THE BOOK--Playing The Percentages In Baseball

---"The really big difference originates with FIP v RA9.  Rivera is a BABIP machine. It’s a huge difference. “
************

Which is why I have little use for fWAR for pitchers.

Note my last sentence.  rWAR at the career level makes much more sense than fWAR.

You can make the argument that you can prefer fWAR at the seasonal level and rWAR at the career level.

---"You can make the argument that you can prefer fWAR at the seasonal level and rWAR at the career level.
**************

Not me, even for seasonal, looking backwards. I’d rather they do something as simple as divide credit for BABIP 50/50 between pitchers and fielders, then give all the credit (in effect) to the defense.

I definitely don’t disagree with you.

There’s 4 aspects to pitching that leads to runs:
- non-fielding PA
- fielding PA
- running game non-PA (which you can break into pre-PA and post-PA)
- sequencing

They all involve the pitcher’s performance to some extent.

The (current) philosophy at Fangraphs (and really all the sites) is that you have a 0 or 100 approach to each aspect.

Since the fielding PA seems to involve the pitcher mostly as incidental, then what’s the point of using the hits allowed when he’s on the mound?  That’s one philosophy.  The other philosophy is that: well, he’s on the mound, and even if it’s T-ball, I’m going to link it to that particular T-ball tee.  So, you have a political 0/100 choice to make.

David would advocate for something not so extreme, based on his above comments.  To make it 50/50, or really, something other than 0/100 to acknowledge this fact.

So, it requires a philosophical shift.

More disturbing however is that SB, CS, PB, WP, BK, PK are all ignored.  Again, you can definitely make it 50/50 with the catcher.

The sequencing of events is again a similar issue.

On the other hand, Rally with rWAR has taken one philosophical stand on it by focusing on runs allowed, which encapsulates everything the pitcher is involved in (and giving him 100% credit after making some fielding adjustments).  And fWAR has taken the polar extreme to the other side.

I can really appreciate these two viewpoints on its own, and the reader is then free to weight each one to his heart’s content.  My preference, because I’ve thought about this way too much, and have decided that I need way more time to think about it, is to split the difference.

It’s like the 4 Game Scores I introduced.  Each reader can merge them as he wants.

So, that’s why I can really appreciate that fWAR and rWAR takes two polar views on the matter.

I thought Fangraphs didn’t include LI.  I don’t see it mentioned in the description, but I’m not expert on this stuff.

studes: let’s see if we can infer it easily enough.

This is their leaderboards for 1972-2011, for relief pitchers:

http://www.fangraphs.com/leaders.aspx?pos=all&stats=rel&lg=all&qual=y&type=8&season=2011&month=0&season1=1972&ind=0

Rivera at 38, Gossage at 30, Lee Smith at 28, Wagner at 25.  That’s an average of 30.

If we look at their FIP-, Rivera is 60, Gossage is 79, Lee Smith is 73, Wagner is 63.  That’s an average of 69% of league average.

Replacement level is about 107% of league average (not sure what Fangraphs actually uses… 107% is my number).

Those 4 pitchers averaged 1200 IP.  So, if we presume 4.5 runs per 9 IP as the league average, we get:

(107% - 69%) x 4.5/9 * 1200 = +228 runs above replacement

That’s 23 WAR, with no leverage.

Since their WAR is actually 30, that means a leverage aspect of 1.30 was applied.

So, SOME leverage was applied.  It might be too low.

But, this is a quick look as to how you can figure it out, if you can find the actual statement.

In any case, the reader out there can use this as a lesson as to how to calculate WAR.

Let me talk a bit about replacement level for SP and RP.

For a league average of 4.70 RA9, the replacement level is 5.00 for relief and 6.00 for starters.

However, if you rely solely on FIP, we have to remember that relief pitchers have a lower FIP than their RA9.  That’s because relief pitchers have a lower BABIP than starters.

