests that we cannot necessarily just apply a 2 WARP adjust across the board???

I've just sort of gotten comfortable that I understand WARP so if it is not reliable now is the time to find out!

35. Tangotiger Posted: October 03, 2003 at 09:20 PM (#514990)

As a general rule of thumb, subtract 2 wins / 162 GP for all position players, IF you think that a replacement level should be -2 wins relative to average.

What does -2 wins relative to average imply? This would make a team of all replacement level players worth -36 wins relative to average (or 45 wins). That is, 9 hitters x (-2) + 9 pitchers x (-2) = -36. If you believe a team of all replacement players would win 45 games, then subtract 2 wins / 162 GP.

36. Tangotiger Posted: October 03, 2003 at 09:30 PM (#514991)

To clarify, WARP sets the replacement level at essentially -4 wins for position players, and -2 wins for pitchers (that by itself should bother you), relative to average.

This works out to having a replacement team of players winning 27 games.

If you are fine with this, then use WARP as-is. Otherwise, WARP overstates hitters by 2 wins / 162.

37. Marc Posted: October 03, 2003 at 10:21 PM (#514992)

Thanks Tango. 2003 was a great laboratory wasn't it? Somebody mentioned the Detroit Tigers above. Your option A (#65) is a team that is 2 wins better than the Tigers. It seems to me that the Tigers were basically a team of replacement players with 3-4 substantive exceptions--thus (what?) 5-7 games better than replacement? I realize this is a very gross method, but that suggests pure replacement at somewhere between 27 and 45. This is not to say 45 is wrong, but rather to say "it's in the ballpark." Is this logical? Are the Tigers a bunch of lab rats?

38. Tangotiger Posted: October 04, 2003 at 03:08 AM (#514993)

A bunch of "replacements" would probably be closer to say what an expansion team would give you, and not necessarily what the worst team you can probably have. Inside those Tigers is some good players, but also some players that no team would ever play, and the only reason they are playing is that the Tigers' mismanagement has no other options.

I think if you were to look at what say all expansion teams 1st year records were from 1961 to 1977, you'll probably find they averaged (and median) about 55 to 60 wins.

A replacement-level player is a player that a team that has average resources and average options would have this player as their bench player. This does not apply to the 03 Tigers and 62 Mets. There are probably a dozen players on these teams that other teams would never want at that point.

But, like I said, if you believe that 27 wins is what you want, that's fine. My warning is more to those who believe that replacement level is 45 wins, and think that's what they are getting with WARP.

39. Tangotiger Posted: October 04, 2003 at 01:08 PM (#514995)

You can't take a metric and fudge the numbers for the hitters so that they come out better than pitchers.

You have to start with a premise and let the metric carry you there.

You can very well make the case that the replacement level setting for pitchers should be 1.8 wins below average and for hitters it should be 2.7 win below average. But to say -2 and -4? I mean, just look at that Duane Kuiper 78 season, or that Tim Wallach 93 season.

Or look at that Oquendo season. 9 wins? Isn't that higher than Halladay and Hudsono and Prior this year?

40. Tangotiger Posted: October 04, 2003 at 01:34 PM (#514996)

Babe Ruth is credited with +376 fielding runs above replacement, when in my view, he should get +79 runs above average. He has +297 runs too much, or about +30 too many wins. His WARP1 is +242, and his Tango-WARP1 is about +212.

Ted Williams? WARP1 = +170, T_WARP = +142.

Walter Johnson? WARP1 = +205.

Willie Mays? WARP1 = +199, and T_WARP = +155.

41. Tangotiger Posted: October 04, 2003 at 02:41 PM (#514997)

42. tangotiger Posted: October 08, 2003 at 03:55 PM (#514999)

Joe, if you will be revising, I would do as I described, and take the difference between FRAA and FRAR and divide by 10. That's the "overcompensation" factor. The "2 wins / season" is a generic rule of thumb.

43. jimd Posted: October 09, 2003 at 03:30 PM (#515002)

Should I divide by 10 or by league R/G?

It wouldn't be league R/G because all of their runs (pitching, fielding, batting) have been normalized to an idealized 4.5 R/G league. Whether the constant is 10 or something else close to that, I'll leave that to those who have studied it more closely.

44. RobC Posted: October 09, 2003 at 09:00 PM (#515003)

Joe,

Based on the 4.5 R/G in jimd's post, you should divide by 9.

