How Significant Is Batting Order?

Most sabermetric analyses of batting order find that the most optimal batting order is worth between five and 15 runs over a typical batting order. From this, it is often concluded that batting order isn’t very important. Is that the correct conclusion?

I enjoy thinking about batting order in general and in specific cases such as this. But I don’t want to get into the particulars of methodology today. The Book Blog has plenty of posts and links on the subject. Jack Moore spurred an interesting discussion in his post yesterday, although it was a coincidence that he posted it the day before I did this. This is a different angle on the issue. Is the reported difference — five to 15 runs a season over a typical lineup (which is different than the worst possible lineup), according to The Book — is “significant” or “meaningful.” What are some other examples of five-to-15 run differences?

In recent run environments, five runs is actually closer to one marginal win than to zero. What else would be worth five runs during a season? Imagine that the Chicago White Sox’ projected 2011 right fielder Carlos Quentin gets hurt (it’s a bizarre hypothetical, I know, but try to suspend your belief) and has to miss 40 games, or about 150 plate appearances. Assume those plate appearances go to Mark Teahen. Using their Marcel projected wOBAs (.356 for Quentin, .311 for Teahen), over 150 plate appearances that would cost the White Sox about five runs offensively. Is having Teahen hit for 40 games instead of Quentin significant?

What if a team could gain 10 runs (about one win) by using an optimized rather than their typical batting order? We could pick on Teahen and Quentin again and talk about the difference over 300 plate appearance (about half a season), but let’s look at something else. In 2010, the Kansas City Royals’ stud closer Joakim Soria was worth 2.1 Wins Above Replacement. Among the relievers worth about one win less than Soria according to FanGraphs WAR were the following: Kevin Jepsen (1.1), Jon Rauch (1.1), Nick Masset (1.0), and Kyle Farnsworth (1.0). How significant is the difference in value between those decent relievers and one of the top relievers in the game?

How about 15 runs? Let’s use a different kind of example. We’re assuming the “five to fifteen” figure is basically correct for purposes of this post (and I have no reason to doubt that it is). Tango and MGL have often pointed out (following Pete Palmer, as Tangotiger noted yesterday) that even one of the worst imaginable single lineup moves — having the pitcher hit in the cleanup spot — would cost an average of 16 runs a season (about 0.1 runs a game). When just considering lineups in general, that can be looked at as pointing out how we can overreact to single lineup decisions. However, there’s another way of using that thought. If a team’s typical batting (again, not the same as the worst possible order) is 15 runs worse over a season than an optimized lineup, it’s practically the same as having the pitcher hit cleanup all season. If a manager hit his pitchers cleanup all season, would that be something worth getting worked up over?

While the tone of this post has obviously been to the effect that batting order can be significant, you may have noticed something. The “argument” (if it can even be called that) for batting order’s significance is enthymematic, that is, it is missing a premise: that five to 15 runs is significant. Yes, I gave some examples to that effect, but one could just as well argue based on the examples above that what I have shown is that the effect of injuries, the value of “elite” closers versus middling relievers, and even hitting the pitcher fourth are relatively insignificant. I have some sympathy with all those points.

I won’t offer a detailed argument one way or the other. I will note that teams did seem to think a marginal win was worth about five million dollars when signing free agents during this past offseason, so if they were willing to really go for an optimized batting order over the typical one that would be like getting an extra $2.5 to $7.5 million worth of value. Of course, there are other obstacles to implementing optimal lineups as opposed to dealing with the issues above: the reaction of the fans and media, the unwillingness of front offices to impinge too much on the manager’s traditional duties (especially based on stuff like simulations and Markov models), and the likely response of most players.

On the other hand, wins are wins, and money is money. Teams talk big about doing “whatever it takes.” As the league gets smarter, it gets more difficult to find the new market inefficiency, the Extra 2%, as it were. Just as each better move in improving a batting order only adds a tiny bit but can add up to as many as 15 runs (one or two wins), each one-to-two-win-per-season strategy (batting orders, better bullpen usage, efficient platooning, etc.) can add to possibly four wins, and then we’re in expensive free agent territory in terms of value. Complaining about batting order for one game is kind of silly, but that can be said of lots of singular decisions in baseball. Over a full season, consistently using sub-optimal strategies adds up. Is batting order significant enough to analyze and (as is our right as fans) complain about? I dunno. Let’s reconvene after the next time a team misses the playoffs by one game, or, even better after a team puts the pitchers in the cleanup spot for a whole season.

