Wednesday, November 29, 2006

Team Run Production in 2006

Statistics such as runs created and OPS (on base percentage plus slugging percentage) are good for summarizing the offensive performance of a player. However, they don’t tell you much about what kind of hitter a player is. How to best break down offense into statistics that describe skills is always under debate and there is no one best way to do it. The combination of statistics most commonly used by analysts is batting average (BA), on base percentage (OBP) and slugging average (SLG). These are the statistics I typically use too but the combination does have some problems.


First, BA, OBP and SLG are highly correlated. For the more statistically inclined, a look at individual seasons of players in the American League between 2001 and 2005 shows that the correlations range from .63 to .68 with the highest correlation between BA and OBP. In short, this says that the three stats are measuring similar things. Logically, consider that batting average is essentially a subset of both OBP and SLG and thus the three of them are closely related.


Another problem (and this is related to the first issue described above) is that neither OBP nor SLG measures a pure skill. OBP combines the ability to get hits with the ability to draw walks. These are two very different skills which can be loosely defined as making contact (hits) and having a good eye (walks). Similarly, slugging combines the ability to get hits with the ability to get extra base hits. Again, these are two different skills – making contact and hitting for power.


BA can be subtracted from SLG to give us isolated power (ISO), a purer measure of slugging ability than SLG. Similarly, BA can be “subtracted” from OBP to give us extra on base percentage (EOBP). Because BA and OBP have different denominators (at bats and plate appearances respectively), it is best not to use the equation OBP-BA. Instead, we subtract hits from times reached based in the numerator and use plate appearances as the denominator. That is, EOBP=(BB+HBP)/PA. So we now have BA, ISO and EOBP. The correlations between these three statistics range from .01 to .42 with the highest correlation between ISO and EOBP. These correlations are much lower that those between BA, OBP and SLG. This is because BA, ISO and EOBP are more independent of one another than BA, OBP and SLG.


Furthermore, using AL team data from 2001 to 2005, I found that BA, ISO and EOBP explain 93% of the variation in team runs scored. I also found that BA, OBP and SLG explain 93% of the variation in runs scored. So the two sets of statistics have the same very strong explanatory power but the former trio is appealing because it more purely measures three separate skills.


Finally, BA in isolation, explains 63% of the variation in runs, ISO explains 59% and EOBP explains 49%. So neither of the three statistics explains a great deal by itself. All three must be used together to explain runs scored.


I have not included base running in run production because the only readily available statistic is stolen bases which is not highly correlated with runs scored. I can not analyze other base running statistics at this time without a lot of data entry and potential copyright issues. So base running will have to wait for another time. Even without it though, we are explaining most of what we need to know about runs scored as that 93% figure is very high.



Table 1 below presents the BA, ISO and EOBP for all American League teams in 2006. Table 2 shows the ranks for the three statistics. From the tables, we can see that the Tigers finished 5th in runs scored despite ranking 9th in BA and 13th in EOBP. This further illustrates how the Tigers had the most efficient offense in the American League in 2006. Much like last year and most other recent years, their biggest weakness was their inability to draw walks. We probably can not count on them being so efficient (lucky?) with their production next year. They will likely need to improve their walk rate if they are to rank as high as or higher than 5th in runs scored again in 2007.


Table 1: American League Team Run Production in 2006


Team

Runs

BA

ISO

EOBP

NY Yankees

930

.285

.177

.112

Cleveland

870

.280

.177

.097

Chicago Sox

868

.280

.184

.089

Texas

835

.278

.168

.087

Detroit

822

.274

.174

.077

Boston

820

.269

.166

.115

Toronto

809

.284

.179

.093

Minnesota

801

.287

.138

.087

Oakland

771

.260

.152

.112

Baltimore

768

.277

.146

.088

LA Angels

766

.274

.150

.085

Kansas City

757

.271

.140

.087

Seattle

756

.272

.153

.076

Tampa Bay

689

.255

.165

.081

Average

804

.275

.162

.092


Table 2: American League Team Run Production Ranks in 2006


Team

Runs Rank

BA Rank

ISO Rank

EOBP Rank

NY Yankees

1

2

3

2

Cleveland

2

4

4

4

Chicago Sox

3

5

1

6

Texas

4

6

6

8

Detroit

5

9

5

13

Boston

6

12

7

1

Toronto

7

3

2

5

Minnesota

8

1

14

9

Oakland

9

13

10

3

Baltimore

10

7

12

7

LA Angels

11

8

11

11

Kansas City

12

11

13

10

Seattle

13

10

9

14

Tampa Bay

14

14

8

12

No comments:

Post a Comment

Twitter

Blog Archive

Subscribe

My Sabermetrics Book

My Sabermetrics Book
One of Baseball America's top ten books of 2010

Other Sabermetrics Books

Stat Counter