A’s Should Also Skip the Starter

A few weeks ago, I advocated for a wild card play-in game strategy that involved beginning the game with a reliever and relying heavily on a team’s bullpen to get them through. In that scenario, we used the Atlanta Braves as the example of how it could work. Well, we’re not yet to the wild card play-in game, but with Texas and Oakland going head to head for the AL West title in a few hours, we’re presented with another situation where skipping the starter makes a lot of sense – specifically, the Oakland A’s should try to maximize their chances of winning Game 162, and they can do so by diminishing A.J. Griffin‘s role in today’s game.

Make no mistake, Griffin has pitched really well for the A’s since they called him up from Sacramento in late June. While his ERA is heavily dependent on an unsustainably low BABIP, his FIP and xFIP are both above average as well, as his 3.76 K/BB ratio is the kind of thing that usually leads to success. This isn’t a knock on A.J. Griffin – the A’s can simply maximize their chances of winning by minimizing the amount of batters he faces.

The times-through-the-order effect has been well chronicled, but this table is worth repeating whenever we have discussions about pitcher usage in games where the outcome has dramatic consequences.

1st PA, SP 0.256 0.317 0.410
2nd PA, SP 0.268 0.328 0.430
3rd PA, SP 0.267 0.326 0.447
1st PA, RP 0.237 0.308 0.372
2nd PA, RP 0.281 0.339 0.463

Whether it is due to repetition, fatigue, or a combination of multiple factors, the reality is that hitters perform better against pitchers the more times they face them in the same game, and by the second or third time through a batting order, even the best starting pitchers aren’t much better than an average relief pitcher. And, while A.J. Griffin has had a nice run as a rookie, he simply can’t yet be considered part of the class of pitchers who can be expected to outperform their teammates in the bullpen.

Notice in that table that starting pitchers as a whole simply don’t perform as well the first time through an order as relievers do, even though starters are generally better pitchers than relievers. This truth goes against the modern way of starter usage in high leverage games, which calls for a starter to essentially pitch until he gets into trouble. The reality is that in many situations, a rally can happen to quickly for the bullpen to get warmed up in time to shut it down, and the game can be lost before the relievers can come in and put out the fire.

The A’s simply shouldn’t ask A.J. Griffin to face more than nine batters tomorrow. By letting him face batters multiple times, the A’s risk allowing Texas to put up runs in a hurry that could end up being the deciding factor. They don’t need to take that risk.

With the September roster expansion, the A’s have 18 pitchers currently on their active roster, though two of those – Brandon McCarthy and Brett Anderson – are injured and will be unavailable to pitch tomorrow. You can probably add Travis Blackley and Jarrod Parker to the list of unavailables as well, as they started the first two games of this series and would be going on zero and one days rest respectively. That brings the total all the way down to 14 available pitchers. Fourteen!

Even if we relegate guys like Jeremy Accardo and Jesse Chavez to the very end of the bullpen, the A’s have enough arms to play match-ups from batter one today. With Grant Balfour, Ryan Cook, Pat Neshek, and Evan Scribner from the right side and Sean Doolittle, Jerry Blevins, and Pedro Figueroa from the left, the A’s have some serious match-up options, and that’s before we even consider starters like Griffin or Tommy Milone, who should be available to throw an inning or two on his normal throw day. Realistically, the plan should probably be to get nine innings out of those nine pitchers, with the rest of the bullpen and Dan Straily being available in case it goes to extras.

Because there are so many specialists in that group, they’ll need to get multiple innings out of several guys, so Griffin, Milone, and perhaps Cook or Doolittle might be needed to pitch across several frames. But, with so many arms available, there’s no reason for any of them to face a single hitter more than once.

The A’s success this year is one of the great examples of baseball being more about the collective efforts of the many rather than the spectacular efforts of the few. That collective approach to winning should be on full display today, as the stakes for this afternoon’s game with Texas couldn’t be much higher. A win today gives them a pass into the playoffs and an extra day off to rest those arms, while a loss forces them into a win-or-go-home situation. The downside of losing is just too large to stick with a traditional starter/bullpen alignment. Griffin should essentially be viewed as just another cog in a very deep bullpen that should be used from the outset to do whatever they can to keep Texas from scoring. If he gets you six outs, that’s great. If he gets you nine, that’s a miracle. He shouldn’t be asked to even attempt to get a 10th. Not with expanded rosters, and not in a game that means this much.