However, here is an interesting thought I had concerning the "double counting" of replacement. Hitters is repl hitting + repl fielding. Pitchers is repl pitching + repl hitting. Thus, pitchers are getting a "double counting" of replacement too.

Also, I believe you adjust WS by 6 or 6.5 per season as a replacement level and that was too even out with WARP. If you are now adjusting WARP by 2 wins per season, you would need to increase the WS adjustment.

45. RobC Posted: October 10, 2003 at 01:55 PM (#515005)

Joe,

There was a discussion at one time about WS replacement level and calculations were done to equate WS with WARP with regards to replacement level. Now, you are adjusting WARP to a "more reasonable" replacement level (this isnt just a fielding adjustment, it could have been applied to the hitting while leaving fielding alone - exact same effect). If, you chose the WS replacement level to match with WARP, then you need to readjust WS to match the "more reasonable" level.

WS unadjusted replacement level = 0

46. tangotiger Posted: October 10, 2003 at 10:43 PM (#515006)

If you use the .29 exponent in PythagaPat, the runs-per-win (RPW) converter for a 4.5 RPG league is 9.52. If you use .28, it's 9.73.

It's easy enough to figure, as you just figure the win% for a 4.51 RS and 4.49 RA, and work from there.

If you use the James' Pythag, the RPW is 9. James' Pythag is not accurate enough for these purposes.

When I look at empirical data, my RPW converter is 9.7 for a 4.5 RPG team.

47. Marc Posted: October 10, 2003 at 11:26 PM (#515007)

Joe, I've now figured out WARP1-2-3, kinda like Lotus, right? But help me again with adjWARP1 and adjWARP3. What exactly are the adj? Thx.

Also I assume the adjWS is just to season length and including NA but not pre-NA??

48. ronw Posted: October 28, 2003 at 07:31 PM (#515010)

Maybe I missed it, but do you have Kip Selbach's WARP1 pennants added?

49. ronw Posted: October 28, 2003 at 07:31 PM (#515011)

Sorry, Selbach's WARP3 pennants added, I see WARP1.

50. tangotiger Posted: October 31, 2003 at 03:56 PM (#515013)

Clay addressed the issue of replacement level on his Chat. I posted my thoughts on the matter at the above link.

51. ronw Posted: November 14, 2003 at 01:20 AM (#515014)

For some 1890's pitching comparison, can we get WARP1 PA numbers for Ted Breitenstein and Pink Hawley? They will not get a vote from me, but their numbers might help the case for Griffith and McGinnity.

52. jimd Posted: April 28, 2004 at 01:53 PM (#515018)

I want to correct for the WARP overvaluing of defense

I think the issue you're talking about is WARP having too low a replacement level (though it's still better than Win Shares in that regard). This is because replacement level hitters have something to offer defensively (e.g. Rey Ordonez), and replacement level fielders have something to offer offensively (e.g. Mo Vaughn). Nobody is replacement level at both (and lasts long). So the overall impact is spurious value - which is accumulated proportional to playing time - because the the real replacement level is higher than either of the component replacement levels. (I have not conducted any study to verify that this is a real problem; this is just my understanding of the alleged problem as previously described.)

Tango's proposed remedy was to use FRAA. The problem with this is that it penalizes glove positions (SS, C, etc.) far more than it does bat positions (1B, LF, etc.). They accumulate more fielding runs over replacment in a given amount of playing time so they are losing more when FRAA is substituted. Mo would lose about 10 runs in a full season, Rey would lose about 20. The truth probably lies somewhere in between.

Unfortunately, I have no easy fix for that. Pre-WWII, batting time probably approximates playing time pretty well (wild guess on my part). Defensive substitutions will gum up those works; maybe the modern custom of retiring instead of accepting a reduced role will minimize the damage there for most HOM candidates.

53. jimd Posted: April 28, 2004 at 09:08 PM (#515021)

WS has a built in replacement level of about .200.

The .200 level is only an offensive replacement level.

54. jimd Posted: May 20, 2004 at 05:00 PM (#637126)

WS has a built in replacement level of about .200.

The .200 level is only an offensive replacement level.
A team at the offensive zero-point (RS = .5*LgRS) with an average defense will be a .200 team.
A team at the defensive zero-point (RA = 1.5*LgRS) with an average offense will be a .308 team.
The radically different replacement levels IMO result in offense getting overrated when compared to defense.

