Do long extra-inning games wear teams out?

On August 9 and 10 of 2014 nineteen-inning games were played. This renewed memories of the Pirates dismal 12-29 record after their 19 inning victory at St. Louis August 19, 2012 and caused some speculation that the Red Sox, Angels, Tigers, and Jays might suffer from the load to their relief corps and fatigue. I've been meaning to take a look at this for a while so here's what I have so far. (the summary is that I found no fatigue effect) I looked at the Retrosheet games 1914-2013 and took only games in which the teams had at least 10 games played before and after the extra inning game. Then I found the team's W-L pct that year, the team's record in the 10 games before, and the team's record in the 10 games after.

For all 27543 such extra inning games the season average percentage was .500
the W-L before was 137334-137210 .500 [there were 886 ties in the 275430 games]
the W-L after was 137247-137347 .500 [836 ties]

When I looked at exceptionally long games (143 game longer than 18 innings)
I found season .500 before .507 after .489
The before was 713-693 the after was 697-728 so they differ by only 0.96 standard deviations, not statistically significant.

For moderately long games [14-18 innings] the difference was smaller, 12526-12512-192 .500 before and 12518-12625-87 .498 after.

For games of each length 10 through 22 innings

Inn #G season before after
10 12154 .500 .499 .499
11 06841 .500 .500 .501
12 03752 .500 .500 .499
13 02130 .502 .502 .502
14 01232 .503 .501 .499
15 00619 .497 .504 .491
16 00366 .502 .502 .499
17 00182 .504 .501 .513
18 00124 .495 .473 .497
19 00066 .517 .518 .486
20 00030 .495 .502 .497
21 00018 .489 .489 .492
22 00018 .475 .480 .475
Some #G totals are odd because one of the teams had fewer than 10 games before or after the extra inning game.

I also broke these down into winning and losing teams. The team quality and before-and-after breakdowns are

Wins 13558 .510    68508-66653-419 .507     68526-66652-402 .507
Loss 13552 .491    66654-68452-414 .493     66621-68512-387 .493
Ties 00433 .499    02172-02105-053 .508     02100-02183-047 .490

For the games lasting more than 18 innings the breakdown is 
Wins 65 .519    318-329 .491 326-322 .503  (t-score 0.42)
Loss 65 .480    332-312 .516 305-342 .471  (t-score 1.62)
Ties 13 .510    066-062 .516 066-064 .508

Winning teams, before and after games lasting 10 innings through 20 innings
The t-score is the number of standard errors that the team dropped after the long game. (if there is no difference then roughly 1 in 20 t-scores would be bigger than 2)

Inn #G season     before           after
10 6010 .508  30371-29530 .507 30366-29570 .507 (t-> 0.13)
11 3381 .510  17022-16699 .505 17176-16522 .510 (t-> -1.28)
12 1856 .508  09471-09033 .512 09342-09160 .505 (t-> 1.33)
13 1037 .511  05225-05216 .505 05232-05098 .506 (t-> -0.87)
14 0595 .514  03004-02927 .506 03015-02926 .507 (t-> -0.11)
15 0298 .506  01499-01467 .505 01455-01514 .490 (t-> 1.18)
16 0171 .506  00864-00843 .506 00831-00875 .487 (t-> 1.11)
17 0086 .519  00444-00415 .517 00479-00380 .558 (t-> -1.69)
18 0059 .503  00290-00298 .493 00304-00285 .516 (t-> -0.79)
19 0029 .531  00145-00143 .503 00141-00149 .486 (t-> 0.42)
20 0015 .505  00069-00081 .460 00077-00073 .513 (t-> -0.93)
Losing teams, before and after games lasting 10 innings through 20 innings
Inn #G season     before           after
10 6012 .489  29599-30343 .494 29531-30424 .493 (t-> 0.43)
11 3366 .491  16510-17065 .492 16558-16996 .493 (t-> -0.45)
12 1849 .491  08981-09448 .487 09073-09369 .492 (t-> -0.89)
13 1050 .492  05244-05216 .501 05207-05271 .497 (t-> 0.64)
14 0595 .492  02958-02977 .498 02923-03008 .493 (t-> 0.61)
15 0297 .490  01496-01463 .506 01477-01481 .499 (t-> 0.48)
16 0171 .494  00842-00854 .496 00859-00844 .504 (t-> -0.46)
17 0088 .489  00433-00445 .493 00408-00468 .466 (t-> 1.15)
18 0059 .485  00259-00329 .440 00280-00309 .475 (t-> -1.20)
19 0029 .503  00152-00135 .530 00140-00148 .486 (t-> 1.04)
20 0015 .486  00082-00067 .550 00072-00078 .480 (t-> 1.22)

