THE BOOK--Playing The Percentages In Baseball

This doesn’t mean much, but I once spent a while browsing through my old original 1969 Baseball Encyclopedia, which has separate records for pitchers as starters vs. relievers.  Pretty interesting to see howq Lefty Grove, etc. did as relievers.  But I’d say that the average difference was less than half an earned run; maybe .3.

Of course, the circumstances were very different back then as most relivers were really starters slumming it on the side.  The Encyclopedia doesn’t have comparative strikeout rates.

Perhaps a dumb question, but what would you say was the cause of the increased K rate as a reliever?  Being able to put a little extra “oomph” on the fastball, since he only had to go an inning or two?  Only being seen once by each batter?

In the NL, a starter gets to pitch against pitchers for a PA or two each game; relievers really don’t.  All else being equal, this would point to starters having a bit better K rate, ERA, and the rest.  Has there ever been any studies as to why pitchers do better as relievers?

Smoltz’s BABIPs may have been about the same (what were they?), but my sense is that closers as a group do post better BABIP than starters.  I think DSG and/or Emeigh have found that result.  I’m sure that higher K-rate is the MOST important advantage a reliever has, but not the only one.

* * *

I’m sure the starter/reliever gap was smaller in the past, but mainly because relief outings were longer then (around 2 IP).  In the past 20 years or so the average outing by a closer or top setup man has declined about .5 IP, and that seems to make a real difference in pitchers’ ability to go “all out.” The ERA+ of top relievers has improved dramatically as their outings have become shorter.  I would think that if Tom Seaver, say, had been allowed to pitch in 3- or 4-out intervals, he could have dropped his ERA by a run or more.

I think it was .290 and .287, but it was estimated as (H-HR)/(TBF-BB-HBP-SO-HR).

***

Yes, I have also found the BABIP to be lower for the ace reliever.

***

In THE BOOK, I show how the starter does each time through the order.  I believe the K rate went down each time.

***

Also, in THE BOOK, when I looked at how a pitcher does in the starter/reliever role, it looked pretty clear that when a pitcher was an emergency reliever, that he didn’t show any difference.  I wish I had taken a closer look then, by also looking at the K-rate.  It’s possible that the wOBA (excluding the K as a PA) may have been the same for all starter/reliever breakdowns I did there, and the difference was simply the extra K.

Oh well.  Any takers?  It’s the section called “Dual Roles”, I think, in the starter chapter.

We can also look at Goose Gossage.  His walk rate in the year before he was a starter was 10%.  As a starter it was 9%.  Then, his next two years were 8% and 10%.  Essentially, his walk rate was unchanged.

His HR rate was 1.7% as a starter, and 1.7% in the two years after (and only 0.5% in the year before).  This is HR/PA.

His BABIP was .257 in the year before, .280 as a starter, and .221/.222 after.  (Big difference!)

His K rate was 23% before starter, 14% as starter, then 29/23% after starter.  This change is consistent with what we found with Smoltzie.

How about Tom Gordon?  11% walk rate in his career as a starter, and 9.4% as a reliever.  K rate of 19% as starter, and 28% as reliever.  Again, exactly the same pattern as Gossage and Smoltz.  His BABIP also much better as a reliever (by 35 points).

Eck?  16% K rate as starter and 25% as a reliever.  BABIP exactly the same, but much better walk rate.

Wakefield?  This guy should need no pacing, right?  9% walk rate in both roles.  Exactly the same BABIP rate. K rate as starter/reliever: 16%/21%.

(HR rates similar for all pitchers.)

Terry Mulholland on the other hand… virtually exactly the same rates as both starter and reliever in all categories.  Of course, his role shift has him pitching in late30s/40s as a reliever.  Then again, so did Eck.  But, he’s not a fastball pitcher.

My guess is that it’s all tied to the fastball.  The other guys can “bring it on” in a limited role, while the junk ballers can’t.  Probably why we see so many fastball relievers, since they “improve” their true talent levels, without modifying their skillset, simply by leveraging it based on their usage pattern.

I’d love to see more study on this.

***

I’m trying to remember what I wrote in THE BOOK.  The pitcher as a reliever had a wOBA of .327 as reliever and .355 as a starter?  Something like that.  Like I said, I wish I had included the K rate.  Let’s assume the K rate as a starter was 15%, and as a reliever it was 23%.  If we throw out the extra 8% of PA as a reliever that resulted in a strikeout, and calculated their wOBA on the other 92% of PA, that .327 would become..... .327/.92 = ..... .355.

Now, all we need is proof.

Fascinating stuff!  Tom, I’d love to see what happens if you adjust for the “first time through the order difference” to see what is left.  What is left would be the “I can throw harder for one inning.”

