Garcia or not to Garcia, and when to take out your starter

Posted: October 7, 2013 in In-game strategy, Managers, Post-season

Let’s start with Freddie Garcia, to whom the Braves have entrusted the fate of their 2013 post-season – to some extent at least. Is Garcia a terrible starter? I don’t think anyone knows that. The Braves apparently think not. They think that he is at least better than Alex Wood or Paul Maholm. (I’ll get to Wood in a second.) In addition, they seem to think that his post-season experience will help them/him today. Freddy Gonzalez was quoted as saying this:

“You start looking at the rotation and see how young they are and how inexperienced they are, then you look at Freddy Garcia’s experience, you look at his postseason starts and postseason innings, so you give him an opportunity to do that,” Gonzalez said. “I think he knows how to maneuver himself through a Major League lineup. He’s shown that this year.”

While I think that is more of an excuse for Teheran’s poor performance last night than a ringing endorsement of Garcia, there may be some merit to that argument. I also think that Atlanta’s decision to use Garcia over Maholm and also Alex Wood’s demotion to the pen is mostly predicated on the Brave’s incorrect emphasis on recent, small samples. Garcia has something like a 1.65 ERA with the Braves since being acquired on Sept. 1 or so. Oh, that’s in 27 IP. I don’t have to tell you how meaningless 27 innings are in terms of future performance.

Still, there is some argument as to what Garcia’s true talent level is now. I think it is around 14% worse than the average starter. I think Steamer (a very good forecasting system designed by the very smart and knowledgeable Jared Cross) rates him about the same, ZIPS likes him more (around a league average pitcher), with Pecota somewhere in between. Don’t take those assessments as the gospel.

Dave Cameron, on Fangraphs today, points out that he has been around league average, xFIP-wise, for the last few years, however, he is likely on the last legs of his career at the present time, his regular FIP has not been good, and no team even wanted him until the Braves picked him up off the scrap heap.

So, all told, given all the evidence, it is likely that Garcia is a well-below average starter.

What about Alex Wood or Maholm? Maholm actually has decent numbers and decent projections across the board, a little below average, perhaps akin to a typical #3 or #4 starter. He appears to be a better choice than Garcia. Cameron, in that same article referenced above, says:

And while Maholm has a pretty solid track record, his results were pretty terrible after a strong start, allowing a .368 wOBA in the second half of the season. His massive platoon splits — righties have a career .349 wOBA against him — are also a pretty poor fit for a start against a team that features guys like Hanley Ramirez and Yasiel Puig, so it’s understandable than the Braves didn’t particularly want to give Maholm the ball in this game either.

All of that is nonsense. How he has done “after a strong start” is irrelevant other than how it has changed our projection for him (a little, but not much – we have plenty of data on him), assuming that he is healthy. And Maholm being a poor fit for the Dodgers because of the righties Puig and Ramirez? Huh? Who cares who the righties are? It is how MANY righties that counts. The Dodgers have 3 lefties in their lineup out of 8, which is 38% of their lineup, while Maholm has actually faced only 19% lefties in his career. (I am not sure that Schumaker remains in CF versus a “tough” lefty pitcher.) Managers typically stack their lineups with righties against him. If anything, the Dodgers lineup is ILL-suited to face a lefty pitcher with large splits! So now, Maholm is looking a LOT better than Garcia.

What about Alex Wood (second round 2012 draft pick and high on the Braves’ prospect hierarchy)? I love his projection. I have him projected at 10% better than a league average starter – around a #2 starter. His 2013 rookie season has been spectacular (and under the radar). Go look at all the pertinent numbers. And his minor league numbers are above reproach. Go look at them too. Why was he relegated to the pen? I assume because he had 2 bad starts in September, which is ridiculous, but typical behavior by stupid teams. And yes, for the record, I consider the Braves, and especially Fredi Gonzalez, stupid teams/managers, at least when it comes to convention versus sabermetrics, or more appropriately, fiction versus fact (yeah, yeah, I know they’ve won lots of games over the years, yada, yada, yada).

Wood appears to be, by far and away, the best choice to start today’s game. Of course he is already in the pen and has already pitched as a reliever in the post-season, so that cannot happen. Big mistake by Atlanta, fighting desperately not to go home for good. How much of a mistake? With Maholm over Garcia, I have the Braves winning 2.5% more often than with Garcia pitching. With Wood on the hill, the Braves win a colossal 7.9% more often than with Garcia! So I see your, “Wood had 2 bad games in September,” and I raise you, “8% in WE!” That’s twice the value of Joe Maddon’s “I’ll bat Delmon F. Young rather Matt Joyce because, you know, well, you know, I’m Joe Maddon!”

