Modifying OPS

by salb918 on Jun 30, 2005 11:48 PM EDT Comment 28 comments

OPS is a quick and dirty stat that allows us to measure the rate of a player's offensive contribution, but we all know OPS doesn't tell the whole story. More advanced and revealing metrics like VORP and EqA are currently in vogue, but I've been in search of a stat that is easily calculated using the "Big Three" rate states (AVG/OBP/SLG). If I want VORP and EqA, I'm going to Baseball Prospectus, not calculating it myself.

Lately I've been playing around with some numbers and have come up with a modified version of OPS which I call mOPS (modified OPS):

mOPS = 2.5*OBP + ISO

where
OBP = on-base percentage
ISO = isolated power = slugging average - batting average

I should point out that this formula does not consider stolen bases, sac flies, or other "secondary effects." (These secondary effects can make a difference in the valuation of say, Juan Pierre or Scott Podsednik, but we'll have to live with that innacuracy for now.) mOPS is quick and easy to calculate, and - in my opinion - places the appropriate weight to OBP (Life is OBP; OBP Is life) and doesn't double count batting average in the way simply adding OBP and SLG does (that is, a player with a high batting average can have his OBP and SLG driven by his AVG - why give credit twice?)

The league leaders in mOPS are:

Player            OPS    mOPS
Derrek Lee        1.163    1.468
Brian Roberts        1.055    1.3475
Albert Pujols        1.027    1.323
Bobby Abreu        0.982    1.3205
Alex Rodriguez        1.013    1.32
Nick Johnson        0.952    1.298
David Dellucci        0.925    1.281
Adam Dunn        0.944    1.28
Jim Edmonds        0.969    1.2765
Vladimir Guerrero    1.016    1.274
Travis Hafner        0.947    1.258
Brian Giles        0.913    1.2525
Carlos Delgado        0.959    1.252

Notice how mOPS gives "extra" credit to players like Bobby Abreu, Nick Johnson, and Brian Giles - guys who just don't make outs while hitting the ball all over the yard. These are guys that have often been slapped with the underrated tag. Now we know why. Bobby Abreu boosters can point to this stat: he's having a season on par with Alex Rodriguez.

OPS overrates guys like Miguel Tejada, who is 9th in OPS but only 21st in mOPS, somewhere between Gary Sheffield and Miguel Cabrera. (By the way, I love Miguel Tejada. It's just that mOPS is cruel and heartless.) Even AL MVP Vladimir Guerrero is somewhat overrated by OPS, because his OBP and SLG are driven by his .340+ AVG.

I was interested in some feedback on mOPS. In a week or two, I can post my methodology, but for now I am interested in hearing what you all think.

0 recs | Comment 28 comments |

Story-email Email Printer Print

Comments

one question

why 2.5? why not 2.4? or just 2? why 2.5? why is 2.5 the 'appropriate' weight?

interesting idea though. ill need to ponder.

by cephyn on Jul 1, 2005 2:29 AM EDT reply actions

Like I said

I'll post my methodology later. In short, though, I found an offensive measure that I liked better than Runs Created, SecAvg, etc. I'll describe this measure later.

I then performed a multivariate regression using OBP and ISO. The ratio of their coefficients was approximately 2.5, which is what you see here.

More on that when I have the time, but that's the short of it.

by salb918 on Jul 1, 2005 9:28 AM EDT up reply actions

I'm going to guess...

That the reason you picked 2.5 is because Beane believes OBP to be worth 3 times as much as slugging, and James believes it to be worth 2 times as much as slugging, so you went with the medium figure since you decreased the slugging number by using ISO instead.

Just a hunch, I'm probably off base by a mile. I look forward to hearing your methodology, and will even post it as an article if you'd like.

"I don't set the rosters, I just make fun of the guy who does" - Rob Neyer

by Marc Normandin on Jul 1, 2005 10:29 AM EDT up reply actions

Not quite

I didn't just "pick" 2.5. That would be silly, and it's the very reason I think that as advanced EqA and VORP are, they are fundamentally flawed. The weighting of on-base activities is heuristic.

More to come.

by salb918 on Jul 1, 2005 10:48 AM EDT up reply actions

I appreciate the interest in "2.5"

But what I was really after was whether or not people felt (intuitively) that this type of analysis gives appropriate credit to people like Giles and Abreu.