I haven’t looked to see how much of an effect that is.  Perhaps the repl level, if you use only FIP, might be 4.80 for relievers and 6.10 for starters.  This would have to be studied.

So, this calculation:
“(107% - 69%) x 4.5/9 * 1200 = +228 runs above replacement “

Can very well be:
(102% - 69%) x 4.5/9 * 1200 = +198 runs above replacement

With that being 20 wins, and seeing a WAR of 30 wins, then a leverage of 1.5 gets you from unleveraged 20 wins to leveraged 30 wins.

So, it’s pretty sensitive to the replacement level.

That’s because relief pitchers have a lower BABIP than starters.

tango, I don’t believe that’s true for all relief pitchers.  I found it to be true for elite relievers, but not all.

studes:

#1: great starters have a lower BABIP than bad starters

#2: great relievers have a lower BABIP than bad relievers

#3: the same pitcher, pitching in relief, has a 17 point drop in BABIP than if they were pitching as starters

So, follow me here… let’s say you have a bad starter, his BABIP is .317… you put him in relief, and, his BABIP becomes .300. 

Therefore, it looks like bad relievers aren’t affected, but in fact, they were hugely affected.

It uses LI, yeah. Appelman and I discussed it on Twitter and I built this spreadsheet to calculate Fangraphs’ WAR values for relivers:

http://www.editgrid.com/user/cwyers/rivera_walkthrough

That’s Rivera’s career numbers - because I used career totals (instead of going by season) and omitted park adjustements it’s not QUITE there, but you can get the sense of what they’re doing.

And Studes is right - elite relievers have a lower BABIP, perhaps, but as a population BABIP is the same for starters and relievers.

Colin must have cross-posted, because this is only true because we don’t have the same population:

but as a population BABIP is the same for starters and relievers

You have way more bad pitchers in relief than you have bad pitchers as starters.  But, pitching in relief confers an advantage to your BABIP (and HR and K).

I don’t have my data handy, but let’s look at 2010:
http://www.baseball-reference.com/leagues/split.cgi?t=p&year=2010&lg=MLB#sprel

Relievers have 16.1 HR per 700 PA, while starters have 18.0 per 700 PA.  What this masks is that pitching in relief reduces your HR rate.  So, their talent level is to have a higher than 18 HR per 700 PA, but by pitching in relief they end up giving up only 16.1 HR.  It makes it seem as if there is no difference.

Same deal with BABIP.  In 2010, relievers were .297 and starters were .298.  But that’s only because relievers are given a huge advantage.

The quick analogy is that pitching in relief is like pitching in Oakland or San Diego, and pitching as a starter is like pitching at Coors.  If you just look at their stats, it looks like the relief and starter population gave up the same number of runs.  Well, they did, but only because of their circumstances.

OK, but BABIP isn’t the only thing that improves when a pitcher moves to the bullpen.  K rates do, too.  I forget about BB and HR rates.

Overall, FIP is about the same as ERA for all relievers, probably because many of them would never be starters in the majors in the first place.

Studes: google on this blog

Rule of 17

So...if a nameless website uses only FIP, their replacement level differential ought to be greater than it is for ERA?

It’s here:

http://www.insidethebook.com/ee/index.php/site/comments/starter_v_relief_1953_2008/

Tom, you said:

I know about the selection bias.  It was discussed at length in The Book.  (For those who don’t have it, read it for free on Amazon.com.) The next step is to do what I did in The Book, for all these pitchers.  I’d bet you’ll find results that match those.

I’m betting you won’t. Did you ever do the study where you control for selection bias?

Rule of 17

Thanks for the reminder.

Colin: I responded in the other thread.

Thanks for this post. Every once in awhile someone asks me what the difference is between fWAR and rWAR for pitchers. Slightly less frequently, someone asks me why I favor rWAR for pitchers.

Far less frequently, someone makes an argument that Mo’s not as good as we think he is based on fWAR and I feel the need to engage him/her.

I’m happy that I now have this for reference.