Article: Geeking Out: OPS Minus Batting Average

[fairweather]

Obviously my previous post was facetious, but I really get sick of hearing about Mauer's career BA. The guy is big and strong he should be hitting for more power. Remember 09? Where did that go?

[Seth Stohs]

Have I ever said I have a B.S. in Mathematics? I love the numbers in baseball and how to interpret them. There is almost no 2 people that place the same emphasis on the same categories. One of the many reasons I absolutely love this game.

 

I was only smart enough to get a math minor, so you win... I completely agree with you, and I think through all of the discussion, I really just like putting the Triple Sash together and realizing that all three numbers tell you a little something about the player. Batting average says one thing. On-Base Percentage says something else. Slugging says yet another thing. OPS tries to bring it all together. The only thing I wonder about is batting average really counting twice. Like has been pointed out above, you could also call it Isolated Power + OBP, or SLG + Isolated Patience.

[Seth Stohs]

Obviously my previous post was facetious, but I really get sick of hearing about Mauer's career BA. The guy is big and strong he should be hitting for more power. Remember 09? Where did that go?

 

Mauer is a terrific BA guy, and that's not a bad thing. he's an even bigger OBP guy, which is more important. And, where did '09 go? It appears he's approaching those levels again in 2013 in a bigger ballpark.

[Brock Beauchamp]

Obviously my previous post was facetious, but I really get sick of hearing about Mauer's career BA. The guy is big and strong he should be hitting for more power. Remember 09? Where did that go?

 

A Google search reveals that the Twins no longer play home games in the Metrodome.

 

Who knew?

[edavis0308]

I've seen it mentioned a few times already, but it really is too bad that stolen bases cant carry any weight in OPS. You could have a .650 OPS but have 80SB a year and have inferior players look better than you by numbers alone.

[ThePuck]

I've seen it mentioned a few times already, but it really is too bad that stolen bases cant carry any weight in OPS. You could have a .650 OPS but have 80SB a year and have inferior players look better than you by numbers alone.

 

Guess that depends on how you value stolen bases

[jay]

I've seen it mentioned a few times already, but it really is too bad that stolen bases cant carry any weight in OPS. You could have a .650 OPS but have 80SB a year and have inferior players look better than you by numbers alone.

 

I agree. OPS is known to undervalue OBP to start with, but SBs compound it further. Even wOBA, an excellent stat, doesn't account for value added by a SB.

 

Since a SB essentially adds a base to the original hit, I wonder what it would look like if we added (SB-CS) to the SLG portion?

[Brock Beauchamp]

I agree. OPS is known to undervalue OBP to start with, but SBs compound it further. Even wOBA, an excellent stat, doesn't account for value added by a SB.

 

Since a SB essentially adds a base to the original hit, I wonder what it would look like if we added (SB-CS) to the SLG portion?

 

You'd want to do more than just add (SB-CS). If no one is on base when the hit/walk occurs, a later CS completely erases any positive effect of the hit/walk so OBP should also be adjusted. If someone is on base when the hit/walk occurred and there is a later CS, you'll need to find some way to measure the positive effect of moving the runner forward but erasing the actual "on-base" benefit of the hit/walk.

[jay]

You'd want to do more than just add (SB-CS). If no one is on base when the hit/walk occurs, a later CS completely erases any positive effect of the hit/walk so OBP should also be adjusted. If someone is on base when the hit/walk occurred and there is a later CS, you'll need to find some way to measure the positive effect of moving the runner forward but erasing the actual "on-base" benefit of the hit/walk.

 

You're right. The first situation would seem to happen a lot more often than the second. To keep it relatively simple, what about (SB-2*CS)? This would remove credit for the original hit/walk?

[Don't Feed the Greed Guy]

Okay, what's the best metric for compiling offensive and defensive statistics? WAR? And if so, I understand that Baseball Reference and Fangraphs have different ways of calculating WAR. Are there other metrics that do this?

[jay]

Okay' date=' what's the best metric for compiling offensive and defensive statistics? WAR? And if so, I understand that Baseball Reference and Fangraphs have different ways of calculating WAR. Are there other metrics that do this?[/quote']

 

I think wRC+ and oWAR are the best offensive stats. They both capture all aspects of offensive value including the bat, baserunning, and SBs. As for the defense... well, I don't have a great answer for that one.

 

The difference in WAR (wins above replacement) between FG and BR is based off differences in their definitions of replacement value. Not many other stats that I know of have this influence.

[ashbury]

I've seen it mentioned a few times already, but it really is too bad that stolen bases cant carry any weight in OPS. You could have a .650 OPS but have 80SB a year and have inferior players look better than you by numbers alone.