counting the number of t-scores for each of the ranges (-2,-1.501),(-1.5,-1.001),(-1,-.501),(-.5,-.001),(0,.499),(.5,.999),(1,1.499),(1.5,1.999) gives

-2.0 W
-1.5 WL
-1.0 WWWL
-0.5 WLL
+0.0 WWLL
+0.5 LL
+1.0 WWWLLL
+1.5 

The distribution does not look very different from what a no-effect random set of data would look like so if there is an effect it seems that it's small enough that more than a few thousand games are required to tease it out.

For games of at least 19 innings the results in the 143 games before and after. i.e. in the game before the extra-inning game (-01) the 143 teams won 69 games, lost 70, and tied 4. In the games immediately following the extra inning game (+01) the 143 teams won 68 and lost 75.

-10  68-75-0 .476
-09  69-74-0 .483
-08  81-61-1 .570
-07  72-70-1 .507
-06  68-74-1 .479
-05  69-73-1 .486
-04  75-68-0 .524
-03  78-63-2 .553
-02  67-75-1 .472
-01  69-70-4 .496
Long extra-inning game
+01  68-75-0 .476
+02  69-74-0 .483
+03  81-61-1 .570
+04  72-70-1 .507
+05  68-74-1 .479
+06  69-73-1 .486
+07  75-68-0 .524
+08  78-63-2 .553
+09  67-75-1 .472
+10  69-70-4 .496
With no effect we should see 2/3 of the results within -12 to +12 games over .500 and 95% within 24 games of .500. the actual tallies are 16 of 20 and 20 of 20.

In more recent seasons it has become more common to use several relief pitchers during the first 9 innings, meaning that bullpen might get more stressed in long games.
Restricting the data to the 20 year period 1994-2013 gives

Games of at least 19 innings:
20 teams expected W-L .509   Before 93-107 .465  After 90-100 .450
Winning teams 10, .523    Before 37-63 .370  After 47-53 .470
Losing teams  10, .495    Before 56-44 .560  After 43-57 .430

Games 14-18 innings
543 teams expected W-L .501  Before 2677-2753 .493  After 2726-2703-1 .502
Winning teams 271, .505      Before 1339-1371 .494  After 1360-1350 .502
Losing teams  272, .498      Before 1338-1382 .492  After 1366-1353-1 .502

For games of each length 10 through 22 innings

Inn #G season before after
10 3300 .500 .500 .496
11 1676 .501 .497 .498
12 0928 .500 .493 .502
13 0513 .503 .511 .498
14 0294 .504 .501 .509
15 0124 .498 .481 .494
16 0060 .501 .470 .497
17 0033 .498 .506 .503
18 0032 .498 .497 .481
19 0012 .532 .450 .433
20 0006 .491 .483 .567
21 0
22 0002 .423 .500 .200

The team quality and before-and-after breakdowns are

Wins 3492 .508    17600-17319-1 .504    17545-17366-9 .503
Loss 3486 .493    17193-17665-2 .493    17187-17664-9 .493
Ties 0002 .611    00014-00006-0 .700    00010-00010-0 .500

For the games lasting more than 18 innings the breakdown is 
Wins 10 .523    37-63 .370   47-53 .470 (t-score 1.44)
Loss 10 .495    56-44 .560   43-57 .430  (t-score 1.85)
Ties 0

