THE BOOK--Playing The Percentages In Baseball

Maybe I missed something in that thread...how could OPS+ now be better than wRC+? Or are you just saying they’d be [more] comparable (if Sean makes that change)?

I’d like to see a proof, but only for curiosity’s sake.

Did I say that?  Can you quote me please?

I’m saying that if Sean makes the change, they’d be a worthy challenger.  Actually, I’m saying exactly what I said.  I’m not sure where the ambiguity lies.

Channeling Fermat with this proof stuff here?

Looks like wRC+ is adjusted for playing time after seeng Drews numbers (137) relative to Bays(142). 

If so, I would prefer to see it on same scale as wRC and not adjusted for playing time, so we can see how many runs were added or subtracted due to park effects.  You can not create runs if you don’t play.  Runs are cumulative and a counting stat, while OPS is the product of 2 rate stats and is unaffected by playing time.

ALL of the “+” families are indexed so that average is 100.

If you want to get specific, then it’s “wRC/PA+”.  Do we REALLY need to see that in the column header?  Can’t I define the non-Merriam-Webster term “wRC+” any way I like to?

I love that FanGraphs implemented this so quickly.

I think there is one difference in the construction of OPS+ and wRC+ that still gives OPS+ some value despite its inferior core, which is that OPS+ removes pitchers’ hitting from consideration.  As wRC+ stands, NL and pre-DH AL hitters have an advantage over modern AL hitters in that the league R/PA will be higher in the modern AL just because of the DH.  A hitter who moves to the AL will suffer in wRC+ without even considering the difference in league quality just because he’s not being compared against so many pitchers anymore.  If you only want to know how a hitter compares to other non-pitchers in his league, and not how NL hitters compare to the league with all the pitchers thrown in to drive wRC+ up a bit, OPS+ might still provide something wRC+ doesn’t.

Of course, this would probably kill the simplicity of calculating wRC+ that made it so easy to implement.

Does wRC+ include SB and CS because wOBA already incorporates them, or did Dave decide to add them to the calculation?

Also, it would be nice if the Fangraphs glossary included explanations for this and some of the other non-obvious advanced hitter metrics.

Kincaid, I’ve been thinking about this same problem.  It would be easy enough to exclude pitcher’s batting from league averages.  This is done for the wOBA weights (honestly doesn’t make much of a difference there), but wasn’t sure how I should handle the league averages, so they’re simply just that, averages.

When I do my stuff, I not only ALWAYS remove pitchers-as-hitters, but I also remove hitters-as-pitchers, be it for hitter, pitcher, or fielder evaluation.

Tom/#3

The ambiguity (to me) lies here: “then OPS+ would be a worthy challenger.” To me, “worthy challenger” means that OPS+ might be better than wRC+ some of the time*. If you didn’t mean that, but instead meant that OPS+ would just be much closer than it is now (if Sean changes it), then ignore my first comment.

*To illustrate...In terms of pure speed, a go-kart is not a worthy challenger to a Maserati, but Tyson Gay is a worthy challenger to Usain Bolt.

I’m going to look into removing pitcher’s from the league averages, because it’s something I should be doing, but I honestly it’s going to be a very small difference.  Something like 2, maybe 3 more wRC+

Pitchers-as-batters account for something like 6% of all NL batting, IIRC.

We can think of it this way:
-nonpitchers 0 runs per 700 PA
-pitchers-as-batters -100 runs per 700 PA

If you take 94% of the first and 6% of the second, you get -6 runs per 700 PA.  If a hitter is 86 RC, compared to the league average of 80, when he should be compared to 80, that would mean 108 instead of 100.

I haven’t worked it out if I’m empirically correct, but that would be my first presumption.  You would get a similar impact when comparing the ERA to the league average in the DH and non-DH league.

So, I would say that David is wrong about the degree of impact, but I could be wrong.

"when he should be compared to 80"… that should be 86.

I suspect that both NL and AL are currently combined to calculate an average player’s RC.  If so, then the degree of impact would be close to David’s estimate of 3 runs.