Print This Post

Matt Klaassen reads and writes obituaries in the Greater Toronto Area. If you can't get enough of him, follow him on Twitter.

70 Responses to “How Significant Is Batting Order?”

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

    Dang… looks a lot less insignificant when you put it into dollars: 2.5-7.5 mil!

    I’m sure it’ll evolve like most other things are evolving in baseball – slowly. You’ll see more and more advanced tactics permeate dugouts, but that will be the last of the old school pillars to topple, for sure.

    Vote -1 Vote +1

    • Telo says:

      Or maybe the intentional walk will be the last to fall…

      Vote -1 Vote +1

      • Bronnt says:

        It will continue to remain a legitimate strategy for National League teams when the pitcher is on deck.

        Vote -1 Vote +1

      • Telo says:

        Right. Obviously there will always be instances when it’s correct to issue an IBB, but in probably 80-90% of times it’s currently used, it’s better to just pitch to the batter.

        Vote -1 Vote +1

      • Daniel says:


        Just one more reason for the NL to finally use the DH. There’s nothing exciting or even remotely interesting about watching pitchers struggle to hit .120.

        Vote -1 Vote +1

      • superhans says:


        Yea… I am certainly biased, but I prefer the brand of baseball the AL serves up. Proponents of the NL/No DH will claim that there is more nuance and strategy, while… I would say that it’s just more time consuming pitching changes, holes in your lineup/useless atbats, and chances for the skipper to screw the pooch. The DH just makes sense to me. But again, definitely biased…

        Vote -1 Vote +1

      • Daniel says:

        The heart of baseball is the duel between the pitcher and batter. I have never been able to grasp why someone would think it’s better for one of those batters to be consistently terrible.

        Vote -1 Vote +1

      • PiratesBreak500 says:

        I dunno- personally I love watching Zambrano or Owings bat. Yes, they’re not typical, but I love seeing a pitcher get a line drive hit after an intentional walk.

        Vote -1 Vote +1

      • kick me in the GO NATS says:

        I will go to my grave wondering why anyone would prefer the AL style to the NL style if they had watched both types. In the NL strategy matters far more!!! An 8 year old could game manage an AL team. More relievers get used in the NL because in close games yu have to strategize. Far more useful pinch hits and double switches happen. Teams simply have to be MUCH DEEPER in talent to be good than the AL.

        Vote -1 Vote +1

      • RC says:

        “In the NL strategy matters far more!!! ”

        There’s no strategy to the double switch. Everyone in the stadium knows whats going to happen before it does.

        In fact, I’d argue that late in a game, there’s less strategy. You know the manager is taking the SP out when it comes to his spot in the lineup. In the AL, theres no guarantee how long the manager will try to get by with the SP.

        Vote -1 Vote +1

      • Joel says:

        Not everyone in the park knows what is going to happen. Too, your argument is contrarian but not correct. NL managers have to make a decision whether to keep in a pitcher who is performing well or make the change as his batting slot comes up. In the AL, managers wait, see signs of fatigue, remove the pitcher. In no world can the AL managers task be more complex–the decision matrix for the NL manager by definition is more complex and requires more manipulation–it will always be the AL managers matrix plus one or more options.

        Vote -1 Vote +1

  2. MikeS says:

    I can’t get past the idea of Quentin being hurt. He ahs played 3 full seasons over the last 5 years. He played in 130 games twice! What do you want form the guy?

    Vote -1 Vote +1

  3. PL says:

    Considering 1 run is the difference between a win and a loss, adding 5-15 runs is indeed, very significant by the nature of the game.

    Vote -1 Vote +1

    • tangotiger says:

      No, that’s not true at all. 10 runs is worth 1 win. Whether one wants to call 0.5 to 1.5 wins significant or not is a matter of taste. One man’s significance is another man’s insignificance.

      But 1 win is 1 win is 1 win.

      Undiscussed here is whether players would object, and thereby have their performance change as a result of their egos bruised. Bruised egos may cost more than optimal placement of players.

      Setting all that aside, all other things equal, do things the right way.

      Vote -1 Vote +1

      • RC says:

        “No, that’s not true at all. 10 runs is worth 1 win”

        No, 10 Runs, averaged over 10s of thousands of seasons, is typically worth 1 win.

        10 runs can be worth much more, or much less than that in 1 specific season.

        Vote -1 Vote +1

      • superhans says:

        Are you arguing with Tango? This should work out well for you.