The 50-50 split in WS is elegant and esthetically pleasing. But that doesn't mean that it's correct. There's no underlying theoretical reason that I can see for the hitting replacement level to be at .200 and defensive replacement level to be at .300. Some of Tango's old WS discussions talked about a 40-60 split (it's 61 or 62 for defense, I can't remember); if that's true then offensive WS are inflated by about 20% at the expense of pitching/fielding.

It's obvious that WARPs mistakes tend to overvalue the tough defensive positions, I believe that's why guy like McPhee, Wallace and Bennett do too well. So I don't think it's an issue to correct for them more.

I don't believe that's true; you're expecting WARP to echo the results from Win Shares.

However you're neglecting to take into account that the game before 1920 was much more of an infield game than the modern game has become after Cravath and Ruth revolutionized hitting with the uppercut swing. Infielders and catchers were making many more plays back then; it stands to reason that they might be somewhat more important defensively relative to OF'ers than in the modern game. Firstbase is a little more important too due to the extra catches required and the lack of a basket glove to make those catches routine. WS has no mechanism to adapt to these kinds of changes in the game, hence WS overrates OF'ers from this era, seeing the responsibilities of each position as constant through time, which account in part for the OF glut. (Overrating offense helps create it too.)

Compared to Win Shares, WARP is a different opinion; which is closer to the "truth" is a matter for debate (see above on replacement level).

WARP ... gives about 2 extra wins a season too, though it's a little more for the glove men.

You may be right, but my gut feeling is that WARP's correction is not that high, or alternatively that the WS correction is too low. My Win Shares 19thC rankings was cluttered with long-career low-value journeymen like Elmer Smith and Tom Brown who did little but hang around just above WARP replacment level but accumulated Win Shares (over 300 season-adjusted apiece) but few HOM votes. They didn't make the WARP rankings either, I'm guessing because of the replacement level difference.

I don't think the defensive sub issue is a problem,

It isn't yet. It becomes one if we abandon the idea of using FRAA and instead begin trying to create an "overall replacement level penalty" that is proportional to playing time.

55. DavidFoss Posted: May 20, 2004 at 09:19 PM (#637900)

Short side question... I'm having trouble getting BP's numbers to add up.

Is WARP = BRARP + FRAR + PRAR or BRAR + FRAR + PRAR or something else?

Is there a link detailing their calcuations?

Thanks...

56. jimd Posted: May 20, 2004 at 09:50 PM (#637964)

Is WARP = BRARP + FRAR + PRAR or BRAR + FRAR + PRAR or something else?

That's what Joe and I were debating (before the Primer site switch).

Is there a link detailing their calcuations?

There used to be a link to sketchy documentation if you clicked on the column headings. It didn't tell all the details but it sometimes gave an overview, and at least usually gave an indication of what the stat was useful for. Last time I checked that had broken during Prospectus' site switch.

57. DavidFoss Posted: May 20, 2004 at 10:20 PM (#638018)

Well, I found this by guessing at the broken links:

http://www.baseballprospectus.com/dt/glossary.shtml

But I think that's just the sketchy stuff you were talking about.

BRARP + FRAR + PRAR looks like a double positional adjustment if you ask me... but that may be what they are doing. BRAR + FRAR would make more sense but thats also a double-replacement level. But, I would pick something like BRARP + FRAA for a single positional adjustment and single replacement.

Needless to say, you could find a flaw any way you slice it. Trying to boil everything down to one number will do that.

58. sunnyday2 Posted: May 20, 2004 at 10:40 PM (#638046)

If i recall, it was neft who did a great article way back in the 1983 (?) BRJ ranking SSs, back when ranking on a numerical basis was new and different. The method was:

an offensive number like an EqA and maybe it was EqA, something revolving around .300

Then a defensive number revolving around zero but also expressed as a .000 number

And a baserunning + or - number revolving around .000

then just add 'em up. The interesting thing was that at a glance it looked like the offensive number was predominant because it was this big .2XX-.3XX number, and guess what, the player's final total rating was also somewhere in the .2XX-.3XX-maybe .4XX range, while those other numbers were all in the .0XX or even .00X range.

But on closer examination, the fact is that the defensive numbers ranged of -.050 to +.099 or a range of .150. The offensive numbers ranged around .300 but only from about .225 to .350, or a range of about .125. So in the end the defense was slightly more determinative (if that's a word) of the final rating and ranking.

