Kimbrel, Schmimbrel

Posted: October 8, 2013 in In-game strategy, Managers, Post-season

By now, if you are a Braves fan, or even if you’re not, you’ve read or heard that Fredi Gonzalez should have brought his closer, Kimbrel, into the game either, at the start of the 8th inning, or, after Puig reached second on a lead-off double.

The reasons are, one, he should be able to pitch 6 outs in a post-season elimination game, the batters in the 8th are better than the batters in the 9th, or just very good in general, and a runner on 2nd with no outs in 8th inning with a 1-run lead is a high leverage situation.

Let’s look at these one at a time to see if they have any merit, and if so, how much (on this site, we like to quantify merit!).

Could Kimbrel have pitched 6 outs and if so, should he have? Unfortunately, we have no way of answering that question. Most closers don’t, a few do, and a few more do during the post season. I don’t have any numbers handy, but we know that Mariano was often called on by Torre to do just that, and he seemed to fare just fine. Kimbrel was not overworked, in the post-season you don’t play every day, and of course, he has the entire off-season to recover no matter what happens. That still doesn’t mean that it is correct for him to go 6 outs, and that doesn’t tell us how much that increases the Braves’ chances of winning even if he can go 2 innings at full strength, although we can estimate the latter.

Fredi was quoted as saying that he was planning to use Kimbrel for 4 outs. As many people have pointed out, that seems kind of arbitrary, and if 4, why not 6? Well, by that logic, if 6, why not 7? Etc. So, I am not crazy about that “argument” in and of itself.

Now, the pitcher he did use, his set-up man, Carpenter, is no slouch. He allows 79% of league average runs (all numbers I am quoting are my projections), which is equivalent to a little better than a league average closer. Then again, Kimbrel is other-wordly. He is Mo before Mo was Mo. He is Eck when Eck was unhittable. He is as unhittable as I think a human pitcher can be. I have him as 47.25% of league average, which is more than a half run better than Clayton Kershaw, the best starter in the world.

It really doesn’t matter how good Carpenter is. What counts is how much better Kimbrel is than the next best pitchers in terms of whom to bring in when. And despite Carpenter being excellent, Kimbrel is 1.28 runs per 9 better, which is enormous. That is like the difference between a regular closer and a really bad long reliever. So, using Kimbrel rather than Carpenter in appropriate situations is definitely a large advantage. The argument that, “Carpenter is a very good reliever in his own right,” is not really relevant to this discission.

Anyway, let’s take a look at what kind of advantage there is to pitching Kimbrel in the 8th and 9th rather than Carpenter and then Kimbrel. We’ll assume that they both pitch exactly one inning in one case, that Kimbrel pitches exactly 2 innings in the other case (obviously if Atlanta were to take a big lead, they could sit him in the 9th, or pitch Carpenter in the 9th). We’ll also try out Kimbrel in the 8th and Carpenter in the 9th.

For that, I am going to use my game simulator. The sim is particularly suited to these kinds of 1 or 2 inning scenarios where we assume that we know the talent level of the pitcher, and that of the batters (although that is not that important), and we don’t have to worry about any possible pitching changes.

Here is what we get:

With Kimbrel in the 8th and Carpenter the 9th: WE=.7819

With Carp in the 8th and Kimbrel the 9th: WE=.7857

With Kimbrel pitching the 8th and 9th: WE=.8258

Those numbers appear to make complete sense. It is better to pitch Kimbrel in the 9th than the 8th, if he only pitches an inning, but not by much. And it is very advantageous to pitch him for 2 innings, assuming that he is able to pitch at full strength for 6 outs. Apparently Fredi does not think that he can. I don’t know that we can argue against that. We have nothing to really hang our hat on. Obviously if he is affected, he may be no better than if he pitches an inning and Carpenter pitches an inning. In that sense, Carpenter’s talent is important. It may be that a stretched-out Kimbrel for 2 innings is no better that a full-on Kimbrel for 1 and Carpenter for 1. For what it is worth, in tonight’s Detroit/Oakland game, Leyland said that he is reluctant to use his closer, Benoit, for more than 1 inning. Of course Leyland is definitely cut from the conventional manager cloth.