        Yes, in a given season 10 runs may be more or less valuable than 1 win… but…. since we talking generally about the relative worth of a run with no specific frame of reference, we want to speak in Expected Value, or averaged wins per run.

        Who cares about a specific season, when we don’t have a specific season to talk about!

        Vote -1 Vote +1

      • Jono411 says:

        10 extra runs of expectation equals 1 extra expected win over the course of the season. given that we have to make decisions before, not after, seeing the outcome, all we care about is the expectation value.

        Vote -1 Vote +1

      • RC says:

        “Who cares about a specific season, when we don’t have a specific season to talk about!”

        The point is there’s a lot of variance, and when your average pennant race/wildcard seems to be a game or two, something that could swing you on average a game, and sometimes much more, seems to be very valuable.

        Vote -1 Vote +1

      • Telo says:

        “something that could swing you on average a game, and sometimes much more, seems to be very valuable.”

        And sometimes less! That’s why it’s an average.

        “The point is there’s a lot of variance, ”

        Right! There is variance. And when you average it out it comes out to about 10 runs = 1 win. Sometimes 10 extra runs would net you 2 games, sometimes a half a game, sometimes zero games!

        There is no way around it – you can’t claim 10 runs is worth more than a 1 win.

        Vote -1 Vote +1

      • Kris says:

        Many o’ managers have been fired for outliers when regression was imminent. Many o’ managers have been fired because 10 R != 1W over a handful of games or even months. When you go against conventional wisdom, you are not allowed room for regression. There’s nothing in a managers contract that states he can only be fired once the numbers become statistically significant.

        Anyone can claim that X = Y during a set period. You can point to it and say that it was a tiny sample size, but that’s hardly going to save anyone’s job. The weatherman’s right 90% of the time, but everyone loses their mind and remembers the 10%.

        Should we be busting nuts for sub-optimal strategies? Sure, why not. The best part of baseball is arbitrarily blaming events on clubhouse disruption resulting from managerial decisions or terrible personalities.

        1 win is only equal to 1 win if we’re to assume that emotional well-being does not affect performance or is a zero-sum game.

        Just because we cannot quantify the value of a particular win on future events doesn’t mean that the effect doesn’t exist.

        Vote -1 Vote +1

    • Oakland Dan says:

      This reasoning is misguided. 10 runs is far, far more likely to be worth zero wins than 10 wins.

      Vote -1 Vote +1

  4. DonChrysler says:

    While it is true that the most extreme changes possible to a lineup could add or subtract 5-15 runs in a season, I think people are still right to say that we worry too much about batting order. The stat is usually quoted in response to people’s questions about singular batting order decisions, e.g. whether X hitter should hit leadoff, bottom of the order, whatever. The 5-15 run stat basically implies that if the universe of possible outcomes for a batting order only has that size of impact, then singular decisions might not even amount to 1 run over the course of a season. From this perspective, these moves aren’t significant.

    That said, my perspective on the issue has always been that since baseball is often played on very fine margins, why not try to do the right thing with your order? If these types of decisions are nearly cost-free (although Tango’s point is certainly a good one), then why not make them, however small the benefit?

    Vote -1 Vote +1

    • Bronnt says:

      Yeah, I certainly agree with your first paragraph. The relationship of how much attention batting order receives to its actual impact is entirely disproportionate.

      Vote -1 Vote +1

    • Jono411 says:

      “While it is true that the most extreme changes possible to a lineup could add or subtract 5-15 runs in a season”

      Not true. The most extreme changes possible to a lineup could add or subtract more like 50 runs in a season. The 5-15 number is the difference between a typical mlb lineup (so already a fairly well-optimized one) and a truly optimal one.

      Vote -1 Vote +1

  5. Bronnt says:

    I disagree with your argument on many levels. The first is that your average team is going to end up using 60 different batting orders in a season anyway. Projecting the total run differential at 5 games requires (IIRC) the assumption that you’re using the exact same line-up for 162 games. Nobody is doing that. Maybe it’s problematic that managers use so many batting orders-they can be very reactionary-but every team will have injuries, and every team will have players who just need to take a day off.

    The second is that your “bad” batting order isn’t generally that bad. Okay, having the 2007 Juan Pierre lead off while the 2007 Matt Kemp bats 7th or 8th is pretty terrible. But in most cases, your unoptimized batting order is basically that you have Chase Utley batting third instead of second. It’s a very small change-Utley might 11-12 extra plates appearances over 150 games, so you might pick up an extra run. Maybe. But then again, you’re dealing with big league players who have big league egos-keeping their feathers unruffled could ALSO have a very marginal increase in run value. Sure, it’s intangible, unmeasurable, but we’re talking about the tiniest degrees of impact. We’d be ignorant to think that we can measure every thing which provides the tiniest data fluctuation.

