cderosa
joeashp

In Ask Bill, joeashp noted that I hadn't specified what my problem was with OPS. I’ll take that bait.

On-base Plus Slugging (OPS) was a lamentable turn for sabermetrics because it isn’t an attempt to measure offense, it’s just adding up good stuff willy-nilly.

Around the time that Bill James was explaining to a national readership how Runs Created pretty accurately predicts a team’s runs scored, there was a writer in a prominent newspaper arguing that if you added guys’ hits, homers, RBIs, and doubles (or something like that), it proved Don Mattingly was the best player in baseball. In other words, if you added up all the stuff Don Mattingly was really good at, then Don Mattingly was the best.

That’s what OPS is to me: people who already believe that on base percentage and slugging percentage are the most important stats, telling you that when you add these together, it gives you the best hitters. That’s a poor argument to approach anyone who isn’t already down with the basic sabermetric understanding of how runs are created.

And if you are already versed in those arguments, then OPS isn’t enough for you. You’ve got to fine-tune it for the offensive environment, and hence, OPS+. In his question to Bill, joeashp proposes fixing OPS to get around a double-counting issue. Now you are going to a lot of trouble to refine a stat that was never intended to make an accurate measure of offense in the first place. Its only selling point was its accessibility.

The way I remember it, this thing had been around for a while but it was Rob Neyer’s ESPN column that popularized it (that column popularized many good things, despite OPS). Neyer’s argument was that OPS correlated with team runs *almost* as well as Runs Created and other complex measures anyway, and it was a lot easier to figure.

While easy to figure, though, it has no compelling logic and pushes us farther than we need to be away from expressing value in runs. If somebody has a 1.107 or a .718 OPS, what’s that? It’s not runs. You can’t even describe it as bases-per-something, because you’ve added together elements with different denominators. So you need to internalize a new set of standards to contextualize a stat that was just supposed to be a convenient prop.

Any time you combine strong offensive indicators, you are going to get some strong-looking connection with team runs. That doesn’t mean that just adding the good stuff together is the best way to go. It isn’t much more complex to figure on-base *times* slugging than it is to figure on-base *plus* slugging. On-base percentage times slugging percentage (OBP x SLG) is just basic Runs Created divided by at bats. With no real added difficulty in terms of calculation, you’d have a stat that is grounded in a demonstrated predictive relationship with actual team runs, that is expressed in runs, and is easily toggled between a rate and a total.

So that's why I've always felt OPS was a wrong-headed detour.

Chris DeRosa

11:38 AM Aug 26th

cderosa
Well I think there are times when you don't want to introduce park effect adjustments. Any time you adjust for park effects, you are entering the argument of how best to adjust for those effects. Are you using one year, three years, five years? Are the seasons weighted? Are you mixing different methods because some parks have been around longer than others?
To the extent that the author is actually rating pitchers, saying who is better than who, yes, he should deal with the park issue as best as he can. I think we'd agree the site's "Total Runs," which doesn't deal with park effects (or even outs used), doesn't cut it as a stand-alone rating. But sometimes when you are working on a smaller part of the question, and it isn't helpful to get bogged down in gray areas.
Here, the issue the author seems to be after is the variance between pitchers' ERAs and their results on a PA-by-PA basis. A good opportunity, I think, for the site to take better advantage of the information it has on hand. The Component ERA was designed to study this very issue.
As to OPS, I thought I was being generous by half an ass.
7:35 AM Aug 19th

cderosa
This site actually publishes Component ERA in the player profiles, which addresses the very question. There's no need at all to resort to the half-assed OPS stat.
8:29 AM Aug 16th

jemanji
By 'generally believed,' John is referring to a vast consensus that strand rate is not sustainable from one year to the next. He might as well have said it is 'generally believed' that fit shortstops are more desirable than real fat ones.

I don't mean to be snotty, but as a poster on a Bill James site, quibbles like that seem like they set a pretty low standard for idea exchange.

.........

It is self-evident that a pitcher's OBP+SLG will edit out things like 'batting average with runners in scoring position', and that ERA will capture them. It is also self-evident why this has value.

It's okay if you couldn't deduce that for yourself, but it comes off as poor form to lecture John for sloppy thinking when the reverse was the case.

Your friend,
Jeff


5:10 PM Aug 15th