How much is gained by bringing in Kimbrel with 2 outs in the 8th, as Fredi was planning, and how much would have been gained had he brought in Kimbrel after Puig reached?

Kimbrel entering with 2 outs in the 8th: WE=.7971

So that adds a little over 1% to Atlanta’s WE, as expected. Of course, we don’t know that by coming in for a 4 out save, that Kimbrel is not less effective than for a 3 out save. So maybe there is no gain.

Kimbrel entering with Puig on 2nd, no outs in the 8th: WE=.6674

Kimbrel not entering with Puig on 2nd, but he comes in with 2 outs if the Braves still have the lead (Fredi’s strategy):WE=.6227

By not bringing in Kimbrel at this point, we lose 4.5% in WE rather than 4%, which is the gain by bringing him in to start the inning (as opposed to waiting until the 9th). There is only a 2.8% gain when allowing him to pitch the 8th and 9th (.8258) compared to 2 outs in the 8th (.7971). I hope you are able to follow all that.

Remember that the 4.5% gain when Puig reaches second is still predicated on Kimbrel being just as effective for 6 outs as for 4 or 3 outs, and again, we don’t know that to be true. We almost have to assume that there is some penalty, no? Sure, if he throws 11 pitches in the 8th, he is probably fine. But, part of the equation are those rare times when he throws 20 or 30 pitches in the 8th and either can’t pitch the 9th or can’t pitch as effectively in the 9th.

Now what about these claims of, “In the 8th, the 5, 6, and 7 batters are due, and in the 9th, 8, 9, and 1 (of course, it is around 50/50 that the 8th will go 1,2,3 no matter who is pitching), so the 9th could just as easily be 9, 1, and 2, or worse (for Atlanta), and you want your better pitcher against your opponent’s better batters?”

Is that one of those things that makes sense if you don’t think about it too much? I’m not sure I am convinced of that argument. I’m not even sure what it means. It’s not like your better pitcher gets the good batters out all the time but the worse pitcher doesn’t. The better pitcher reduces the production of both the good and bad hitters. I am not even sure what to test.

I am going to try this: Have the good and bad part of the order come up in the 9th and see what the difference is between the closer and set-up guy.

I am looking at the bottom of the 9th inning, protecting a 1 run lead, with the leadoff batter up. Let’s see the difference between using a closer like Kimbrel and a pitcher like Carpenter.

We=.8681 with Kimbrel and top of the order in the 9th.

We=.8065 with Carpenter and top of the order in the 9th.

That is a huge difference. 6.2% of WE (which is why Kimbrel is so valuable!). But what about with the bottom of the order in the 9th? Under this theory, we should see less of a difference, since to some extent we don’t need Kimbrel.

WE=.9003 with Kimbrel and bottom of the order in the 9th.

WE=.8547 with Carpenter and bottom of the order in the 9th.

Here we have only a 4.5% advantage for Kimbrel over Carpenter.  So at least in this situation, that theory appears to hold some water. Of course an extra 4.5% is nothing to sneeze at. If you had a 1-run lead in the 9th, even with the bottom of the order coming up, you definitely want your best pitcher!

Bottom line here is that even at its worst, not bringing in Kimbrel to start the 8th, or with Puig on second, costs on the order of 4 or 5%. That is huge for one decision, but certainly doesn’t mean that their manager cost them the game. And that is the worst (or best) case scenario. If pitching 2 innings has any negative effect on Kimbrel, which it might, then that 4 or 5% can certainly disappear quickly.

  1. Adam R says:

    If 47.25% of league average is more than half a run better than Kershaw, and Kershaw is the best starter in the world, doesn’t that mean that every starter in the world allows more runs than average? The only way I can think of that this could be true is that all relievers are below the league average and all starters above it, but I assume that starters pitch more innings than relievers, so I remain skeptical.