Now back in '83 I thought that was funny, but for SSs I've come to believe it is right. Players get about 700 PAs in a year while SSs make as many as a 1000 plays (not so much anymore, granted). what is important of course is not the raw number of plays, but the variance. Take an offensive EqA against, what? As always, defense is hard to peg, with FAs pretty much useless. if you used FAs, of course, the range would be smaller than for an offense EqA. But factor in range, not just errors, and i would guess that the variation is greeater on defense, and therefore the defensive ratings are relatively more important toward the value of a SS.

this is perhaps more true then than now. But taking history as a whole, I think it is true. I haven't entirely put this theory to work, because if I had, Joe tinker would probably be on my ballot.

(I am on a borrowed computer on which I can log-in and it's shift key has a different feel to it than what I am used to. sorry.)

As for WARP, handle with care. And WS, handle with care. I even use TPR, but handle with extreme care.

59. JoeD has the Imperial March Stuck in His Head Posted: May 21, 2004 at 07:58 AM (#639002)

For WARP, the base for it will be BRARP+FRAA+PRAR.

For WS, I'll adjust offense for the .200 replacement level if I can figure out a way. I like the .308 for pitching, so I'm fine with a .308 overall replacement level, it might be a little low, but if it's easier that could work.

Thoughts? I'll probably be setting up the spreadsheet this weekend - pending house-hunting - then it's a question of populating it with data . . .

60. EricC Posted: May 21, 2004 at 11:44 AM (#639022)

Joe, What is the formula for pennants added?

61. JoeD has the Imperial March Stuck in His Head Posted: May 25, 2004 at 08:39 AM (#643101)

Eric - just saw your note, but I don't have it handy (it's not easily memorizable). I won't be home until Wednesday morning, when I get back I'll write it in for you.

If you have a 2002 (I think) Baseball Prospectus, the article introducing the method has the formula.

62. TomH Posted: February 14, 2005 at 07:57 PM (#1145131)

not sure of the best thread to put this (could be in the WS vs WARP battle), but here goes:

Pennants Added, or any other metric, depends heavily on the setting of Replacement Level

Some use 'average'. Some use .300 OWP, or the Win Shares or WARP level, or WS minus X or WARP minus Y. We all agree that there is no 'right' answer, but in every case, we are using a linear scale above our indiviudal placement of 'replacement'.

I've come to believe that the scale should NOT be linear, at least at the bottom (below average), but rather should be a square-like function (curve bending upward).

Why? Well, we mostly agree that below-average play has SOME value. But it does not, one-for-one, have as much valie as above-average play, since in some circumstances the player could have been replaced by a better one, or platooned, or confined to part-time, pinch-hitting, -running, or playing defense. If he was a pithcer, he could have been moved into slop relief, where the payback (wins) per inning is less critical.

This is easily seen in sim games, where there is the quandry of drafting a 6th starting pitcher or 4th OFer of marginal quality (for injury protection), or picking up a 50 inning reliever or fragile hitter (Griffey). There still is value in a 4.50 ERA, but the difference is smaller from 4.5 to 5.0 than it was from 4.0 to 4.5; in other words, we would all take a one 4.0 guy and one 5.0 guy over two 4.5s, right?

If I were to generate an uber-stat "Hanrahan's pennants added", I would set up a very low (Win-Share-like) bottom replacement level, but give much less than full-run (or -win) credit for the difference bewteen 'average' play and
'replacement level' play; ramping up to full credit for all play above average.

63. DavidFoss Posted: February 14, 2005 at 08:39 PM (#1145226)

TomH,

Check the newer "2.0" version of this thread... Pennants Added 2.0

There is a lengthy discussion about replacement level, there.

64. TomH Posted: June 21, 2005 at 02:21 PM (#1419743)

An article was published in SABR's By The Numbers this week attempting to measure Pennants Added. You can see the article (pdf) at www.philbirnbaum.com, Feb 2005 issue, page 15. Dan Levitt is the author.
His conclusions were that
- two "5 wins above average (WAA)" seasons are worth 20% more in Pennants than five "two WAA" seasons.
- one "10 WAA" is worth 19% more pennants than two "5 WAAs".

65. Paul Wendt Posted: August 21, 2007 at 02:11 AM (#2493008)

Today Joe Dimino posted lots of Pennants Added data in the Dan Rosenheck thread. (There he derived Pennants Added from DanR's wins above replacement.)

Pennants Added : How does the thought-experiment work?