I promise, I will send more information as to how I arrived at the 2.5 figure. Should I send it as an attachment to Mr. Normandin or Mr. Scotto or should I post it as a diary? In any case, I may not get around to that for a week or so.

by salb918 on Jul 1, 2005 11:14 AM EDT reply actions

RE: delivery

Send it to either of us, whichever one you like better. Just kidding of course. Send it to one of us and we'll post it up. Just know if you have it ready Monday I won't be in all afternoon or most of the evening, since I will be in Pawtucket.

"I don't set the rosters, I just make fun of the guy who does" - Rob Neyer

by Marc Normandin on Jul 1, 2005 11:46 AM EDT up reply actions

Didn't mean to be rude

but much of my opinion hinges on the 2.5, thats why i asked right away.

What I like - its quick, simple, and an interesting angle to attack the "question"

What I don't like - well, the 2.5 -- I mean, that's a LOT of weight added to OBP. A LOT. It just drowns out the ISO, I think. Almost to the point of why even bother with the ISO, just look at OBP. Unless you're a monster who slugs at a ridiculous rate above average (Babe Ruth, Barry Bonds) it won't even make a dent. If you really feel OBP is The Thing, and you're going to almost triple the value of it, just use OBP.

I'd like to see a comparison list of mOPS vs. OBP and see how much they differ by, both current and some historical stuff. I think that's necessary in order to assess the value of mOPS. For me, a straight OBP+ISO would be just as useful, and easier to calculate. I'm open though, and await your argument - i want to be convinced. 8)

by cephyn on Jul 1, 2005 12:43 PM EDT up reply actions

Good comments

I hadn't thought too much about how much OBP is drowning out ISO. You are right in saying that it is a whole lot.

But look at OBP + SLG. SLG is driven by batting average for many, which correlates very well to OBP. In effect, you are counting up some OBP there as well.

Part of the usefulness is that guys who have very low ISOs have reduced values compared to guys with the similar OBPs. It's not that I think OBP is The Thing...it's just that it appears to be the thing.

I also would like to see some historical comps, and I agree that mOPS has limited usefulness until we can assess it in a historical context. Unfortunately, I don't have an electronic encyclopedia of such things, so making spreasheets is a royal pain in the butt. Does anyone have an electronic encyclopedia that might be willing to do some historical comps?

Thanks for your comments. I've been doing this on my own, and sometimes get too wrapped up in my own ideas w/o getting outside criticism.

by salb918 on Jul 1, 2005 12:50 PM EDT up reply actions

Additionally

What if you have a leadoff hitter who gets on base a lot but doesn't SLG high...say .400 and .400. Bats .300 his mOPS would be.... (.400 x 2.5) + (.400-.300)= (1 + .100) = 1.100
OPS= .800

Now let's look at a bomber who doesn't get on as much
.247/.337/.547

OPS = .884
mOPS= (.843 + .3) = 1.143

The slugger is ahead. You're telling me that Rob Deer is better than Rickey Henderson. I reject that. That's also why I'm not crazy about regular OPS - it rewards a guy who mashes more than a guy who leads off. Both are important, but a good leadoff guy is just as important as a masher, IMO. If I ask you who do you want, a .400 OBP or a .337 OBP, what will you say? If OPS is limited and not good enough, mOPS has similar issues. To even the above situation out, you'd need to further weight OBP, in which case, just look at OBP.

Thoughts?

by cephyn on Jul 1, 2005 1:10 PM EDT up reply actions

maybe

one way to correct this would be to do some of the corrections to slugging and BA for steals, caught stealing. but that defeats the simplicity of the calculation...

by cephyn on Jul 1, 2005 1:24 PM EDT up reply actions

Numbers

The problem with the numbers game is that our preconceived notions can cause us to reject conclusions.

What mOPS says is that, well, these two have equal values. The 400/400 player has a different type of value as the 337/547 bomber - but when you use a dashboard metric, you have to deal with some loss of information. I think it is perfectly acceptable to say that the two have equal values, although in actual lineup construction, one must realize they have different types of value.

