Isolated On Base

If you follow baseball advanced stats at all, you're probably familiar with Isolated Slugging (ISO). It's a stat that separates a player's slugging percentage from his batting average, to get an idea of how much of his slugging is from extra base hits vs. how much is just a load of singles propping up a lack of power (I didn't say anything, Jon Jay, I don't know why you're looking at me like that).

The cool thing about ISO is its simplicity. You can calculate it in your head, just based on the slash line you would get on the back of a baseball card, or on the jumbotron at a stadium.

Is there a comparable stat for OBP? My hypothesis is that we can subtract batting average from OBP to get a fairly good estimate of walk rate. In homage to ISO, I call this stat IBO.

Comparing OBP to average is trickier that slugging and average. Both SLG and AVG are denominated on AB. OBP is more complicated: (H+BB+HBP) / (AB +BB+HBP+SF), per Baseball-Reference.

To test this out, I'm using the Lahman Baseball database data from 2014, for all qualified hitters (a total of 135 players).

First step, I created a table to store IBO values. I'm going to get a little hand-wavey here, because I don't want to post all the database structure code. Trust me on this, I'm pretty sure I subtracted one number from another successfully.*

That accomplished, I wrote an R script to pull the data and do some MATH!

library(RMySQL)

ammlb = dbConnect(MySQL(), user='myuser', password='mypassword', dbname='AMMLB', host='localhost')

rs = dbSendQuery(ammlb, "select * from vw_QualifiedIBO where yearID = 2014")
tempdata = fetch(rs, n=-1)
query_results = data.frame(tempdata)
leaders <- query_results[c("nameLast", "nameFirst", "BBPct", "IBO")]
leaders <- leaders[order(-leaders$BBPct),]
print(leaders)

The leaders are who you would expect, the guys who walk a lot:

Last Name	First Name	BB%	IBO
Santana	Carlos	0.1712	0.1341
Bautista	Jose	0.1548	0.1176
Stanton	Giancarlo	0.1473	0.1074
LaRoche	Adam	0.1399	0.1027
Carpenter	Matt	0.1344	0.1025
Smith	Seth	0.1327	0.1009
Werth	Jayson	0.132	0.1022
Fowler	Dexter	0.131	0.0985
McCutchen	Andrew	0.1296	0.0966
Freeman	Freddie	0.1271	0.0973
Dozier	Brian	0.1264	0.1027
Ortiz	David	0.1246	0.093
Crisp	Coco	0.1234	0.0902
Granderson	Curtis	0.1208	0.0987
Valbuena	Luis	0.119	0.0917
Rizzo	Anthony	0.1185	0.1001
Trout	Mike	0.1177	0.0899
Duda	Lucas	0.1158	0.0961
Mauer	Joe	0.1158	0.0841
Moss	Brandon	0.1155	0.1005
Zobrist	Ben	0.115	0.0824
Davis	Chris	0.1145	0.104
Encarnacion	Edwin	0.1144	0.0859
Teixeira	Mark	0.1142	0.0971
Holliday	Matt	0.1109	0.0985
Choo	Shin-Soo	0.1096	0.0985
Donaldson	Josh	0.1094	0.0875
Ramirez	Hanley	0.1094	0.0862
Martinez	Victor	0.1092	0.0736
Yelich	Christian	0.1065	0.0788
Rollins	Jimmy	0.1056	0.0799
Crawford	Brandon	0.105	0.0774
Puig	Yasiel	0.105	0.0867
Howard	Ryan	0.1034	0.087
Heyward	Jason	0.1032	0.0808
Gordon	Alex	0.1011	0.0851
Lucroy	Jonathan	0.1008	0.0716
Montero	Miguel	0.1	0.0852
Upton	BJ	0.0984	0.0786
Carter	Chris	0.0979	0.0809
McGehee	Casey	0.097	0.0673
Upton	Justin	0.0938	0.0719
Beltre	Adrian	0.0928	0.0634
Peralta	Jhonny	0.0924	0.0735
Cano	Robinson	0.0917	0.0677
Plouffe	Trevor	0.0911	0.0705
Lowrie	Jed	0.0904	0.0719
Kipnis	Jason	0.0903	0.0705
Gardner	Brett	0.0899	0.0715
Jennings	Desmond	0.0882	0.0746
Cabrera	Miguel	0.0876	0.0582
Markakis	Nick	0.0873	0.0666
Kemp	Matt	0.0868	0.0591
Rendon	Anthony	0.0852	0.0639
Gonzalez	Adrian	0.0848	0.059
Jones	Garrett	0.0841	0.063
Pedroia	Dustin	0.0837	0.0589
Abreu	Jose	0.082	0.0661
Escobar	Yunel	0.0819	0.0654
Longoria	Evan	0.0815	0.0673
Cruz	Nelson	0.0811	0.0625
Bruce	Jay	0.0809	0.0643
Eaton	Adam	0.0802	0.0615
Utley	Chase	0.0798	0.069
Seager	Kyle	0.0796	0.066
Aoki	Nori	0.0795	0.0643
Walker	Neil	0.0789	0.0706
Frazier	Todd	0.0788	0.0634
Posey	Buster	0.0777	0.0528
Ellsbury	Jacoby	0.0773	0.0568
Brantley	Michael	0.0769	0.0573
Span	Denard	0.0752	0.0533
Freese	David	0.0744	0.0612
Chisenhall	Lonnie	0.0739	0.0625
Pence	Hunter	0.0734	0.055
Gomez	Carlos	0.0731	0.0721
Wright	David	0.0717	0.055
Kendrick	Howie	0.0715	0.0538
Gillaspie	Conor	0.0711	0.0537
Calhoun	Kole	0.071	0.0534
Desmond	Ian	0.071	0.0587
Braun	Ryan	0.0707	0.0581
Cabrera	Melky	0.0695	0.0495
Pujols	Albert	0.0691	0.052
Andrus	Elvis	0.0681	0.0508
Butler	Billy	0.068	0.052
Martin	Leonys	0.0677	0.0508
Ozuna	Marcell	0.067	0.048
Castro	Jason	0.0665	0.0642
Brown	Domonic	0.0664	0.0505
Bogaerts	Xander	0.0659	0.0575
Hosmer	Eric	0.064	0.0477
Mercer	Jordy	0.0636	0.0506
Loney	James	0.063	0.0464
Castellanos	Nick	0.0622	0.0468
LeMahieu	DJ	0.0621	0.0473
Morneau	Justin	0.0618	0.0449
Castro	Starlin	0.0615	0.0475
Navarro	Dioner	0.0615	0.0429
Sandoval	Pablo	0.0611	0.0456
Murphy	Daniel	0.0607	0.0432
Marte	Starling	0.0606	0.0651
McCann	Brian	0.0595	0.0539
Ackley	Dustin	0.0594	0.0481
Davis	Khris	0.0583	0.0552
Reyes	Jose	0.0582	0.0408
Infante	Omar	0.0579	0.0428
Viciedo	Dayan	0.0568	0.0492
Hamilton	Billy	0.0565	0.042
Aybar	Erick	0.0564	0.0429
Jeter	Derek	0.0559	0.047
Simmons	Andrelton	0.0557	0.0413
Byrd	Marlon	0.0549	0.0484
Hill	Aaron	0.0518	0.043
Hardy	JJ	0.0512	0.0408
Segura	Jean	0.0512	0.0432
Altuve	Jose	0.051	0.0359
Blackmon	Charlie	0.0483	0.0465
Dominguez	Matt	0.0479	0.0417
Gordon	Dee	0.0479	0.0371
Cozart	Zack	0.0465	0.0464
Gomes	Yan	0.0463	0.0343
Adams	Matt	0.0462	0.0331
Hechavarria	Adeiny	0.0457	0.0315
Rios	Alex	0.0441	0.0304
Harrison	Josh	0.0401	0.0313
Kinsler	Ian	0.0401	0.0322
Ramirez	Aramis	0.0395	0.0442
Hunter	Torii	0.0392	0.0331
Johnson	Chris	0.0378	0.0294
Escobar	Alcides	0.0376	0.032
Ramirez	Alexei	0.0366	0.0316
Perez	Salvador	0.0363	0.0293
Jones	Adam	0.0279	0.0298
Revere	Ben	0.021	0.0185

So how do those numbers actually correlate?

That looks pretty good... but who are we, the old scouts from Moneyball? Let's do the numbers.

print("Correlation between BBPct and IBO:")
correlation <- cor(leaders$BBPct, leaders$IBO)
print(correlation)
[1] 0.9704839

0.97 correlation. Yeah, that seems alright. So we can get a general sense of what IBO looks like (at least last year):

.090+ Among the league leaders
.065 Above average
.050 Middle of the pack
.040 Below average
.019 Ben Revere

EDIT: it occurs to me that quartiles might be a useful measure, and I figured out you can do it with a single command in R!

quantile(leaders$IBO)

[1]   0%     25%     50%     75%     100%
   0.01850 0.04715 0.06150 0.08035 0.13410

So that's nice, but can we use it to extrapolate walk rate?

Tune in next time!

* Ha! Double check on me when this gets up to BitBucket. I'm working on getting it ready to open source.