    My third point is more specific-you can sometimes manipulate your batting order to take advantage of the mistakes the other manager is making. Maybe having that righty with the .360 wOBA batting 4th looks like a mistake since your .380 wOBA lefty gets pushed to fifth, but you get a little extra value from going lefty/righty/lefty in the heart of your line-up. Either the opposing manager uses three pitchers because he’s a slave to platoon splits, or he leaves in a LOOGY to pitch to a good right-handed batter. Again, it’s going to be a very small, in fact tiny benefit, but we’re talking on the order of tiny benefits over a full season anyway.

    Vote -1 Vote +1

    • Jono411 says:

      “The first is that your average team is going to end up using 60 different batting orders in a season anyway.”

      So? Given the set of players a manager is going to play for a given game, the difference between using the optimal lineup for those players and one a manager would typically use adds up to 5-15 runs over the course of the season, or approximately 0.06 runs each game.

      “The second is that your “bad” batting order isn’t generally that bad.”

      That was the point of this article – your “bad” batting order is 5-15 runs worse than an optimal one over the course of a season. You say that it isn’t “that bad”, but I’d say that $2.5-$7.5 million of lost marginal value IS “that bad”.

      “But then again, you’re dealing with big league players who have big league egos-keeping their feathers unruffled could ALSO have a very marginal increase in run value”

      The only valid point here, but do you really think that batters perform differently depending on what spot in the order their manager puts them? I find this argument very hard to believe and I’ve never seen any evidence that it’s actually true.

      “but you get a little extra value from going lefty/righty/lefty in the heart of your line-up.”

      well yea, but that’s part of optimizing the lineup – never put lefties back to back

      Vote -1 Vote +1

      • Dan says:

        No, your “bad” batting order isn’t that bad. Your optimal batting order is 5 – 15 runs better than the least efficient batting order. The one that most managers use, is not that bad.

        “well yea, but that’s part of optimizing the lineup – never put lefties back to back”

        I am pretty sure that the way you statistically optimize a batting order has nothing to do with handedness.

        Vote -1 Vote +1

      • Dan says:

        Sorry, my reply was supposed to be to Bronnt.

        Vote -1 Vote +1

    • Bronnt says:

      “So? Given the set of players a manager is going to play for a given game, the difference between using the optimal lineup for those players and one a manager would typically use adds up to 5-15 runs over the course of the season, or approximately 0.06 runs each game.”

      People will end up moved around a lot, also. Occasionally, a manager is going to stumble upon an optimized line-up, or at least, something very close to it. Some batting orders will inevitably be less bad than others, and a lot of them will be dictated by the fact that you’ve got some really bad hitters playing instead of your really good ones. It’s more important to have Albert Pujols in the line-up, anywhere, than it is where you put him. And nobody is putting him in the 9th spot anyway.

      Vote -1 Vote +1

      • Jono411 says:

        “It’s more important to have Albert Pujols in the line-up, anywhere, than it is where you put him.”

        obviously…putting pujols in the lineup gives you ~80 more runs, putting him in the right spot gives you maybe 10 more runs.

        “and a lot of them will be dictated by the fact that you’ve got some really bad hitters playing instead of your really good ones.”

        again, obviously…you cant play your best hitters every game, but given the 9 hitters the manager has decided to play a game, there’s no reason not to optimize that group

        i don’t disagree with anything you just said, i just don’t see what your point is…

        Vote -1 Vote +1

      • Bronnt says:

        My point is that I think even the 5-15 run amounts are exaggerated. It’s obviously much less when you consider that certain games, you’ll have very valuable hitters missing. When your relative value of each hitter is very similar, and not very good, on those days, batting order does not matter very much. And teams will have several of those games come up.

        And when you consider that batting order isn’t static, again, it reduces the effect that it matters. Assuming that your projected Opening Day lineup is about 5 runs worse over 162 games than your optimized line-up. You’re actually not going to see a difference in value of 5 runs. Personnel will be shuffled, guys will rest, and the actual value should end up even less. And when you’re talking about potentially less than half a win over a whole season, it just doesn’t matter that much.