Print This Post

Dave is a co-founder of USSMariner.com and contributes to the Wall Street Journal.

49 Responses to “A’s Should Also Skip the Starter”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Aggie E says:

    Its a nice article, but wont happen.

    Vote -1 Vote +1

    • Wrong Context says:

      So what if it won’t happen? That’s not at all what this website is about.

      If you’re interested in what’s going to happen, listen to the pregame show or watch espn.

      Vote -1 Vote +1

      • Ron says:

        Watching ESPN just gives you the run down on what Lebron James and friends think on Twitter or how this affects the Red Sox and Yankees. While I doubt that there will be a strict one time through the order plan, I bet Griffin will be on a short leash and if he shows any sign of trouble he will be out.

        Vote -1 Vote +1

  2. Pinstripe Wizard says:

    I thought I saw where Cook and Balfour have been used for four straight days. Granted they will almost certainly be called on if needed in the 8th and 9th tonight, but I doubt you could expect Cook to “pitch across several frames.”

    +27 Vote -1 Vote +1

    • bobbee says:

      All of the A’s high leverage relievers are taxed right now.

      Vote -1 Vote +1

    • Matt says:

      As others have mentioned below, in addition to 4 in a row for Balfour and Cook, Doolittle has been used 3 in a row and Neshek is on paternity leave.

      Normally I would be in favor of this, but besides the very good Blevins, the choices for replacing Griffin’s innings are some good but very gassed relievers or Scribner and Figueroa. I don’t think you want Chavez or Accardo anywhere near this game unless it’s a blowout. Milone just went Sunday and hasn’t come out of the bullpen at any level since 2009. Straily has really struggled, and thus I like your idea of not counting on him early but saving him for extra innings since he does have the ability to go long.

      I’m not saying you don’t use a quick hook on Griffin tonight, but despite having all those pitchers on the roster there really aren’t many great options on this particular day. It’s not really the game to go experimenting with things like whether Milone can pitch in relief.

      Vote -1 Vote +1

  3. Frankie B says:

    Interesting article, but I disagree in this particular case. Cook and Balfour have made four straight appearances, Doolittle three, all high leverage situations. It’s asking a lot to get multiple frames out of them, even though a win gives the A’s an extra day or two off (to Saturday, maybe even Sunday if they get homefield advantage). You also want Griffin, a rookie, to have his head on straight going into the playoffs. Yanking him after three innings, especially if he’s pitched well, doesn’t send the right message. With this combination of circumstances, while I might be inclined to have a quick hook if I’m Bob Melvin, I let the game play out before making a move.

    +6 Vote -1 Vote +1

  4. sporkless says:

    What if you employed this strategy and found yourself down by 8 runs after two innings? And what if you found yourself up by 8 runs?

    Those aren’t reasons not to do it, but the strategic choices are probably a little more dynamic. Interesting bit of game theory.

    Vote -1 Vote +1

    • Mike D says:

      You’d shouldnt be down by 8 in two innings because even before you get to that point your starter should be gone at any sign of major trouble ahead of him

      Vote -1 Vote +1

  5. Paul says:

    The A’s are where they are because they are tied for second in the AL in ERA, by using at least two pitchers who most teams would not have relied on to pick up major innings. In fact, at least half the teams in the league still would not have taken Straily and Griffin at mid-season, continuing to believe that they were correct to pass on them round after round in their respective drafts. Yes, hideous from all fork marks, but Moneyball lives.

    As for Griffin starting, Dave’s ruminations on relievers starting was interesting, but Griffin did shut out Texas in their own park in his second MLB start. Home ERA is quite a bit better than his stellar overall. And then there is again that whole part about them not being here without him you know, starting a lot of games.