All that being said, what is the absolute best alternative? We’ve all mentioned it before, and Dave Cameron reminds us:

I’ve written about the case for skipping the starter in elimination games many times before, and this is the kind of game where an all-hands-on-deck approach is most warranted. Garcia might be their best option to start the game, but there’s no reason to let him try and pitch himself out of impending doom. If the Dodgers start stringing some hits together, get someone up in the bullpen, and be ready to make the move to put out any fires that arise in any inning in order to keep the game close.

Yes, of course. You don’t need or want a starter when you don’t have an ace, especially in a post-season elimination game! Pick your 3 best relievers, although you can save your closer and possibly set up guy for the last couple of innings, and roll them out there one at a time, either until they face the order 1 time or they come to bat. Speaking of – the first time Gonzalez lets Garcia bat rather than a pinch hitter and then removing him from the game, the first time I will say what an idiot he is. Also, when you bring in your relievers try and pay attention to the L/R matchups. For example, you have Crawford leading off and Gonzalez batting 4th. Make sure that your lefty reliever pitches to both of them. Basically when you switch pitchers, make sure that at least for the first batter, you have the platoon advantage.

Now, lets get to this idea of when to take out your starter other than for a pinch hitter or after he has gone through the lineup once (if you are going to only allow him to pitch one through the lineup), or 2 or 3 times.

Cameron above says, “If the Dodgers start stringing some hits together, get someone up in the bullpen, and be ready to make the move to put out any fires that arise in any inning in order to keep the game close.” David Schoenfeld, in his Sweetspot article today, talks about how Fredi should have taken out Teheran after he allowed X number of runs or hits. I am afraid all of that is nonsense. The reason is this:

There is no evidence (that is code for, “it ain’t true”) that  how a starter has pitched at any point in the game has any relevance (predictive value) to how is going to pitch to the next batter or for the rest of the game. Given that, your decision to remove your starter or not should have nothing to do with whether he has just gotten hammered or is pitching a shutout through 6 or 7 innings.

What does matter is two things and two things only: How good he is in general (independent of how he has pitched thus far in this game) – basically his current projection, and how many times he has faced the batting order. (Of course, if a pitcher has thrown a lot of pitches and is tired, it may be time to take him out no matter what.)  We know that each time through the order, a pitcher gets decidedly worse, independent of how many pitches he has thrown and independent of how that batting order has done thus far in the game. This is an extremely important concept, and one that NO manager has any inkling of. (Managers know about the “times through the order thing – you hear announcers talk about it all the time. But they often ignore it and they use other incorrect assumptions to trump it – such as, “My guy has been pitching great, so surely I should leave him in there even though he is facing the lineup for the 3rd time.”)

What are the consequences of using the ubiquitous, “Ill leave him in there if he is pitching well, and I will take him out when he allows X number of runs/hits balances against how many innings/pitches he has thrown among other things?” Well, since it really doesn’t matter when you take him out along those lines, what happens is when you take out a good or great pitcher after he has gotten hammered and put in a worse pitcher (after accounting for any time through the order penalties, and platoon considerations, etc.), you cost yourself WE. What happens way more often is that you allow an inferior pitcher to continue pitching when he has thus far pitched well. That costs WE as well, and usually a lot. It is not uncommon to see a bad pitcher pitch an extra inning or 3 when there is a reliever who is a run better that should be in the game. For 2 innings, that is .22 runs or around 2% in WE. That’s a lot to give up.

Caveat: I realize that you can’t be yanking pitchers left and right when they are pitching well and always allowing pitchers who are getting hammered to keep pitching. You will lose the trust of your players and perhaps impede the development of some of your pitchers – but surely there is a happy medium.

BTW, and I don’t mean to pick on Schoenfeld, as he is very smart and writes mostly good, sabermetric-oriented articles on ESPN, here is my response to today’s Sweetspot piece (which is NOT good – the piece that is):

In your article on Teheran and the Braves, you are assuming that Teheran would have continued to pitch poorly at any point in time that he DID pitch poorly and you would have taken him out.

Do you have any evidence of that? Making an argument that he “should have” taken him out because he did in fact allow more runs is a ridiculous argument and you know that.

If Kershaw or Scherzer allows a few runs and then allows 5 more, “should” the manager have taken him out too before those 5 runs scored?

Obviously in all case, had the manager taken the pitcher out, those subsequent runs would likely not have scored, but, absent a crystal ball, the manager cannot know that at the time he has to make the decision.

Now, the point about having fast hooks in general in the post-season is a good one. That is because you don’t worry about tiring your pen during a 162 game season. But, what does a fast or slow hook mean? If there is no evidence that how a starter HAS been pitching has any predictive value for the rest of the game (which I am afraid there isn’t), then when SHOULD you take out a starter?