I know it is difficult to criticize thoroughly until my methodology is available. For that, I apologize.

by salb918 on Jul 1, 2005 2:06 PM EDT up reply actions

loss of information

i think we lose too much information here. I think OPS is a limited statistic. I thing mOPS is similarly limited. So I question the point of creating a new limited metric. 8(

I like the concept of mOPS though.

I respect the preconceived notion argument. But I really just can't accept Rob Deer for the HoF.

Im still pondering.

by cephyn on Jul 1, 2005 2:31 PM EDT up reply actions

You needn't accept Mr. Deer for the HoF

That's just crazy talk

Deer: 220/324/442 OPS = 766 mOPS = 1032

Rickey: 279/401/419 OPS = 820 mOPS = 1142

I'll leave it up to Mr. Normandin to calculate JAWS, etc.

by salb918 on Jul 1, 2005 2:38 PM EDT up reply actions

Multiplier and Some Other Thoughts

First off, nice work on this, and I'd love to see your methodology, whenever you put it together. This is certainly interesting stuff.

I ran some regressions and some trial and error, and it seems to best correlate with run scoring with the multiplier somewhere between 2.5 and 2.8, so, in that sense, it looks pretty good.

I might be making a mistake in my calcs, but I'm only getting the correlation of mOPS to run scoring up to an R-squared value of .8916 (when I play around with the mulitplier), and I think that there are other options with similar accuracy.

The problem with isolated power over slugging is that, in my opinion, singles and walks are not created equal. Singles are worth more runs.

There's a balance between double-counting singles and weighting them the same as walks.

Here's a familiar formula:

         H + TB + 1.5*(BB + HBP) + SB
EqA  =   ----------------------------
          AB + BB + HBP + CS + SB/3

If you string out the H + TB + 1.5(BB+HBP) component of EqA, you get 2*1B + 3*2B + 4*3B +5*HR + 1.5*(BB+HBP).

Which isn't all that far off from another run estimation formula I like:

XR = (.50 x 1B) + (.72 x 2B) + (1.04 x 3B) + (1.44 x HR) + (.34 x (HP+TBB-IBB)) +(.25 x IBB)+ (.18 x SB) + (-.32 x CS) + (-.090 x (AB - H - K)) + (-.098 x K)+ (-.37 x GIDP) + (.37 x SF) + (.04 x SH)

or ERP or Batting Runs or anything.

I like the concept of modifying OPS into something more valuable. OPS is quick and dirty and with a slight modification can be a lot more. But I think that the problem with undervaluing the base hit will keep down the R-squared.

OPS overvalues singles, mOPS seems to undervalue them. Somewhere, there's a balance. And yes, I'd love to see the methodology, because I could be completely butchering this.

by Dan Scotto on Jul 1, 2005 1:34 PM EDT reply actions

Offensive figure of merit.

"I ran some regressions and some trial and error, and it seems to best correlate with run scoring with the multiplier somewhere between 2.5 and 2.8, so, in that sense, it looks pretty good."

What do you mean by run scoring? Is this on a teamwide basis?

I created my own offensive figure of merit, and by playing with the multiplier, I can get r-squared up near 0.97. So by "run scoring" you mean offensive figure of merit, and we are using different figures. I will email you and Marc about my offensive figure of merit and you can decide from there.

The balance between singles and walks you refer to: formulas like EqA and ERP and whatnot - this could be my ignorance, but what is the origin of the multipliers in these equations? 1.5? -0.37? I'm wondering because I don't know. My 2.5 comes from...well, wait for the email. Sorry for all this vagueness.

Keep the commentary flying!

by salb918 on Jul 1, 2005 2:15 PM EDT up reply actions

About to check my email....

I used team figures from 2000-2004. I might be wrong in my methodology; I just played around with Excel.

by Dan Scotto on Jul 1, 2005 11:22 PM EDT up reply actions

I have to agree

a walk is not quite as good as a single. OPS respects this by counting hits twice and walks once. Is a single twice as good as a walk? no.

For a completely random thought, what about (BA/2)+OBP+ISO -- does it tell us what we need to know now? Gives extra weight to hits, the bulk is OBP, and ISO rewards power.

I still think its flawed though.

by cephyn on Jul 1, 2005 2:37 PM EDT up reply actions