Does the player under assessment go to the 1896 Baltimore Orioles with probability approx. 1/220 (uniformly for all major league seasons from 1893) * 1/12 (uniformly to all teams in that league)?
If so then probability for 1996 Baltimore Orioles is approx 1/220 * 1/14, probability for 1956 Baltimore Orioles is approx 1/220 * 1/8, and so on. For every team-season, approximately 1/220 * 1/[size of league].

Or does the player go uniformly to a year? or to a club-season?

Then what? The 1896 Baltimore Orioles win the pennant with our player, but they won it without him ==> 0 pennants added. The 1996 Orioles win the division with any player worth > 4.0 wins above replacement, tie with a player worth 4.0, and lose with a player worth < 4.0 warp. ==> 1, 1/2, or 0 pennants added depending on player value. Is that the method?

66. Jim Sp Posted: September 30, 2007 at 07:49 PM (#2551856)

Joe,
Can you please subtract one from Tom Glavine's pennants added?

Thanks.

67. David Concepcion de la Desviacion Estandar (Dan R) Posted: September 30, 2007 at 10:19 PM (#2552314)

Hahah. I was there. It was atrocious. Absolutely embarrassing. Then again, the Mets did strand 8 baserunners in the first three innings. It's hard to only blame Glavine when the Mets only scored one run.

Pathetic.

68. JoeD has the Imperial March Stuck in His Head Posted: October 01, 2007 at 04:20 PM (#2552969)

Yeah really. Glavine wasn't even hit all that hard from what I could see, it was just a nightmare (for him) that spiraled out of control.

69. Dr. Chaleeko Posted: October 01, 2007 at 05:05 PM (#2553056)

Glavine wasn't even hit all that hard from what I could see, it was just a nightmare (for him) that spiraled out of control.

Payback for the NLCS last year when Glavine pitched poorly but didn't break. The game where Pujols said something about how the Card should have smacked him around, and everyone jumped Albert's sh*t for criticizing the veteran.

70. Paul Wendt Posted: September 28, 2008 at 01:46 PM (#2959134)

Jim Sp,
Do you subtract one of Pedro's pennants added or wait until Sunday/Monday night when the Mets are in or out of the playoff?

--

Today they will close down Shea Stadium in the last game of the season. The Mets are tied with the Brewers so I imagine it will be a bigger deal than the Yankees last weekend. It's all over the 'net.

Well, there is this.
>>
MLB.com replays
Shea's greatest moments
MLB.com offers Shea Stadium's top moments through Sunday's pregame show at noon. Watch >
<<

I doubt they have anything on Tommy Agee.

71. Paul Wendt Posted: September 28, 2008 at 01:46 PM (#2959135)

The preface above provides some Pennants Added statistics based on
: WARP3
: WARP1
: Win Shares.
Do we have a source with greater scope?

In "Dan Rosenheck's WARP data" (#336-350, Joe Dimino and others)
there is some discussion of Pennants Added based on
: DanR's WARP.

IIUC the Pennants Added statistics here in Hall of Merit tables and prose are based on differing measures of player-season value and differing timespans 1871 to yyyy. Within-season variation in team W-L records is generally decreasing 1871 to 2008 --variation in year yyyy decreases as yyyy decreases--
so the same must be true of within-timespan variation --variation in span 1871 to yyyy decreases as yyyy decreases.
IIUC that is more or less important to Pennants Added depending on the adjustments for standard deviation that are built into player-season value.

72. JoeD has the Imperial March Stuck in His Head Posted: October 09, 2008 at 05:04 AM (#2975029)

Paul I've calculated Pennants Added for every player season in DanR's WARP spreadsheet. I've actually put it into a MS Access Database. I will upload it tomorrow, I'm on my Mac right now, and don't have access (no pun intended) to the file right now . . .

73. JoeD has the Imperial March Stuck in His Head Posted: October 09, 2008 at 10:45 PM (#2975530)

Paul the file is 13 MB zipped, so I can't upload it to the site. And gmail can't accept files over 10 MB. Do you have an FTP site I can use or something?

74. Paul Wendt Posted: October 11, 2008 at 02:08 AM (#2976768)

YouSendIt.com provides free upload and download, limited to 7 days and 100MB, iirc. That is their loss leader. For me you send notice to pgw02472@yahoo.com although you may send to everyone on the HallofMerit elist.

You must be Registered and Logged In to post comments.

<< Back to main

Tuesday, July 01, 2003

Pennants Added (updated for 1914 ballot)

Reader Comments and Retorts

BBTF Partner

Support BBTF

Bookmarks

Hot Topics

Syndicate