Wednesday, July 12, 2017

The 2005 White Sox Went 14-5 Against The Indians Despite A -.033 OPS Differential In Those Games (and the Indians had an overall OPS differential of .098 while the Sox had .040)

All data from Baseball Reference and the BR Play Index. My formula for estimating winning pct based on OPS differential is

Pct = .5 + 1.33*OPSDIFF

That is based on a regression. So it estimates the Indians to have a .631 pct or 102 wins while the Sox would have .553 and 90 wins. Yet the Sox actually went 99-63 and the Indians went 93-69.

In the 19 games, the Sox had a .667 OPS and the Indians had .700. There were nine 1-run games and the Sox won them all. The White Sox only outscored the Indians 80-75 in their games.

These stats slightly over state the improbability, but only slightly. The Sox had a 3 game lead with a three game left in Cleveland. But given that at that point the Sox led the season series 11-5, they were assured of having the tie-breaker on their side if the two teams had an identical record since the worst the Sox could do against the Indians would be 11-8, thus taking the season series.

The Sox won all of those last three of those games, two by 1 run and one by 2 runs. But that still means that the Sox had been 7-0 in all the 1-run games between the teams before that. They had only outscored the Indians 70-69 in their first 16 games. And still went 11-5.

The Sox had a 15 game lead at the close of play on Aug. 1. By Sept. 22, that had fallen to 1.5 games. It was still 1.5 on Sept. 24 with 8 games left.

The Sox lost 2 of 3 to the Indians over Sept. 19-21 in Chicago, winning only the middle game 7-6 in 10 innings. By winning that one, they got their lead up to 3.5 games. A loss would have meant a 1.5 game lead (which it still fell to anyway two days later).

The Indians outhit the Sox 14-11 in that game, out HRed the Sox 3-2 and out walked them 5-4. The Indians committed 2 errors and had 1 unearned run.

For the whole season, The Sox had the following OPS differentials in High, Medium and Low leverage situations: .102, .018, .034. The Indians had .044, .098, .123.  I do have a regression equation that uses the OPS differentials in the High, Medium and Low cases. Here it is

Pct = .5 + .306*LOW +.420*MED + .564*HIGH

That estimates the Sox to win 93.2 games and the Indians 97.8. Not as big as 12 mentioned at the beginning based on total OPS, but still fairly large. Maybe the Sox had a much higher OPS in High leverage cases in the games between the two teams.

In games between the two teams, here is the Sox OPS in High, Medium and Low leverage situations: .798, .661, .599. For the Indians they were .584, .739, .748. So here are the Sox OPS differentials in their games with the Indians in High, Medium and Low leverage situations: .214, -.078, -.149.

In games between the two teams, the Sox had an OPS of .729 with runners on and the Indians had .651. From the 7th inning on, with the score tied or either team ahead by 1 run, the Sox had an OPS of .784 and the Indians had .705.

Here are the scores of all the games between the two teams that year and the OPS each team had. It shows the OPS each team had. The Diff column is the Sox OPS minus the Indians OPS. The Indians had a higher OPS in 11 of the games, yet won only 5. The Indians had a higher OPS in 8 of the first 16 games and won only 5 of those.


Date Loc Sox R Ind R Inn Sox OPS Ind OPS Diff
 Apr 4 CHI 1 0
0.351 0.181 0.170
 Apr 6 CHI 4 3
0.655 0.454 0.201
 Apr 7 CHI 5 11 11 0.648 0.653 -0.005
 Apr 11 CLE 2 1
0.654 0.618 0.036
 Apr 13 CLE 5 4 10 0.644 0.616 0.028
 Apr 14 CLE 6 8
0.664 0.660 0.004
 Jun 3 CHI 6 4
0.727 0.704 0.023
 Jun 4 CHI 6 5
0.731 0.707 0.024
 Jun 5 CHI 4 6 12 0.729 0.708 0.021
 Jul 14 CLE 1 0
0.740 0.753 -0.013
 Jul 15 CLE 7 1
0.739 0.751 -0.012
 Jul 16 CLE 7 5
0.739 0.748 -0.009
 Jul 17 CLE 4 0
0.739 0.747 -0.008
 Sep 19 CHI 5 7
0.749 0.782 -0.033
 Sep 20 CHI 7 6 10 0.750 0.784 -0.034
 Sep 21 CHI 0 8
0.748 0.786 -0.038
 Sep 30 CLE 3 2 13 0.748 0.790 -0.042
 Oct 1 CLE 4 3
0.748 0.788 -0.040
 Oct 2 CLE 3 1
0.747 0.787 -0.040