I would submit that you “should” take out a starter any time you have a better option in the pen, talent-wise (and L/R matchup-wise), without being ridiculous, and without burning up your relievers for that game or for the rest of the series (e.g., you don’t put your best reliever in the game in the 3rd inning even though he is better than the starter). You can wait until fairly late in the game OR until he HAS pitched badly ONLY FOR PR purposes and for the purposes of not offending your players!

That means that the talent of the starter has a lot to do with it. If your starter is Teheran, then very much you want him to pitch as little as possible REGARDLESS OF HOW HE IS DOING, since almost every one of your relievers is better than he is, especially after Teheran has faced the order 1 or 2 times. If your starter is Kershaw then most of your relievers are worse than he is at almost at any point in the game.

But, to fault Gonzalez for not taking out Teheran because he went on to allow a few more, is not a logical argument, unless, again, you think that Freddy was supposed to know that Teheran would come out and allow 2 more runs and 4 more hits. If you don’t take out him out after 2 good innings, it is NOT correct to take him out after 2 bad innings! You can’t have your cake and eat it too! As you say, he was one pitch away from getting Crawford out in the 2nd inning. If it is correct to leave him in there if he gets Crawford out, how can you say that because he threw a slider which was not even awful, that it is now correct to take him out? That makes little sense.

Wouldn’t the reliever who replaced him have just about the same chance as allowing those 4 hits and 2 runs? Would you fault a blackjack player for hitting a 12 against a dealer’s 10 if the player busts and the dealer had a 6 in the hole?

Finally, what is the Braves’ obsession with Gattis? He is clearly a very poor defender in LF (he played a little LF in the minors, but mostly catcher). And he is not that good on offense. Why not play B.J. Upton in left? Oh, I forgot for a second – the Braves are also obsessed with small samples and, “What have you done for me lately?” They think Upton is a terrible hitter. But, even if he is not nearly as good as we used to think that he was, he is a true CF’er and therefore going to be a well-above average defender in LF. He is probably 20 runs better (per 150 games) than Gattis in LF. Is it reasonable to think that he is anywhere near 20 runs worse than Gattis on offense? Again, if we don’t attach some ridiculous weight to how he has done in 2013, I don’t think there is any argument to make in favor of that. In fact, if I run both players through my sim, Upton and  Gattis, with Upton in the lineup, the Braves add another 2.5% to their WE!

I mean, how much win expectancy can the Braves and Gonzalez give away before the game even starts!

  1. mgl59 says:

    Perfect example of what I was talking about – 6th inning, Morton comes out to face the order for the 3rd time. Morton is not a good pitcher to start with. He is a below average starter. The third time through the order, he is terrible. It doesn’t matter that he hasn’t allowed any runs yet. And it doesn’t matter that he has thrown only 80 pitches. We can prove that. We can go back in history and look at all roughly average pitchers, who have thrown, say, between 70 and 90 pitches, have not allowed a run (or we can do, say, 0-2 runs) , and who are facing the order for the 3rd time. What we will find is that they are not well-below average. Or you can take my word for it.

    In either case, that is what we can expect from Morton in the 6th and beyond. I assume we have at least one alternative in the pen who is consderably better than that. So why it Morton starting the 6th? Because that is what almost all managers would do. That’s not a good reason. In fact, that is not even a reason. Again, during the regular season, we have bullpen conservation issues to worry about and we have player egos to worry about. In the post-season, I think we can legitamtely throw those things out the window.

  2. mgl59 says:

    Same for Wacha of course, no hitter or not. And what a terrible steal attempt by Harrison in the 8th. Short lead, not a great jump (despite what Brenly and company said), looked back at the plate more than once (which slows you up) and an early slide. With that pitch and that throw, a good base stealer should be easily safe.

  3. TomC says:

    Just kill me now (Garcia leading off the 5th)

  4. mgl59 says:

    Yes, that is pretty bad. I sort of predicted that. How much does that cost in WE? With a pinch hitter, you are looking at around an RE of .56. WIth a pitcher at bat, he gets out 85% of the time resulting in an RE of .35. When he reaches base, it is an RE of around 1.2. So .85 * .35 + .15 * 1.2, or .48. So you lose .08 runs or so. Around 1% in WE. Not too terrible I guess. But, you also want to get Garcia out of the game! That is going to gain you at least 1 run in RA for 1 or 2 innings! That is another .17 runs or 2% in WE. So letting him hit and having him pitch another 1.5 innings or so costs around 3% in WE. That is not wood!