I’m sure I’m not the only one who is uncomfortable with the inclusion of context neutral SB and CS run values in wOBA.  When I see that 45% of Braun’s SB’s last year were stealing 3B (and more than a few with 2 out), I get a little concerned about the “stat padders”.

Does anyone publish aggregate RA, WPA and LI data for stolen base attempts?  I’ve seen a few people manually add it up for their favorite team but never a leaderboard for the league.

Tango, you’re right that it’s more than 2 or 3, I quickly calculated something wrong.  I’m seeing about a .008 yearly difference in R/PA with pitcher’s excluded, which ends up being pretty much what you did.

Peter does bring up a good point that when I’m subtracting wOBA to get to wRAA, I’m just using the yearly wOBA average. 

The leagues make running some of these calculations very annoying....

For those keeping track, I’ve just adjusted the league R/PA to exclude pitchers’ batting.

Instead of a “proof”, can you demonstrate which players were or are undervalued by OPS+ as-is (from 2000-) that are opportunities revealed by wRC+ or modified OPS+?

There are two kinds of players that are undervalued by OPS+ that are properly valued by RC+: guys with lots of walks, and guys with lots of steals.

As an example, let’s take a look at Rickey Henderson’s career RC+ and career OPS+:
145
127

How about guys with lots of walks but few steals?  Wade Boggs sounds like our man:
135
130

And guys with lots of steal?  How about Vince Coleman:
99
83

If on the other hand Sean implemented the 1.2 / 0.8 weighting as I suggest, Wade Boggs’ OPS+ comes in at:
415/337*1.2 + 443/414*0.8 - 1 = 133.4
As opposed to 130.2

It comes into play when you see guys bunched at 130, and really it should be 133 for some guys and 127 for the other guys.

Say for example Jim Rice:
352/337*1.2 + 502/407*0.8 - 1 = 124.4
As opposed to 127.8

So, where once you might think that Boggs’ 130 and Rice’s 128 are “close”, the real numbers show it as 133 for Boggs and 124 for Rice.  Quite a larger difference.

And RC+ shows them as: 135, 132, in part because Boggs was a terrible base-stealer.

In any case, we know how to do it right.  There is exactly one person in the world that calculates OPS+.  I don’t think there’s anybody out there that’s actually calculating OPS+ by themselves, is there?  And I’m saying Sean should calculate it the right way.  He’s doing it that way because Pete Palmer did it that way.  It’s a nice approximation.

But, given how much OPS+ is used, and how much people are trusting Sean, he should get it right.

Tango, what are the weights for SB and CS in wOBA? 

And why did you choose to include them in wOBA, rather than treating baserunning separately?  (Don’t necessarily disagree, just curious on your thinking.)

wOBA weights for all events by year:

http://tangotiger.net/bdb/lwts_woba_for_bdb.txt

Explanation of how weights are derived:

http://www.insidethebook.com/ee/index.php/site/comments/woba_year_by_year_calculations/

Actually, Fangraphs is David’s site.  I don’t really mind either way whether wOBA has SB or not, as long as there’s a reason to have it or not.  In The Book, since we were dealing with batter/pitcher matchups, there was no need to bring in SB in there.

Including just SB and CS, with those weights, will substantially undervalue the baserunning value of basestealers.  To highlight an extreme example, Brett Butler essentially gets no credit for his basestealing (+3 for career), while according to Rally’s data he was +76 on the bases.  I suppose the argument is that it’s better to give basestealers some credit—at least the high percentage guys—as opposed to no credit at all (OPS+).  On the other hand, I think people know that baserunning isn’t included in OPS+, while this kind of muddies the waters by giving partial credit.

And ROE is not included in wOBA?  Not a huge deal, but that surprised me.

My version of wOBA includes it.  If you are referring to the link, that was based SOLELY on the idea of creating weights for data in the Baseball Databank database.

Palmer and James’ LWTS, RC includes basestealing but not base-advancing.  I don’t think it’s confusing on that level.

I also find that SB-CS to be a decent approximator of basestealing + base-advancing.

That is, SB-2CS is your basestealing metric, while SB is akin to the base-advancing metric.  Add the two together,and you get 2*(SB-CS), or simply SB-CS.