Wednesday, July 5, 2017

Team OPS Differentials And Expected Wins At The Halfway Point (Or Who Is Leveraging Wins)

This is through games Monday, July 3. Some teams are doing better than expected and some worse. But it will turn out that the biggest outliers see their expected win differences shrink quite a bit when High, Medium and Low leverage situations are taken into account. To predict or estimate a team's winning pct I use

Pct = 1.3465*OPSDIFF + .5

That is based on a regression done on all teams from 2010-2014. I used team averages over the period for both pct and OPSDIFF.

The table below shows each team's OPS so far this season along with their OPS allowed (OPSA), their actual winning pct and then the differential in their win total from what the formula would predict.

The Orioles have a large negative OPS differential yet are still close to a .500 team at .488. So they have won about 9 more games than expected. I will have some discussion below about the Orioles as well as the Twins and Yankees (who have won about 8 fewer games than expected). Also, at the end of this post is a table with each team's actual win and loss totals through Monday.


TEAM OPS OPSA Pct W Diff
Baltimore 0.730 0.821 0.488 9.05
Minnesota 0.740 0.797 0.512 7.29
Colorado 0.755 0.756 0.576 6.61
LA Angels 0.697 0.740 0.494 4.54
Kansas City 0.716 0.734 0.512 2.99
San Diego 0.682 0.764 0.415 2.05
Milwaukee 0.773 0.769 0.529 2.04
Atlanta 0.734 0.757 0.494 2.01
San Francisco 0.679 0.776 0.393 1.97
Pittsburgh 0.707 0.763 0.446 1.76
Boston 0.763 0.720 0.578 1.69
Texas 0.739 0.762 0.482 1.07
Seattle 0.748 0.766 0.488 1.04
Cincinnati 0.774 0.826 0.427 -0.26
Arizona 0.782 0.682 0.627 -0.68
Houston 0.836 0.699 0.675 -0.81
Toronto 0.721 0.749 0.451 -0.91
Detroit 0.759 0.784 0.444 -1.77
NY Mets 0.770 0.781 0.463 -1.79
Philadelphia 0.697 0.794 0.346 -1.92
LA Dodgers 0.790 0.656 0.655 -2.16
Chicago Sox 0.737 0.753 0.451 -2.23
Chicago Cubs 0.741 0.719 0.500 -2.43
St. Louis 0.746 0.729 0.488 -2.88
Washington 0.815 0.721 0.590 -3.01
Cleveland 0.768 0.707 0.543 -3.15
Miami 0.742 0.748 0.444 -3.85
Oakland 0.731 0.751 0.422 -4.26
Tampa Bay 0.777 0.729 0.512 -4.43
NY Yankees 0.803 0.697 0.543 -8.06

Orioles: Their run differential is -75, so I thought maybe they had a great record in 1-run games. It is good at 12-8, but I don't think that accounts for winning 9 extra games. With runners in scoring position, their OPS is .856 while just .711 with none on. So far this year, for all of MLB, those numbers are .771 and .734. Maybe that has helped the Orioles score more runs than expected. But maybe not.

Over the years 2010 - 2012 here is the regression generated team runs per game based on OBP and SLG

R/G = 14.71*OBP +  10.37*SLG - 4.57

The O's have an overall OBP of .308 and SLG of .422. The equation estimates that they would score 4.34 runs per game while it was actually 4.44. Not a big difference.

Their pitching might explain things. They have allowed an .830 OPS with none on but just .786 with RISP. So that is .044 better, combined with the .037 differential for all of MLB in the opposite direction and we have a .081 swing. So that could explain quite a bit. The O's pitchers are just doing well with RISP.

But the O's pitchers have allowed OBP and SLG of .355 & .466. That estimates to 5.48 runs per game while the actual is 5.34. Combine that with the extra 0.10 from offense and we have a swing of .24 runs per game. Over 82 games, that is about 20 runs or just 2 extra wins.

