July 19, 2000
The Daily Prospectus
The Limits of OPS
After we posted the results of our informal mid-season awards poll last week, we got a number of responses that looked like this:
OK...who does the voting for your MVP award?
You get the idea. And yes, we have been prominent in the push for greater visibility of on-base percentage and slugging percentage. They are the two most important offensive statistics for individuals. OPS, the sum of the two, serves as a much better single number to judge a player than the more traditional batting average.
So now that OPS has made inroads in television coverage, in Internet and print media coverage and even within the game itself, perhaps it's safe to point out that OPS is not, and never has been, a magic bullet. The metric has flaws that make it a poor tool for serious, in-depth evaluation of performance.
The first one is that OPS hides information. In the desire to get to one number, OPS throws together two things that do not have equal value. On-base percentage is more important than slugging percentage to an offense, so players with the same OPS can have much different value as hitters.
OPS AVG OBP SLG
OPS would cloud the fact that these two players are having completely different seasons. Corey Koskie is hitting for a good average and taking a lot of walks, while hitting enough doubles to give him a good slugging percentage. Dean Palmer is simply having a Dean Palmer season: a .270 batting average with some walks and good power. Before adjusting for park, Koskie's season would appear to be more valuable.
You can do this for just about any number between 700 and 1000, if you're bored enough. The point is that OPS doesn't provide enough information to gauge value. As both a writer for and the editor of Baseball Prospectus, I strive to use the notation of AVG/OBP/SLG to sum up a player's performance.
Beyond the clouding effect, OPS factors only a player's performance at the plate, Playing time, base stealing, baserunning, position and defensive performance are ignored. Anyone who votes for MVP, makes a trade or signs a free agent without considering these factors and solely considers OPS--and I strongly doubt anyone does--is just as wrong as the "batting average uber alles" crowd.
Our critic above is guilty of just this fallacy. Yes, Carlos Delgado and Mark McGwire are having phenomenal offensive seasons, but as first baseman of moderate defensive ability, they don't deserve MVP votes ahead of a shortstop like Alex Rodriguez or a second baseman like Jeff Kent. Value over the average, or replacement, player at a position has to be considered, rather than simply raw batting totals.
Finally, OPS is not adjusted for context, and in making serious evaluations of performance, you have to consider context. This is where more advanced metrics, such as Clay Davenport's Equivalent Average (EqA), come into play. EqA adjusts for the player's home park.
Returning to our slow white third basemen, we noted earlier that they had similar OPSs but that Koskie's greater OBP meant he was having the better season offensively, before adjusting for park. Well, if you do adjust for park, you find that the two players actually do have nearly identical EqAs (Koskie: .274, Palmer, .273), courtesy the difference between playing home games in the Metrodome and in Comerica Park.
In this case, the shape and the context of the performances cancel out, leaving two equivalent seasons. That, of course, is not always the case. Take a trio of NL second basemen:
OPS AVG OBP SLG EqA
These three guys are having completely disparate seasons, despite having nearly identical OPSs. And while Ron Belliard and Bret Boone are a similar comp to Koskie and Palmer, with one player's OBP edge being canceled out by a park difference, it's clear that Jay Bell's hitting comes in a bit below both of them.
So while the growing use of OPS is a quantum leap forward for baseball coverage and management, it is important to recognize its limitations. OPS is not the be-all, end-all of performance metrics; it's simply a quick-and-dirty way of evaluating hitting when no better tools are easily available.
By the way, I want to correct an item currently appearing in the NL West Notebook. In it, I wrote that games at Coors Field had seen 38% more runs than an average game, a figure I pulled from Clay Davenport's EqA reports. What I pulled, though, was the park factor, which is merely the adjustment factor for Rockies' players.
In actuality, the environment in Denver has increased run scoring in 2000 by 90%. This doesn't impact the conclusions of the piece; rather, it makes the point even more dramatically that the distorting effects of altitude mandate that they be accepted and factored in to any evaluation of the Rockies.
Thanks to reader Roy White for his eagle eye. Roy, contact Gary Huckabay for your tickets to his one-man musical, "La Cage Aux Folles: The Molting Years."
Joe Sheehan can be reached at email@example.com.