    Vote -1 Vote +1

    • Average_Casey says:

      I don’t think it’s in your best interest to cite ERA as an example on Fangraphs. You probably would intead want to use FIP, opponent wOBA, opponent WRC/WRC+, BABIP, etc. His overall home stats may still be great but ERA does not tell you enough.

      Vote -1 Vote +1

  6. chuckb says:

    The Cardinals have decided, in their infinite wisdom, to go with Kyle Lohse rather than Adam Wainwright or Jaime Garcia against the Braves Friday. The same basic strategy should be employed by them — not allowing Lohse to pitch more than 1 time through the lineup. But it won’t be. The Cards could utilize starters like Miller, Rosenthal, Kelly, and Garcia all in relief in that game before even getting the game to Boggs and/or Motte but if Matheny insists on getting 6 innings out of Lohse, the Braves will win.

    Vote -1 Vote +1

    • Michael says:

      Garcia on the road, I don’t like; and as a Cards fan, I’m sure Carpenter will somehow talk his way into pitching in that game.

      Vote -1 Vote +1

      • chuckb says:

        The Braves wRC+ vs. lefties is 85. Garcia can get them out and should be used.

        I’ll bet we don’t see Carpenter in that game, having just pitched yesterday and being only in the teens in IP this season.

        Vote -1 Vote +1

  7. Aaron (UK) says:

    Shouldn’t Texas do the same thing? Is there a mutual stand-off going on here?

    Apropos of that, strategy in a BAL/NYY Game 163 (sadly now unlikely) would be very similar but even more dramatic because of the shorter turnaround time – it might be in both teams’ best interests to “agree” to pitch their #4 starter for 6 innings (or until 5 runs score) rather than burn their aces and/or bullpens.

    Vote -1 Vote +1

    • Zach says:

      this makes zero sense game theory wise because of how bad the outcome is if you get burnt.

      much better to go all out and be a little shorthanded for at least 2 non-elimination games.

      Vote -1 Vote +1

      • Aaron (UK) says:

        Sure, but – to a point – it’s probably what’s going to happen. Both teams will pitch their starter at least twice through the order unless they absolutely melt down, even though there may be a better strategy.

        But maybe this is evolved cooperation; if they both go Dave’s all-out approach neither benefits in this game and they are both potentially a little bit worse off in the next game.

        Vote -1 Vote +1

  8. JD says:

    In a vacuum, I’d be fine with this. Others have mentioned that Balfour, Cook, and Doolittle have been overworked, but also, Pat Neshek is not with the team. His wife gave birth, and he won’t be back till Friday (or Sat/Sun). That leaves the A’s down one reliable reliever for sure and likely down 2 or 3 more (even if those guys can pitch, there’s a strong likelihood they won’t be as effective. This is especially true of Cook, I think, who looked pretty shaky mechanically last night).

    Vote -1 Vote +1

  9. david hathaway says:

    Brilliant article and i totally agree. I hope Melvin sees this analysis.

    Vote -1 Vote +1

  10. Jason H says:

    Texas would sure be happy if Oakland did this. They’d love for Oakland to burn their best relievers early in the game in low leverage situations.

    …in reality the best strategy for must win games is the one managers have been using forever. Start the game with the best available starter, with the understanding that he is on a short leash. First sign of trouble and you are into the bullpen. Hopefully you are in the 6th inning by then.

    Vote -1 Vote +1

    • Aaron (UK) says:

      Trouble is the first sign of trouble is often a 2- or 3-run HR

      Vote -1 Vote +1

    • Justin says:

      Runs in the first couple of innings count the same as runs in the later innings. The only difference is how the effect the narrative.

      That being said, Dave is advocating for basically the same thing, only with a shorter leash. Let griffin start, but don’t let him more than 2-3 without having somebody up in the bullpen that can step on immediately.

      Vote -1 Vote +1

      • chuckb says:

        Right. That’s basically the strategy Dave suggests, only that leash is really, really short. There’s no need to have a starter in this sort of game, unless he’s a star, face the order the 2nd time through (unless there’s a big lead).