 

If you include the huge run-killing effect of a CS, I suspect that the overall effect won't move the needle very much for most players, and isn't worth the effort for the casual fan.

[ashbury]

So whether you count the hits once or twice, it's about the same, although it might be just a little better to count the hits twice, for whatever reason.

 

I think "for whatever reason" has been well enough expressed in this thread by others.

 

If anything, I'd be curious to see a correlation run on OPS+BA. It comes from my years of experience trying something I thought should work but didn't, so try the opposite. I suspect this has been tried before, however.

 

OPS is an apples plus oranges shorthand to get a read on a player's season, so I don't see a lot of incentive to try to patch it. Other approaches to evaluating offensive results would be better if you really want to be serious.

[ashbury]

A band box in thin air helps

 

Thin air, check. Band box, not so much:

 

Clem's Baseball ~ Stadium dimensions

 

Coors is actually pretty large all the way around the outfield. You have to hark back to long-ago stadiums to find larger dimensions.

[edavis0308]

If you include the huge run-killing effect of a CS, I suspect that the overall effect won't move the needle very much for most players, and isn't worth the effort for the casual fan.

 

 

I'm not sure I buy that. What about this year, Nate McClouth has 24SB compared to 4CS. Are the 4CS killing the positive value of his 24SB?

 

Revere last year would be a prime example I suppose, as he stole 40 and had 9CS with a .675OPS in ~550PA. Extrapolate his numbers for a full year of ~650PA and I bet he easily has 50SB with roughly 12CS.

 

It's practically impossible to actually account for the steals and caught stealing in the math, I get that, and it really is too bad. Further evidence of this - what about people like Span that get picked off at a high rate?

 

EDIT: Actually, Everth Cabrera's year last year - .648OPS in 449PA. 44SB to 4CS. That might be a better example then Revere.

[Winston Smith]

Maybe the Twins should have a free abacus day at the park for stat geeks?

http://t0.gstatic.com/images?q=tbn:ANd9GcTh_b3zGuRSBPGOsjlSH4JTRLwIgReeYInOejHJUTQruegqunFE

[twinstalker]

Not to be glib, but you're taking one flawed, non-sensical measure, doing something non-sensical to it, creating another flawed, non-sensical measure, and comparing them. Ugh.

 

Both are obviously non-sensical because they involve adding together two percentages that are based on different denominators. A measure that makes sense is one with one denominator, probably PAs. Then proper weights need to be assigned to each outcome of each PA. In the end, what you'll get is a percentage that is probably just a standardization of a runs created measure.

 

For example, just intuitively I used to rate players (pre-card) to determine who would have the better strat-o-matic card for our drafts my league conducted (again, prior to the cards coming out). I assigned a value to a walk, hbp, single, double, triple, hr and used plate appearance as the denominator. In slg, you assign weights 1, 2, 3, and 4 with ABs as the denominator, so I had to determine what I thought everything was worth and usually did so with a combination of intuition and some simple regression. I gave singles a weight of 1 and changed or created new weights for the other outcomes.

 

What was nice about my measure was that the distribution was relative close, pct-wise, to slug, so I inherently understood the numbers. I guarantee I drafted better than all the guys using OPS (a non-sensical measure I still am not comfortable understanding what's good what's not good--for good reason).

 

FYI, for those who don't know, strato is a retrospective game, so this has nothing to predicting the future (nor does any of what Seth discusses).

[edavis0308]

Not to be glib, but you're taking one flawed, non-sensical measure, doing something non-sensical to it, creating another flawed, non-sensical measure, and comparing them. Ugh.

 

Both are obviously non-sensical because they involve adding together two percentages that are based on different denominators. A measure that makes sense is one with one denominator, probably PAs. Then proper weights need to be assigned to each outcome of each PA. In the end, what you'll get is a percentage that is probably just a standardization of a runs created measure.

 

For example, just intuitively I used to rate players (pre-card) to determine who would have the better strat-o-matic card for our drafts my league conducted (again, prior to the cards coming out). I assigned a value to a walk, hbp, single, double, triple, hr and used plate appearance as the denominator. In slg, you assign weights 1, 2, 3, and 4 with ABs as the denominator, so I had to determine what I thought everything was worth and usually did so with a combination of intuition and some simple regression. I gave singles a weight of 1 and changed or created new weights for the other outcomes.

 

What was nice about my measure was that the distribution was relative close, pct-wise, to slug, so I inherently understood the numbers. I guarantee I drafted better than all the guys using OPS (a non-sensical measure I still am not comfortable understanding what's good what's not good--for good reason).

 

FYI, for those who don't know, strato is a retrospective game, so this has nothing to predicting the future (nor does any of what Seth discusses).