IOW, to answer someone’s question above, there is definitely an advantage that the short releiver has in terms of only seeing batters one time through the order (generally of course, unless they get raked in that one inning).  As Tango explains in The Book, starters have a large advantage the first time through the order (as compared to subsequent times).  We think there is an additional advantage based on the fact that the reliever does not have to pace himself and therefore can throw harder and possibly even better off-speed pitches.  This is actually tied to the “fitst time through the order” thing, as we don’t really know how much of that effect is batter unfamiliarity and how much is a “tiring” effect.

In any case, it would be great if someone could look at starter/reliever differential for more than one class of pitchers, perhaps the hard throwers versus the soft tossers, or perhaps power versus finesse, however you want to define that (Bill James has his own specific definition, something like K+.5*BB rate, or something like that).

Of course, if we looked at TLV (type, location, and velocity) data, we would REALLY get a better idea as to what is going on.  As Tango and I have said repeatedly, there is a wealth of information to be gleaned from the TLV data - a proverbial cherry patch waiting to be picked!

I looked at this subject a couple months ago.  According to the Neyer/James book, Smoltz’s pitch selection changed when he became a reliever, as did the pitch selection of others that changed between starter and closer (ex. Eck dropped his curve).  For example, it says he stopped throwing the splitter in ‘98 due to the stress it put on his arm, but he brought it back when he became a closer and Eckersley stopped throwing his curve when he became a closer.

Gagne almost doubled his K/9 when he became a closer.

Derek Lowe’s numbers look much different in different roles:

* Reliever
1999 K/9 = 6.59
* Closer
2000 K/9 = 7.78
2001 K/9 = 8.05
* Starter
2002 K/9 = 5.20
2003 K/9 = 4.87
2004 K/9 = 5.17

Kodiak: for this analysis, you’d be MUCH better off putting things in terms of per batter rate.  Per “9 innings” really means “% of outs that are strikeouts”.

***

MGL, I touched upon that in the book.  The typical starter would be .340 the first time, .348 the second, and .354 the third.  The overall net effect was something like a .027 point improvement in “I can throw harder, now”.

What is also interesting (also in THE BOOK) is that I didn’t find any advantage for a reliever pitching on short/no rest, or long rest, meaning that they likely recover pretty fast.

Couple this with the plan to not have your pitcher bat, and it makes me really feed the necessity to make all but 2 or 3 of your pitchers as relievers.

I guess that would be .019 improvement in “I can throw harder, now”, and the other .008 because they don’t have to go through the order 3 times.

Looking at Lowe’s career: 7% walk rate as starter, only 5% as reliever.  His K rate was 14% as starter, and 19% as reliever.  Interestingly, the gap (K-BB) rate was 7% as starter, and 14% as reliever, and that 7% difference is consistent with what we’ve found. Also gave up much fewer HR, but also a much higher BABIP as reliever (.310 to .321).

John Wetteland?  6.3% gap, consistent with the others.  HR rate similar.  BABIP 29 points better as reliever.

Kelvim Escobar seems like an interesting case study for this as well.  He’s been both excellent and average while starting and while relieving, and he’s done significant amounts of both in the same season, which makes him fairly unique. 

My subjective observation tells me that relievers throw significantly more four seam fastballs than starters, and they work up in the zone more often, because the added velocity gives them a better chance for a swing and a miss.  Thus, higher strikeout rates and higher flyball rates.  When moved to the rotation, they throw more two seam fastballs, getting less strikeouts but more groundballs.

Escobar’s walk rate is 10% either way.  His BABIP is 20 points higher as a RELIEVER!  Way more HR as a reliever.  His K rate is 6% points higher, consistent with the others.

it appears to me that umpire strike zones expand at end of games...totally based on perception.

Well, that is a very interesting thought.  One of the weird results from THE BOOK was that as the pitcher goes through the order each time, they were getting worse, but when they reached the fourth time through the order, they actually got better.

Now, I controlled for the quality of pitcher, so that wasn’t it.  What we ended up with is that if the pitcher made it that far, he was probably “on”, and therefore, the sample was biased that way.

However, it is interesting thought to bring up that the game conditions may have actually changed.

So, this is easy enough to check.  Compare how a reliever does when he enters the 5th and 6th innings, as to when he enters the 7th and 8th innings.  Cool thought.  Anyone want to take a stab at it?

Tango: I’m not sure the way you have set it up (5th/6th v 7th/8th) would take into account game conditions....i think it’s a 9th inning phenomena, but i’m clearly not one to know how to check this kinda thing out so maybe i just oughta shut up