        Vote -1 Vote +1

      • vivalajeter says:

        Justin, I disagree with your first sentence. In a video game, you might be correct. But in reality, it just doesn’t work that way because there are human beings throwing the pitches and nerves are involved. If it’s a tight game in the 8th or 9th inning, with the division on the line and a rabid crowd going nuts, do you really think a lower-tier reliever will handle that pressure as well as someone who’s pitched at the end of ballgames all year long?

        In the box score, all runs are equal – but not every pitcher handles the pressure situations the same way.

        Vote -1 Vote +1

      • Justin says:


        Perhaps not, but if the team is in that position having already used its best pitchers, who’s to say that it wouldn’t be a three of four run game if they’d used their scrubs first? At that point it wouldn’t matter if you still have your closer because you’ve already lost. If its a high leverage situation early, by all means use your best pitcher — there’s no guarantee that another such situation will come up later.

        Watching the game right now, I’m guessing the As kind of wish they’d hone to the bullpen earlier.

        Vote -1 Vote +1

  11. drewggy says:

    Why not at least mention that the Rangers should adopt a similar strategy? Dempster is no great shakes, either.

    Vote -1 Vote +1

  12. jcxy says:

    i’m having a tough time looking only at the data you present and feeling confident about your conclusion. i feel like you need to tease out the correlation between how well you go through the lineup the first time with how well you go through the second (or third). in other words, we need to know whether recording a 0-2.5/9 effort predicts something different than a 2.5+/9 effort? my guess, although i don’t know, is that in-game performance impacts future outcomes more than those figures let on because there is a fairly evident systemic bias that starts sometime during the second pass whereby bad starters get lifted and good starters stay in longer. we see this in the similar 2nd and 3rd pass data. the data you use in support of your argument potentially ignores a very important predictor.

    the corollary is that while it’s true that starters might perform “relatively worse”, that’s not the same thing as “bad” and if the As choose to leave cook and balfour on the shelf, “relatively worse” from griffin may be better than their other bullpen arms. which is the other problem i have with the data you cite: you’re assuming that the talent in the As 12 or 14 bullpen arm is similar to the average RP. is this assumption justified? of the 14 arms, maybe 1/2 should be considered “average” by major league standards simply because of roster expansion. thus, i’d wager that the expected avg from a 1st pass RP on the As today is higher than that 0.237 avg you suggest.

    Vote -1 Vote +1

    • DCN says:

      Does the table include data only from batters the starter faces twice, or whether it includes all data and then classify it based on which pass it is.

      The systemic bias you mentioned should run the opposite way (make the difference between later passes seem less pronounced)? The worst pitchers will only show up in the data from earlier passes, so theoretically their later passes would be even worse.

      However, what about batting order bias? The first pass will almost always include all the hitters, but if the pitcher is pulled mid-way through, say, the third pass, the sample will be skewed toward the top of the lineup, and generally better hitters. So you would expect higher rate stats. (Shouldn’t really affect the second pass data too much though).

      Vote -1 Vote +1

      • jcxy says:

        The systematic bias is that starters pitching well will face more batters on the third pass than starters pitching poorly because those poorly pitching starters won’t end up facing as many 3rd pass batter. This impacts the outcome data because the set of pitchers it’s drawn from is different (in a way that we would think influences the results) than the other pools. It’s like considering that there is a practical lower limit to WAR because a team simply won’t play a -10 WAR player enough for him to accumulate that (anti) production. Where your logic is incorrect is in extrapolating this systematic bias logic backwards to it saying something about the 1st pass. There are many ways for a starter to accumulate first pass data, but many fewer ways to accumulate 3rd pass data. In fact, one might just as well imagine that there two (simplistic) tranches that would yield the 1st pass/2nd pass/3rd pass data that’s presented:

        Starter Type — 1st pass performance — 2nd pass — 3rd pass
        Good ——– Excellent ———— Very Good —- Good
        Bad———– Mediocre————-Poor——-

        As you can see, I’m not arguing against the broad trend that the more hitters see you, the better they fare. I’m saying that the data is incomplete and that there might be other influences that are of *better* service to a manager in making in game decisions that are not considered in the broad trends of 1st pass/2nd pass batting data.