Hard to see what is going on. I would have expected a much better record in 1-run games. If we give them to extra wins there, we are still only at 4 extra wins far below the 9 we get. But, their overall predicted pct is .377. Over 20 1-run games, that would be 7.5 wins, or 4.5 less than they actually have, accounting for half the difference.

So I looked at their OPS and OPSA in close and late situations. They are .698 and .710, respectively. I thought they might have a big positive differential here, but they don't.

The O's hitters have in High, Medium & Low leverage cases their hitters have an OPS of .752, .783 and .682. Their pitchers have .741, .783 and .883. So the positive differential in high cases probably helps. See comments on Twins.

Twins: They are 9-4 in 1-run games and they have a -55 run differential. So maybe they have 2-3 more wins than expected in 1-run games but that is far below the extra 7 wins. Their pitchers and hitters don't do any better than they normally do with runners on or with RISP (alot worse in some cases). They have a .633 OPS win close and late situations while they give up .676. Not sure how they are doing so well in 1-run games.

They do allow a only a .689 OPS in high leverage situations while it is .841 & .803 in medium and low cases. So maybe when it is close and late they do really well with runners on. In High, Medium & Low cases, their hitters have .761, .746 and .730. So that means they have a very good differential in high leverage cases.

Yankees: They are 9-16 in 1-run games. This probably explains alot of why they have won 8 fewer games than expected. In High, Medium & Low leverage cases their hitters have an OPS of .716, .822 and .825. Their pitchers have .725, .717 and .672. So the negative differential in high cases probably hurts.

But I do have a regression equation that uses the OPS differentials in the High, Medium and Low cases. Here it is

Pct = .5 + .306*LOW +.420*MED + .564*HIGH

Using this equation, Yankees have won only 3.4 fewer games than expected. The Orioles and Twins have won 3.53 and 2.77 more games than expected, respectively. Only one team ends up more than 5 wins off, the A's at -5.31. So taking into account how teams do based on leverage makes a big difference. Maybe differences still exist because of errors, turning DPs and/or base running. Also, I used OBP & SLG instead of OPS, things might still get more accurate.


TEAM OPSDiff W L Pct
Arizona 0.100 52 31 0.627
Atlanta -0.023 40 41 0.494
Baltimore -0.091 40 42 0.488
Boston 0.043 48 35 0.578
Chicago Cubs 0.022 41 41 0.500
Chicago Sox -0.016 37 45 0.451
Cincinnati -0.052 35 47 0.427
Cleveland 0.061 44 37 0.543
Colorado -0.001 49 36 0.576
Detroit -0.025 36 45 0.444
Houston 0.137 56 27 0.675
Kansas City -0.018 42 40 0.512
LA Angels -0.043 43 44 0.494
LA Dodgers 0.134 55 29 0.655
Miami -0.006 36 45 0.444
Milwaukee 0.004 45 40 0.529
Minnesota -0.057 42 40 0.512
NY Mets -0.011 38 44 0.463
NY Yankees 0.106 44 37 0.543
Oakland -0.020 35 48 0.422
Philadelphia -0.097 28 53 0.346
Pittsburgh -0.056 37 46 0.446
San Diego -0.082 34 48 0.415
San Francisco -0.097 33 51 0.393
Seattle -0.018 41 43 0.488
St. Louis 0.017 40 42 0.488
Tampa Bay 0.048 43 41 0.512
Texas -0.023 40 43 0.482
Toronto -0.028 37 45 0.451
Washington 0.094 49 34 0.590


Friday, June 30, 2017

How Well Did "Replacement" Catchers Actually Hit?

I looked at the top 100 or so catchers in career PAs since 1945. Then I looked for catchers who had a season with a 200+ PA decline followed by a season with a 200+ PA increase. So they went down at least 200 PAs from season 1 to season 2, but then from season 2 to season 3 they went back up at least 200 PAs. They had to be on the same team all three seasons. I then calculated the OPS+ of the "replacement." All data was from Baseball Reference.

The idea is that these catchers were temporarily replaced. Catcher is a position for which it is probably very hard to simply move a guy from another position. In the outfield, you can move a right fielder to center, for example. Maybe move a SS to 2B or a 3B man to 1B. But at catcher, you really need to find another guy who is a catcher.