  2. Jonas M says:

    I am curious about your baseball simulator. What factors does it take into account? Have you written about it anywhere previously?

    • MGL says:

      I talk about it occasionally. I use it to analyze some situations. Sometimes it is very good for that and sometimes it is not so good. It is best for very specific situations where you don’t have to worry about things like pitching changes, pinch hitters, infield playing up or back, etc. It uses projections for all players and assumes an “odds ratio” result for the batter/pitcher matchup including platoon adjustments (using a platoon ratio projection for each player). On top of that it uses defense, base running, arms for the outfielders, GB/FB projections for pitcher and batters, SB/CS projections for runners, pitchers and catchers, pinch hitters, sac bunt attempts, park factors, etc.. Pretty much the whole 9 yards, so, as I said, it is pretty good for analyzing lots of different situations, strategy alternatives, etc.

  3. David says:

    The significance of a 4 out save, of course, is that Fredo is saying that he didn’t have a problem with Kimbrel sitting down between innings and then coming back out again. That’s a “big deal” for those who argue against using one inning RPs for longer outings. So if he’s already allowing for that, then yeah, why not 6 outs? 7 outs would be twice up and down again and that certainly could be too much for a nominal one inning guy. Maybe. Who knows?

    • MGL says:

      Right, I agree that it would be more likely for 6 outs to be closer to 4 or 5 outs, then 6 outs and 7 outs, because of, as you say, the extra time in between innings. But, the bottom line is that we don’t know. As I said in the piece, clearly there are times when 6 outs is going to be taxing or even impossible, so I think it is fairly safe to say that there is some penalty for the 6 out appearance (even if Mariano showed no difference – in fact, he was slightly better in 6 out saves – but that is a relatively small sample).

      I would find it hard to believe that 6 outs is exactly equal to 3, but that 7 outs is worse. I think that every out must be marginally worse and maybe there is an additional penalty for sitting in between innings.

      I mean, every out adds pitches and at some point a short reliever becomes a long reliever, right? And you would NOT think that a curve of # pitches versus effectiveness would look like a set of stairs, right?

      On top of everything, we have the fact that virtually no manager uses closers for more than 3 outs, occasionally 4 outs, even in the post-season with the exception of Mariano. While that fact is is not dispositive of anything, and it could represent an inefficiency (and it could be dependent on what part of the season we are talking about), I think it gives us some evidence that it is not optimal for a short reliever.

  4. TomC says:

    You don’t necessarily have to eat the full penalty if he starts the 8th since you can relieve him in the 9th if he’s had to work too hard, but if you bring him in for the last 4 outs, there’s no way to leverage a efficient outs except in the relatively unlikely case that the game goes to the 10th and his spot doesn’t come up (I think he was due up 9th in the 9th, so 3 runners in the 9th+10th and he’s gone), and even there, you’re eating whatever the double-sitting penalty is in the 10th.

  5. MGL says:

    TomC that is an excellent point! In fact, that applies to using all closers in the 8th rather than the 9th, even if the leverage in the 8th is less than that expected in the 9th. Let me rephrase it, if you don’t mind.

    Let’s say that, as with my sim results, using Kimbrel in the 8th or 9th is relatively close in WE (there is only a .38% difference in WE, although part of that is the fact that better hitters hit in the 8th). Even though you gain a little by waiting until the 9th (and that is true even if you blow the lead in the 8th), if you use Kimbrel, or whoever your best reliever is, in the 8th, you can let him pitch the 9th if he only throws a few pitches in the 8th, or you can take him out after the 8th, if you think he threw too many. On the average, he throws somewhere between 1 and 2 innings, but he never throws as a lot of pitches. As TomC says, you can leverage him based on the number of pitches thrown in the 8th. That is brilliant and almost all closers should be doing that!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s