        Vote -1 Vote +1

      • DCN says:

        Yeah, I got that. I completely understand that good pitchers will face more batters on the third pass. And I agree that it would be better to look at the data more closely.

        What I don’t get it is this:
        “Where your logic is incorrect is in extrapolating this systematic bias logic backwards to it saying something about the 1st pass.”

        Where am I doing this at all?

        If you mean “The worst pitchers will only show up in the data from earlier passes, so theoretically their later passes would be even worse,” I’m not saying the later passes say anything about the earlier passes. I’m saying the same thing you do – that we don’t have third pass data on the worst pitchers. So we’re not really getting a complete picture on how pitchers fare from one inning to another.

        With the batting order bias, here’s what I’m saying:

        Almost all pitchers face the whole lineup on the first pass
        Vast majority pitchers face the whole lineup on the second pass
        Many bad pitchers start getting pulled on the third pass (partway through the lineup)

        So what you have is:
        First pass: almost even distribution of batters 1-9
        Second pass: pretty even distribution of batters 1-9
        Third pass: distribution skewed toward the top of the batting order

        So there’s a bias in the selection of batters as well as the selection of pitchers.

        (One simple way to solve both these problems is to only include data from pitchers who went three whole times through the order. That way you get a better picture of how individual pitchers actually fare on subsequent passes.

        Of course, it’s not good for comparing starters to relievers since you’re throwing out the worst starters from the starter sample. But if you want aggregate data on how starters fare, that’s how you do it.)

        Another issue is in the reliever data – you’re getting a much different selection of relievers on the second pass data than on the first pass data. What’s the dropoff between first and second pass if you only look at relievers who got to the second pass (only a few relievers ever do)?

        Vote -1 Vote +1

  13. Fred Vincy says:

    I don’t think the A’s will do it this strictly, but I have no doubt AJ will be on a strict leash today.

    Vote -1 Vote +1

  14. David says:

    Honestly there’s only a very specific case where a manager really has the stones to do this, and that’s where the available starting pitcher options are just plain bad. A team with a good #1 and #2, for example, who just pitched the last two days … but with lousy choices #3-5.

    The closest this will actually come to happening is to put the starter on a very short leash and pull him at the first sign of trouble. Even so, most managers won’t consider 1 or 2 early inning runs to be “trouble”.

    Vote -1 Vote +1

  15. Chris From Bothell says:

    This suggestion is plausible in part because, unless I’m misremembering, the Rangers are a very impatient team. So they won’t be doing the work to try to cycle through relievers faster.

    But otherwise, the team should be asking their best contact hitters to plan to foul off as many pitches as possible, asking anyone who gets on base to try to draw throws to first and be stealing as much as possible, etc. If you know pitchers are on a short leash, get less aggressive at the plate and more aggressive on the basepaths.

    I’d think the A’s would be willing to use their pen somewhere in between the traditional “starter goes 6 and hands it over”, and the pen-o-rama Dave describes here. It’s just as others here noted, a shorter leash. If you’re trying to avoid a starter getting in trouble and not having a guy warmed up in time to save him, get relievers warmed up on the first hit given up no matter what, and then the reliever’s readiness dictates when they come in more than the score or situation.

    Also, with the bullpen-go-round that will likely be necessary, I’m sure the managers and pitching coaches are practicing their best Zeno’s-paradox-walk-to-the-mound. Also part of controlling the tempo of the game and giving max time for relievers.

    Finally, I imagine they will have some crib notes ahead of time as far as which relievers should be brought in to face which starters. A cool head and a commitment to specific matchups, pitch counts or innings limits should rule the day. If you’re planning on upending your bullpen on the other team, I’d think you’d need to be really disciplined about which guys come in when, and not be a slave to pitcher-specific splits, team-specific splits, or whatever pinch hitters are brought in. Being reactive to pinch hitters basically lets the offense control how you burn your pen and I could see the bullpen usage thing go haywire quickly. (Though admittedly, I have a skewed feel for this sort of thing because I’m an M’s fan and have to suffer through Wedge’s weird strategic choices.)