  5. mgl59 says:

    I have to honestly say that I feel sorry for Braves fans. I may rip on Maddon for trying to outsmart everyone else, but Fredi Gonzalez is just not smart. Surely you have to bring in Kimbrel in the 8th inning with Puig on second and no outs. Carpenter is a very good pitcher, but Kimbrel is other-wordly. He is more than a run per 9 better than Carpenter. By the way, I don’t know if it was just a fluke, but the Braves just played so sloppily the entire series. I almost cannot think of an inning where they didn’t kick the ball around, or run the bases poorly, or lay down a poor bunt, or some such thing.

  6. TomC says:

    Yeah, it’s brutal to watch. Fredi hadn’t even started anybody stretching, much less warming up, when Garcia went back out for the 5th. He kept trying to lose tongiht and finally got there.

    • MGL says:

      I have a vague recollection that in 2010 they played like a Little League team in losing to the Giants, but that could easily be another team and/or another series. Maybe it was Cincinnati in some post-season series. But yeah, it is funny how it seems like those 1% mistakes invariably turn into 100% losses!

      • Alvaro Pizza says:


        Game 3 of the Giants / Braves series was the Brooks Conrad 3 error game. Elliot Johnson and Gattis are the new Brooks Conrad in 2013. In that game Cox also pulled out Kimbrel out of the game to let Mike Dunn pitch to Aubrey Huff. I know Dunn and Huff are both lefties, but Cox panicked because Kimbrel was “inexperienced” and “shouldn’t” pitch to LHB.

        In game 4, Cox let Derek Lowe in too long.

        Last year against the Cardinals, Kimbrel enters the game in the 9th while losing 7 – 4. According to Fredo, Kimbrel just can pitch in the 9th. There is no way Kimbrel can put out a fire in the 7th or 8th inning of an elimination game.

        Fredo “Corleone” Gonzalez and Bobby Cox. WOW.

  7. MGL says:

    Yes, I clearly recall Cox being an awful strategist too. I would be a little careful about some of the things you say above.

    Are you sure that Dunn vs. Huff is not better than Kimbrel versus Huff? I don’t know the answer. I would have to do some research.

    And, what do you mean by, “Leaving Lowe in too long?” If you read my post, for an experienced pitcher at least, like Lowe, allowing some number of hits or runs has no predictive value, so the concept of “leaving a pitcher in too long because he is getting hammered” is a fallacious one. In fact, you never hear someone complain about that unless the pitcher continues to get hammered. I submit, and the data will support it, that pitchers will “continue to get hammered” (however you want to define that) at exactly the same frequency whether they have been pitching excellently or badly.

    So there is no merit to the argument that a manager should take out a pitcher when he is (WAS, really) getting hammered or having (WAS having) trouble with his command, etc. You take out a pitcher when you have a better alternative, with leverage and bullpen conservation in mind, and/or in the NL when that pitcher comes to bat in a more than low leverage situation. Those times are not rigid or mandatory, but they are better than “how a pitcher has done so far.” Of course ALL managers use the latter. It is human nature, and a strong part of human nature at that. And they are for the most part wrong. They end up taking out or leaving in pitchers at sub-optimal times and they reduce their team’s chances of winning the game as compared to a more optimal strategy for relieving your pitchers as the game goes on.

    So, again, I don’t know what you mean by, “Leaving him in too long,” but I suspect that you mean that he was getting hit around, he left him in, and he got hit around some more. In that case, your statement may be correct, but not for those reasons. Him getting hit around has nothing to do with when to take out a pitcher (other than it is convenient – everyone accepts that, you don’t have to explain yourself), and him continuing to get hit around, if that happened, CERTAINLY has nothing to do with his decision which was made before that happened, unless you think he should have been prescient! 😉

  8. Alvaro Pizza says:

    Hi MGL:

    Looking at the numbers; Dunn vs Kimbrel against LHP in 2010 are almost the same. I don’t manage projections so I don’t know them. What I know is Kimbrel is and was at the time a better reliever with better stuff.

    Lowe was NOT getting hammered, but he was a mediocre 37 years old starter that year who had a great september, and Bobby Cox thought the Derek Lowe who was almost an ace with the Red Sox and with the Dodgers had came back just because Lowe had a great september and a good start in Game 1 in San Francisco. Cox had a great bullpen that year even excluding Billy Wagner who had gotten injured in Game 2.

    Available in that bullpen that day: Takashi Saito, Johnny Venters, Craig Kimbrel, Eric O’Flaherty, Mike Minor and Brandon Beachy. After the 5th inning every one of those is better than Lowe facing the lineup the 3rd time.

    You can watch the box score of the game here:

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 )

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s