Winning teams, before and after games lasting 10 innings through 20 innings The t-score is the number of standard errors that the team dropped after the long game. (if there is no difference then roughly 1 in 20 t-scores would be bigger than 2)

10 1650 .507   8353-8146 .506   8225-8271 .499 (t-> 1.39)
11 0840 .509   4201-4199 .500   4275-4124 .509 (t-> -1.15)
12 0466 .508   2364-2296 .507   2386-2272 .512 (t-> -0.48)
13 0255 .510   1306-1244 .512   1252-1296 .491 (t-> 1.48)
14 0147 .507   0750-0720 .510   0743-0727 .505 (t-> 0.26)
15 0062 .504   0277-0343 .447   0298-0322 .481 (t-> -1.20)
16 0030 .502   0144-0156 .480   0149-0151 .497 (t-> -0.41)
17 0016 .511   0085-0075 .531   0082-0078 .513 (t-> 0.34)
18 0016 .497   0083-0077 .519   0088-0072 .550 (t-> -0.56)
19 0006 .563   0023-0037 .383   0025-0035 .417 (t-> -0.37)
20 0003 .465   0009-0021 .300   0020-0010 .667 (t-> -3.05)

Losing teams, before and after games lasting 10 innings through 20 innings

10 1648 .492   8140-8339 .494   8127-8349 .493 (t-> 0.13)
11 0836 .493   4129-4230 .494   4073-4286 .487 (t-> 0.87)
12 0462 .492   2214-2406 .479   2274-2343 .493 (t-> -1.28)
13 0258 .496   1316-1264 .510   1304-1276 .505 (t-> 0.33)
14 0147 .501   0722-0748 .491   0752-0718 .512 (t-> -1.11)
15 0062 .492   0320-0300 .516   0315-0305 .508 (t-> 0.28)
16 0030 .499   0138-0162 .460   0149-0150 .498 (t-> -0.94)
17 0017 .485   0082-0088 .482   0084-0086 .494 (t-> -0.22)
18 0016 .498   0076-0084 .475   0066-0094 .413 (t-> 1.13)
19 0006 .502   0031-0029 .517   0027-0033 .450 (t-> 0.73)
20 0003 .516   0020-0010 .667   0014-0016 .467 (t-> 1.60)

For games of at least 19 innings the results in the 20 games before and after

-10  10-10 .500
-09  10-10 .500
-08  12-08 .600
-07  10-10 .500
-06  07-13 .350
-05  08-12 .400
-04  09-11 .450
-03  09-11 .450
-02  08-12 .400
-01  10-10 .500
Long extra-inning game
+01  10-10 .500
+02  10-10 .500
+03  12-08 .600
+04  10-10 .500
+05  07-13 .350
+06  08-12 .400
+07  09-11 .450
+08  09-11 .450
+09  08-12 .400
+10  10-10 .500

After playing 19 innings August 9 the Red Sox won 4 in a row. Before the game the Red Sox were 3-7. Since the game they are 5-5.
The Angels were 4-6 before the game and 8-2 after.
After playing 19 innings August 10 the Tigers lost two more. Before the game the Tigers were 5-5. Since the game, 5-5.
The Jays were 4-6 in the ten games before the marathon. After game they lost 4 in a row and have a record of 2-7.
The ten games win-loss strings before and after for these teams are
Bos LLLWLLLWLW, Lost 5-4 on road, WWWWLWLLLL
LAA LWWLWWLLLL, Won 5-4 at home, LWWWWLWWWW
Det LWWWLWLLWL, Lost 6-5 on road, LLWWLWLWWL
Tor WWLLLLWLLW, Won 6-5 at home, LLLLWLLWLW

The fours teams were 16-24 .400 before their 19 inning games and so far are 21-19 .525 after.
What are some other ways to look at the data? e-mail.
http://www.rose-hulman.edu/~rickert/BB/xinnfatigue.html
My baseball page