    I realize all the things I note here are probably things every team should be doing in almost every game. Or at least in games where you can upend the pen and be confident you don’t need most of them again for a day or two.

    All the theorizing about this one, and the implications for the play-in game, are making me more interested to watch a regular season game that my team isn’t in, in a long time…. hooray for the second wild card!

    Vote -1 Vote +1

  16. vivalajeter says:

    It’s entirely possible that a team might have a slightly better chance of winning a game using this strategy, but I’ve seen nothing to suggest that the A’s should take this approach today.

    Griffin has done a terrific job keeping runs off the board this year, and his ‘per inning/pitch’ splits don’t show that he suddenly hits a wall after getting through the lineup the first time. In fact, he pitches better in the 45-60 (.673 OPS) and 61-75 (.564 OPS) pitch categories than the average reliever in the 1st PA data listed above (.680 OPS).

    If the A’s had an incredibly deep and fresh bullpen and Griffin was a borderline-at-best starter, then yeah, it might be a decent strategy. But the way things are, I don’t see how this strategy is better than asking him to pitch 5-6 innings, but keeping him on a short leash in case things fall apart early.

    If he throws 3 perfect innings to start the game, do you really want to pull him just because the average reliever is better the first time through a lineup than the average starter is the second time through a lineup?

    Vote -1 Vote +1

    • jpg says:

      Agree. Plus I just think you’re running a greater risk of asking 10 relief guys to get, say 20 outs, than one starter get you the first 15 outs. In other words, your relying on 10 guys to all have their “good stuff” coming out of the pen. The other issue I see is that you can’t play platoon matchups forever when you only have 3 or 4 lefty short men. What happens in the 8th when you’ve burned through all of your lefties early on and Hamilton comes up? Since this plan calls for using your best relievers early and often, now you are potentially asking your 8th best righty, who is rarely asked to face a lefty, to get Hamilton out.

      Vote -1 Vote +1

  17. Wally says:

    Another thing that is not mentioned here is that the quality RP innings are nested with the current standard usage patterns. So, you see a team generally rely on just a few RP a night. This strategy will generally rely on instead of maybe 3 RP on average, but 5 or 6. I’m guessing, in generally, those extra 2 or 3 RP are going to be quite a bit worse than the first 3 used for most relief innings. Further if you isolated the numbers for your actual choice, which is using maybe the 5 or 6th best reliever or letting your starter go through the order the 2nd time, you’d probably find the numbers to be quite a bit different. Especially, if you assume your starter is already better than average as well. To put it statistically, what happens when you compare the 1-sigma better than average starter the second time through the order, versus the 1-sigma worse than average reliever the 1st time through the order.

    I think you have to let Griffin go through the order twice today, or at least give him the chance to. If at any point the second time through the order a batter reaches base, by all means, get someone up in the pen. But the A’s just don’t have the arms available tonight to be in their bullpen in the 3rd inning.

    Vote -1 Vote +1

  18. Neshek’s wife gave birth yesterday, he flew home to FLA and is out for the game.

    Vote -1 Vote +1

  19. Average_Casey says:

    Given the fact that Billy Beane is always willing to buck convention, I think if any team were likely to use the strategy, it would most likely be the A’s. Also, when you account for the fact that they have a very young starter at the end of the season with the largest amount of pressure possible, it would make sense to get him out of the game as fast as possible. While I think that I would have him go through the lineup twice if it’s a low scoing game, I cannot argue with Dave’s idea because it is sound I hope Oakland tries a similar strategy because it would be fun to see it in action.

    Vote -1 Vote +1

  20. clevelander says:

    This is essentially the same strategy used by both teams in the MLB All Star Game every year. So, there is only a 50 50 chance of success. Not good enough.

    Vote -1 Vote +1

  21. Dave (UK) says:

    Are you going to wait until the end of the game to say ‘I told you so!’, Dave?

    Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Current ye@r *