OPS+ vs. EqA vs. wOBA

[AstrosFan]

Which of these offensive metrics do you prefer, and why?

[SABR Matt]

depends.

I prefer OPS+ for quick look-up but for detailed analysis, I prefer wOBA (because it does a better job modeling real runs than OPS or EqA do).

[AstrosFan]

Do you see OPS+ as better at modeling runs than EqA? For quick look-up, those two seem to be about even.

[Tango Tiger]

EqA is much better than OPS+. EqA is a disguised version of Linear Weights, but with park factors thrown in.

EqA would be slightly better than wOBA, because of the park adjustment. But, it wouldn't be that hard for someone to apply park factors to wOBA, and make it superiod to EqA.

Conceptually, wOBA is better than EqA.

[Patriot]

EQA models runs better than OPS+ for my money. For all of the bells and whistles, EQR/EQA is essentially a linear weights method, non-linear only because of the presence of SB and CS in the denominator of the "RAW" statistic. While I have no use for the EQA scale, if you raise it to the 2.5 power and multiply it by 5, you are back to runs/out.

OPS+ is essentially an estimate of runs/out relative to the league average. If you treat it as such, and find the linear weights for an average team, you'll get something like this:

.47S + .81D + 1.16T + 1.50HR + .26W - .087(AB - H)

EQR comes out to something like this:

.52S + .84D + 1.16T + 1.48HR + .36W + .24SB - .23CS -.117(AB-H)

For average teams, little difference. However, OPS+ will severely understate the value of great offensive players, as it does not take OBA or outs directly into account outside of the initial consideration of OBA.

wOBA uses proper linear weights, is easily convertible to runs, and thus is far preferable to either.

[SABR Matt]

Originally Posted by AstrosFan

Do you see OPS+ as better at modeling runs than EqA? For quick look-up, those two seem to be about even.

No...I just see OPS+ as easier to understand. EqA may be disguised linear weights and therefore more logically accurate, but you can more readily look up the components of OPS and figure out how a player accomplished his OPS.

[AstrosFan]

Even assuming we use Total Baseball's version of OPS+, which is much simpler to computer: 100*(OBP/LgOBP+SLG/LgSLG-1)/PF, finding the components is not an easy thing to do. Baseball Reference's league figures are for all the players in the league, but when doing calculations, both TB and B-Ref remove the pitchers. The league components for making the appropriate calculations are not something you can just look up yourself. You have to remove the pitchers yourself, if you wish to calculate it on your own.

[AstrosFan]

Let us assume, for the sake of argument, that wOBA is also adjusted for league and park factors. Without that, it is at a disadvantage to the other two methods, as it would be measuring raw production, without a context element. When we build that in, the playing field becomes more level.

If I have the three at my disposal, and they are all context adjusted, I go with wOBA, because, as has already been mentioned, it best models linear weights and run scoring. But working from scratch, wOBA is at a disadvantage, because historical context adjusted wOBA for all players in history is not something I can just look up, like I can with OPS+ and EqA. In fact, I have a database that makes it easy for me to calculate context adjusted wOBA, and for that reason, it remains my preferred method. But because it does not have the same reference value the other methods do, the answer is not as clear-cut for other people.

[Colin Wyers]

Originally Posted by AstrosFan

Let us assume, for the sake of argument, that wOBA is also adjusted for league and park factors. Without that, it is at a disadvantage to the other two methods, as it would be measuring raw production, without a context element. When we build that in, the playing field becomes more level.

If I have the three at my disposal, and they are all context adjusted, I go with wOBA, because, as has already been mentioned, it best models linear weights and run scoring. But working from scratch, wOBA is at a disadvantage, because historical context adjusted wOBA for all players in history is not something I can just look up, like I can with OPS+ and EqA. In fact, I have a database that makes it easy for me to calculate context adjusted wOBA, and for that reason, it remains my preferred method. But because it does not have the same reference value the other methods do, the answer is not as clear-cut for other people.

I don't know that any one metric is the best, and a lot of it depends on how much time you're willing to invest in the matter and how much accuracy you really need.

The other concern is who you're presenting the data to. OPS+ and EqA come with "branding" that make them a lot easier to bring up in casual and semi-casual conversations than wOBA. The more popular Baseball-Reference.com becomes, the more common it is for casual fans to use and reference OPS+; unless you really need the additional accuracy it's probably best to stick with OPS+ if you're trying to write for a broader audience.