 

I actually jumped into this because the conversation I feel is a nice change of pace from everything. Conceptually, I do find this stuff very interesting. (side note, I just grabbed PA in my numbers. I could have grabbed AB but I didnt, so shame on me perhaps).

 

Oh, and if we can't discuss combining two non-sensical measures, then what on Earth ARE we supposed to discuss?

[Brock Beauchamp]

Both are obviously non-sensical because they involve adding together two percentages that are based on different denominators. A measure that makes sense is one with one denominator, probably PAs. Then proper weights need to be assigned to each outcome of each PA. In the end, what you'll get is a percentage that is probably just a standardization of a runs created measure.

 

This has always irritated the hell out of me regarding OPS. The OBP portion is on a 1.000 scale while the SLG portion is on a 4.000 scale.

 

And not only do they use different scales but then some of the data from each is friggin' duplicated using these different scales. MRARGH.

 

It's some pretty sloppy-ass math. With that said, it's pretty amazing how closely it aligns with runs scored and just how good a measurement it is when you're talking on a team or league level.

[jay]

I'm not sure I buy that. What about this year, Nate McClouth has 24SB compared to 4CS. Are the 4CS killing the positive value of his 24SB?

 

Revere last year would be a prime example I suppose, as he stole 40 and had 9CS with a .675OPS in ~550PA. Extrapolate his numbers for a full year of ~650PA and I bet he easily has 50SB with roughly 12CS.

 

It's practically impossible to actually account for the steals and caught stealing in the math, I get that, and it really is too bad. Further evidence of this - what about people like Span that get picked off at a high rate?

 

EDIT: Actually, Everth Cabrera's year last year - .648OPS in 449PA. 44SB to 4CS. That might be a better example then Revere.

 

I pulled 2012 offensive stats from fangraphs (>150 AB) and threw a spreadsheet together to add in the (SB-2*CS) I proposed earlier. The pool was 401 players. I just added it in to SLG, but it could just as well go in OBP or be calculated separately. Separate might be best for connotation purposes, something like OSG (On-base, speed, and slugging)?

 

Guys like Revere and E Cabrera benefit, along with some guy named Mike Trout that went 49SB/5CS. Interestingly, it doesn't benefit all speed guys. Starlin Castro saw a decrease compared to OPS (25SB/13CS). Jose Tabata got hit the hardest (8SB/12CS).

 

I have to do some actual work for a bit and can't determine if the run correlation is better or worse, but anyone else is welcomed.

 

https://docs.google.com/file/d/0B5Gl53bhfjnPYTFuVENpYVlsMG8/edit?usp=sharing

[Oxtung]

My biggest problem with all of these stats is that the correlation is always from a team perspective. People then take that team correlation and make the assumption that it is true for the individual. That is a logical fallacy. A simple example is this:

 

2+2=4. We know both the addends (the 2's in this case) and therefore we know the sum (the 4). However this does not work in reverse. 4=???? It could be 0+4 or 1+3 or 2+2. Having a sum tells you little about the individual addends.

 

So when you add many teams worth of data together (the sum) for a correlation study that number, regardless of the actual correlation, tells you nothing about the players themselves (the addends).

 

So applying this to players, how can one say that OBP is worth more than SLG?

 

I have another thought brewing in my head but haven't played with it enough...I'll post more later.

[Brock Beauchamp]

I pulled 2012 offensive stats from fangraphs (>150 AB) and threw a spreadsheet together to add in the (SB-2*CS) I proposed earlier. The pool was 401 players. I just added it in to SLG, but it could just as well go in OBP or be calculated separately. Separate might be best for connotation purposes, something like OSG (On-base, speed, and slugging)?

 

Guys like Revere and E Cabrera benefit, along with some guy named Mike Trout that went 49SB/5CS. Interestingly, it doesn't benefit all speed guys. Starlin Castro saw a decrease compared to OPS (25SB/13CS). Jose Tabata got hit the hardest (8SB/12CS).

 

I have to do some actual work for a bit and can't determine if the run correlation is better or worse, but anyone else is welcomed.

 

https://docs.google.com/file/d/0B5Gl53bhfjnPYTFuVENpYVlsMG8/edit?usp=sharing

 

Sounds to me like you're on the right track. I remember Beane saying years ago that if a player isn't successful in ~70% of his SB attempts, he's actually hurting the team.

[ThePuck]

So applying this to players, how can one say that OBP is worth more than SLG?

 

 

Because, as was said more eloquently earlier by Brock, not making an out is the most important thing. Outs are precious commodities.

