Tangotiger Blog

If you take out ROE, shouldn’t you also take out the corresponding number of PA’s from the denominator? For a fair amount of players, ROE can account for 4/5+ runs per year, yet by not taking those out of the PA in the denominator, you’re turning those runs into outs.

(This is my biggest issue with Fangraphs wOBA.)

If you don’t know the number of ROE, how can you take it out of the denominator?

Oh, I totally misunderstood something, my bad.

I don’t think using a single weight for 2b and triples is a good idea. No need to simplify that much. People see that, I think it will turn them off. IMO.

Do people still think about total bases?  What about:

(5*H+4*TB+7*BB)/(10*PA)

David: that’s a good point. 

I actually use the non-SB version, which is why I lump 2B+3B together, reasoning that 3B are mostly “speed doubles” or “park/OF-influenced”.

However, if I’m going to include SB, CS, then you are correct, then I should split up 2B (1.25) from 3B (1.60).

So, that’s what I would suggest: if you don’t include SB, CS, then lump 2B+3B (1.30).  If you do include SB, CS, then split up 2B+3B.

***

Jared: the point of wOBA is to make clear what each event does, so that a lay person can look at it and think “ok, a HR in 2 PA is just a bit more valuable than 2 singles in 2 PA”.

With the H, and TB and BB combination, you have to sit down and think about what it’s doing.  It’s why those combinations have never taken off.  Dave (and studes) both have their own version of such combination, and they are both good.  But, they don’t take off simply because you have to sit and think a bit too much.

Trying to remember 0.7, 0.9, 1.3, 2.0 as the coefficients in wOBA is straightforward, and transparent.

I can say we got to this point because we’ve tried all these other combinations, and the wOBA structure is what resonates.

***

Personally, the LWTS structure is what should really resonate, but many people find it impossible to get past the “0 = average” scale.  And they see a negative as… well, as if it’s some sort of deficit or something.  But, whatever, I don’t want to talk about that.

I can say that since wOBA *is* linear weights, no one has EVER said anything about a hitter having a .300 wOBA, even though that’s no different than being -15 runs in 600 PA.

See, what wOBA cleverly does is maintain the two dimensions: the rate portion, and the quantity (PA) portion, much like ERA and IP keep that separation intact.

Once you combine those two dimension into one (be it in terms of “above average” or “above replacement”), suddenly, that’s when all the problems come in.

I updated the standard so that we have two standards, much like Bill James had the standard version with and without the SB.

Not sure if this is significant at all, but I figured I’d post it here. As part of other research, I ran year-to-year correlations on FanGraphs’ wOBA and the above Standard wOBA (not including SB/CS, but with the 2B/3B distinction) for 2000-2012, min 400 PAs, and got .567 for FG and .585 for standard. This is somewhat surprising to me, since I’d think the customized weights would make wOBA more stable year to year, not less. I’m sure there’s some explanation, but I figured you all would know better than me.

Matt: Fangraphs wOBA includes SB/CS I think, so you’d have to include that.  Even so, I am shocked at the low correlation you are reporting.

Fangraphs removed SB/CS from wOBA, wRAA, and wRC+ last year.

http://www.fangraphs.com/blogs/woba-base-running-changes/

I tried to replicate Matt’s sample, but I got .56 for both wOBAs. If I add in ROE to std_wOBA using Retrosheet I get .55.

I double checked my method against Bill’s here:

http://www.beyondtheboxscore.com/2011/9/1/2393318/what-hitting-metrics-are-consistent-year-to-year

...and got the same .61 that he did for FG wOBA for 2001-2008 min 300 PA (even though Fangraphs’ wOBA formula has changed since that study.)

So I checked only the 2009-2012 sample, and found that yoy correlation for both wOBAs were at or below .5 in those recent seasons.

Ok, I apparently misread Matt’s sample as 2002-2012 (I think because I use that period for everything I do).

Running 2000-2012 I get .59 for both.

Strange. I’m sure I messed something up when I ran it. Thanks James.

I misread what Matt was doing.  I thought he was correlating Fangraphs wOBA against Standard wOBA.  I now see that he was correlating year-to-year wOBA.  So, ignore my prior comment.

Using this:
http://tangotiger.com/index.php/site/article/linear-weight-run-values-2010-2015

These are the wOBA values for 2010-2015:

2.06 HR
1.63 3B
1.26 2B
0.96 SAFE
0.89 1B
0.73 HBP
0.71 NIBB
-  Out
-  K

***

If you merge 2B+3B, and merge SAFE+1B, we get:
1.30 2B+3B
0.89 1B+SAFE

For my personal processing, I go with Standard wOBA:
0.7 BB+HB
0.9 1B+SAFE
1.3 2B+3B
2.0 HR

I know some people like to break out 2B and 3B, but for the most part, it’s not worth it.  I like having simple rules of thumb, and the above scale satisfies that.

Asked this on BJOL, but thought I’d add it here…

Is there a value for wOBAScale that you would advocate be used in conjunction with the two versions?

Yes, 1.2 works well.

Assume the league OBP (and wOBA) is .320.  So, to convert the 2.0 HR into runs:
2.0 minus .320, all divided by 1.2 gives you 1.4

And 1.4 is what we want.  You can try it with singles as well:
0.9 minus .320, all divided by 1.2 gives you 0.48

That’s close to what we want.  The walk as well:
0.7 minus .320, all divided by 1.2 gives you 0.32

The out naturally is -.320/1.2 = -.27