        Vote -1 Vote +1

      • Jono411 says:

        “When your relative value of each hitter is very similar, and not very good, on those days, batting order does not matter very much.”

        but it does matter if the phillies hit utley 3rd every game and keep howard in the 4th spot vs. lefties. that pair of moves probably costs the phillies 10+ runs by themselves. the important part isn’t which spot 6-9 do you put each of your weak hitters, but rather that you get your strongest hitters 1, 2, 4 and the next two 3 and 5.

        “Assuming that your projected Opening Day lineup is about 5 runs worse over 162 games than your optimized line-up.”

        it’s not the projected opening day lineup – it’s every single game’s lineup. typical lineups managers put out are around 0.06 runs per game worse than the optimal one. maybe some days it’s 0.15, some days it’s 0.01, but on average, a typical manager will put out a set of lineups over the course of the year that result in 5-15 fewer expected runs.

        Vote -1 Vote +1

  6. Ryan says:

    That was an interesting perspective, I enjoyed reading this.

    Vote -1 Vote +1

  7. jmarsh123 says:

    Depending on the extremes of the personnel involved, batting order can absolutely have a huge difference.

    Anytime batting order comes up, as a Cubs fan I point to one thing. 2005.

    Derrek Lee had a vintage Pujols year (.335/.418/.662) with 50 2B, 46 HR and….107 RBI. Why? Two of the men who hit in front of him the most had OBP’s of .298 and .254.

    It’s hard to wring maximum value out of a cache of solo homers and 2-out bases empty doubles.

    Vote -1 Vote +1

    • Barkey Walker says:

      How is Rbat calculated again? Pujols 2005 line was (.330/.430/.609) and got 9 higher Rbat than Derrek Lee on baseball reference. I guess 0.12 in walks is worth more than 0.43 in iso? (almost identical PAs, 690 for Lee, 700 for Pujols; almost identical SB/CS numbers.)

      Vote -1 Vote +1

  8. BillWallace says:

    This is exactly the same reaction I’ve always had. These batting order analysis’ always conclude “yeah they could get 10 more runs by batter Jeter 9th but that’s only 1 win so who cares?”.

    Well I think $5 million is a lot of money. How much do managers get paid? Would it not be worth it to pay the manager who could get his team to accept an optimized lineup?

    I get that it’s not easy, like it probably can’t be done with my example, Jeter, but $5million isn’t exactly loose change.

    Vote -1 Vote +1

  9. Xeifrank says:

    Has anyone developed a scientifically “proven” lineup generating tool? As far as I have seen so far, everything is theory as far as lineup construction goes. I think you would need a very sophisticated baseball game engine that could run through thousands of different permutations to even have an idea on which couple hundred lineups you should choose from. Has that been done? If so, where is it published with proven results? Such a sophisticated engine would have to also take into consideration the effects of alternating your L/R hitters in the lineup as much as possible. The simple tools out there now don’t do that.

    In summary, if given a high level of certainty that a lineup was going to save you 10 or so runs per season then of course you should use it. I just have my doubts that the tools/theories out there now gaurantee such an answer (and yes, I’ve read the book).

    Vote -1 Vote +1

    • Rally says:

      There are over 300,000 possible lineups. Once you set a lineup, to run it through a simulator you really need several hundred thousand iterations, maybe a million to strip out the variation of runs per game.

      I’ve built simple lineup estimators, but even with faster computers these days it would take me months to run through the possibilities for a single lineup.

      Vote -1 Vote +1

  10. Randy says:

    Not sure if this was covered (didnt see it, but viewing on my phone is not optimal), but the optimal lineup can very depending on which pitcher is on the mound, whether due to handedness or arm slot or other stylistic difference. Many other factors could affect which lineup is optimal on any given day. It’ll never be perfect, but as a fan who sees Leyland throw crazy lineups out there I’m defintely interested in finding a better way to optimize lineups. Good article.

    Vote -1 Vote +1

  11. Hank says:

    What is the error bar on the run difference measurement? Or better yet what is the error bar on the total runs scored for a season before you take the difference?

    I like the use of the models as it generates critical thinking, but I think folks take the output of models as concrete and don’t assess the potential error/limitations of the model.

    Is this model capable of statistically detecting a 5-15 run difference for a team that scores 700-800 runs? So if scenario A (unoptimized lineup) = 750 runs and scenario B (optimized lineup = 760 runs, the model is good enough to consider that delta real and significant?