[LaBombo]

Looking only at OPS, Dustin Pedroia and Ian Desmond look like virtually the same player (.812 to .809). Except that Pedroia is getting on base nearly 40% of the time while Desmond is getting on base less than 32% of the time. I think that pretty much everybody would take Pedroia's bat over Desmond, and for good reason. It's better.

 

Which is not evident by OPS+, where Pedroia trails Desmond 119 to 121 but quite evident in oWAR, where Pedroia leads Desmond 2.6 to 2.2.

 

+1. Weighted on base pct (wOBA) is finally showing up outside of fangraphs once in a while, and it gives Pedroia a slightly bigger lead over Desmond, .357 to .347. It was a bit surprising that it didn't give Pedroia a bigger edge, but then again its correlation to runs scored is only a few percent higher than OPS.

[LaBombo]

Sounds to me like you're on the right track. I remember Beane saying years ago that if a player isn't successful in ~70% of his SB attempts, he's actually hurting the team.

Interesting that it's been a moving target over the years. Bill James threw out the math-friendly yardstick that SB attempts must succeed twice as often as they fail (66%) to have value. Then 70% became the unofficial threshold in the Beane era. Lately at fangraphs there have been some studies that suggest the real break-even may actually be closer to 75%.

 

Those numbers are general ones that don't reflect which base is being stolen or how many outs there are. One other thing they don't reflect is that pitchers supposedly pitch less effectively with base stealers "creating havoc" (tying infielders' shoelaces together? spraypainting graffiti on their uniforms?). But Stats inc. did a study suggesting that the distraction of straight steal attempts did more to hurt the performance of the batter than that of the pitcher.

[Oxtung]

Because, as was said more eloquently earlier by Brock, not making an out is the most important thing. Outs are precious commodities.

 

I don't think you understood my point. Brock's assuming that what is good for the team is also good for the individual. That leap is not logical as demonstrated by my simple addition example. We know that for a team OBP is slightly more important than SLG when it comes to run creation but that tells us nothing about what is more important from individual hitters.

[Willihammer]

I don't think you understood my point. Brock's assuming that what is good for the team is also good for the individual. That leap is not logical as demonstrated by my simple addition example. We know that for a team OBP is slightly more important than SLG when it comes to run creation but that tells us nothing about what is more important from individual hitters.

 

Right. wOBA judges every batter by the same weighted averages but team scoring environments aren't all identical. A walk in a low OBP environment isn't as valuable as in a high OBP environment (although still valuable). Likewise a home run in a low OBP enviroment is more valuable on average, than one in a high OBP environment. The Blue Jays are built this way.

 

The Red Sox are excellent in every respect, so Pedroia's OBP contribution, still very valuable, might not be as valuable to his team as Desmond's HRs are to the Nats who are pretty terrible offensively in every respect.

[Oxtung]

Right. wOBA judges every batter by the same weighted averages but team scoring environments aren't all identical. A walk in a low OBP environment isn't as valuable as in a high OBP environment (although still valuable). Likewise a home run in a low OBP enviroment is more valuable on average, than one in a high OBP environment. The Blue Jays are built this way.

 

The Red Sox are excellent in every respect, so Pedroia's OBP contribution, still very valuable, might not be as valuable to his team as Desmond's HRs are to the Nats who are pretty terrible offensively in every respect.

 

It reads like you're using data to backup your assertions on OBP vs. SLG w/r to lineup creation. I would love to read the article if you have a link to it because my intuition says it should be the other way around. I would guess on a team with many high OBP but low SLG guys a HR hitter would be more valuable then another OBP guy because his HR's will drive runners in more efficiently since there are many people on base in front of him, especially 1st base if the team is drawing many walks. On the flip side if a team is filled with sluggers a guy who gets on base would be very valuable because he would then be able to get driven in with a very high frequency thereby also increasing the value of the sluggers behind him. I have no data to back this up, just my intuition so if you have that article I would love to read it.

 

Regardless of which way it turns out, our larger points are the same. A players value is largely determined by the lineup he is a part of. It's easy to say Mauer is a better hitter than Clete Thomas but it is much more difficult to tell the difference between 2012 Mauer and 2012 Willingham. Who the better player is largely has to do with what the Twins needed more. Baserunners or Power hitters.

[Oxtung]

Thinking more about this, and just brainstorming here as I type, I wonder if anyone has done a study on the OBP/SLG (or it could be ISO or any power metric) ratio to runs scored.

[ashbury]

This has always irritated the hell out of me regarding OPS. The OBP portion is on a 1.000 scale while the SLG portion is on a 4.000 scale.

 

And it's even worse than that: simply scaling SLG down by 4X doesn't help because the typical value is not 4X the typical OPS. Combining the two values is a doomed exercise.

 

OPS is just a quick-and-dirty, fun, number.