Unfortunately, there were not many cases. Only 11. And I used two seasons for Fisk, 1974 and 1975. He came back to being a regular in 1977. In almost all cases, the catcher had 400+ PAs in both season 1 and season 3.

Then I found the difference in PAs for each player, comparing season 1 to season 2, and season 2 to season 3. A catcher might have 500 PAs in season 1, then 200 in season 2 and 450 in season 3. So in one case I could assign 300 PAs to the "replacement" catcher and 250 in the other.

That means I did two calculations for the OPS+ of the "replacement," one using the difference in PAs between season 1 and season 2 and the other using the difference in PAs between season 2 and season 3. The "replacement" OPS+ is a weighted average of the OPS+ of the X number of PAs for catchers on the team in season 2 who ranked lowest in OPS+.

For example, if a team needed to replace 300 PAs at catcher, I ranked any catchers on the team from highest to lowest in OPS+. Then I came up with a total number of PAs that equaled 300 using the lowest ranked catchers in OPS+ on the team. It might be 100 PAs from one catcher and then 200 from another. Then I calculated their combined OPS+, weighted by PAs.

The table below shows the OPS+ for the replacements. ROPS+1 refers to the OPS+ of the "replacement" assuming the number of PAs needing to be replaced is the difference in the regular starting catcher's PAs from season 1 to season 2. ROPS+2 is for the season 2-3 difference. The last column shows the OPS+ for the regular starter over the three seasons in question.


Player Year ROPS+1 ROPS+2 St. OPS+
Alomar 1999 61 58 76
Crandall 1961 91 82 110
Fisk 1974 61 61 122
Fisk 1975 52 52 122
Girardi  1991 68 59 71
Kendall 1999 26 27 129
Lieberthal  2001 63 63 105
Lopez 1999 67 67 117
Posada 2008 48 46 135
Scioscia 1983 66 66 92
Varitek 2001 78 78 93

Generally, the replacements did much worse. ROPS+1 has a weighted average of 64 while ROPS+2 has a weighted average of 60. The weighted average of the OPS+ for all these catchers, over the three seasons (4 for Fisk) is 109. All catchers with 1000+ career PAs since 1945 had a combined OPS+ of 92 (also a weighted average).

So when teams have had to "replace" their regular starting catcher, temporarily, the "replacements" had an OPS+ in the range of 60-64 while an average catcher would have had 92. But the drop off compared to the actual starters on these specific teams was even greater, since they a had a combined OPS+ of 109.

Why the guys who needed to be replaced tended to be above average is not clear. I did look at the guys with the longest careers and maybe one of the reasons why they had long careers is that they generally hit better than average catchers.

Again, I only had 11 observations. So the conclusions from this might be limited. Maybe this analysis will give other people some ideas on how to approach this issue.

Wednesday, June 7, 2017

Jim Baker On The Best Batter Games Of All Time Using A Fantasy Formula

Fellow sabermetrician Jim Baker has compiled the best individual games of all time. Batters get one point for every total base, run scored, RBI, walk and stolen base (this is a formula many leagues use). The table below has the top 12. None of these guys stole a base in their games. Notice two of them have been from this year and from seemingly unlikely players.



Pts
Player
Date
r
h
2b
3b
hr
rbi
bb
32
Shawn Green
5/23/2002
6
6
1
0
4
7
0
32
Mark Whiten
9/7/1993
4
4
0
0
4
12
0
31
Scooter Gennett
6/6/2017
4
5
0
0
4
10
0
31
Anthony Rendon
4/30/2017
5
6
1
0
3
10
0
31
Gil Hodges
8/31/1950
5
5
0
0
4
9
0
31
Tony Lazzeri
5/24/1936
4
4
0
1
3
11
1
30
Josh Hamilton
5/8/2012
4
5
1
0
4
8
0
30
Fred Lynn
6/18/1975
4
5
0
1
3
10
0
30
Joe Adcock
7/31/1954
5
5
1
0
4
7
0
30
Walker Cooper
7/6/1949
5
6
0
0
3
10
0
29
Mike Schmidt
4/17/1976
4
5
0
0
4
8
0
29
Phil Weintraub
4/30/1944
5
4
2
1
1
11
2