    If you apply that % change to player statistics, I think folks would view it differently (say swapping nine 2WAR player for nine 2.1WAR players)

    Vote -1 Vote +1

    • RC says:

      The difference between an optimized and un-optimized lineup is 50+ runs, not 10.

      What we’re talking about here is a pretty close to optimized lineup vs. a fully optimized one.

      Vote -1 Vote +1

      • Xeifrank says:

        Right but what tool are you going to use to create a “fully optimized” one?

        Vote -1 Vote +1

      • Hank says:

        Sorry unoptimized was a poor choice of words… an “average” lineup (assuming most managers will not have a pitcher or their weakest hitter hitting 1st, 2nd or 4th)

        Again the model is accurate enough to detect a 5-15 run difference over a 700-800 run season? Without knowing the std dev, I’m a bit skeptical that difference has much statistical significance

        Vote -1 Vote +1

  12. CircleChange11 says:

    I agree with the prmise and the conclusion, and especially the “whatever it takes” philosophy.

    Since the “2%” was mentioned, I would like to request a FG analysis article evaluating the +\- scenario of the Rays playing Zobrist at 1B.

    My initial reaction is that is a big waste of Zobrist’s value within the overall team concept. I would do this myself, but I am concerned that I lack the know how to do it accurately. It does seem to reduce his overall defensive value (significantly) as well as reduce his bat value as compared to say, if he were playing 2B. My impression is that it could costing the team 1-2 wins, which for the Rays could be very significant. Whattya think?

    Vote -1 Vote +1

  13. TheGrandSlamwich says:

    On a very similar note, one thing I’ve always wondered about that I haven’t read any specific statistics on is what is the effect of Tony LaRussa batting pitchers 8th instead of 9th? My gut tells me that it is negligible (especially when Ryan was at short), but I’ve never read any research done about it.

    Vote -1 Vote +1

  14. Sun king says:

    The most annoying thing to me about unoptimised lineups is that lineups are basically the only thing a manager can do to really help his team win. If I pay a guy a million dollars to be my manager, he may as well do everything he can to get those extra wins.

    Vote -1 Vote +1

  15. CircleChange11 says:

    I meant to add that it’s a manager’s primary duty to put a team on the field that gives them the best chance to win. That would seemingly include an optimal batting order.

    I recall Harvy Kuehn (sp?) quote, “The lineup is the easiest job, I haven’t spelled a name wrong yet.” You’d like to think he was just being funny.

    As an aside and at risk of polluting the site with more personal stories…

    (1) My dad just got back from spring training and talked with Theriot, who will be hitting leadiff. He told my dad he’s very happy to be in StL because they aren’t asking him to hit for power or walk more. FWIW. Hoping for a high BABIP season.

    (2) He also got to talk with Rasmus for 15 minutes alone. My dad is not a GM, but has had a lot of experience meeting and talking with pros (caravans and things of that nature. He’s not a great baseball mind, but gets a good read on people. His perception of Rasmus, to build on a point Tango made, was “So much talent, so little confidence.” I found that interesting because I get the feeling that TLR is just waiting for Rasmus to return to the dugout to tell him everything he’s doing wrong. Got a great picture of them talking (arm around each other). Sometimes I think we forget just how young and vulnerable some of these guys are, and how little they’ve struggled up to this point.

    Vote -1 Vote +1

  16. John H. C. says:

    Not to mention that putting together an optimal lineup requires knowing the overall performance levels of a player in advance. Since there a decent amount of variance between a player’s true talent level and his performance level, who knows what the actual optimal lineup will end up being.

    Or to put it another way, the optimal lineup for the Mets might be to have Angel Pagan leadoff and David Wright bat second, depending on how they end up performing this year. But if Reyes is resurgent, he’s probably the optimal leadoff man.

    Vote -1 Vote +1

  17. delv says:

    Rather than “How significant is batting order,” perhaps a better question is “why do we care about batting order so much?” There are certain things that baseball fans obsess over more than others, and batting order seems to be one of them. Maybe it relates to the value of narrative, something that people in general enjoy. Or maybe it’s the idea of “ranking” hitters (in their order), another favorite pastime of baseball fans.

    People don’t generally pay attention to defensive positioning much, but I figure that the traditional layout that baseball teams use currently, while likely very close to optimal, is probably not fully optimized. Here’s a question: was the shortstop (#6 on the field) originally a “short-center?”

    Vote -1 Vote +1

  18. CircleChange11 says:

    Optimal batting orders would also have to incorporate platoon splits and performances/talent against pitcher “types”.

    It also needs to be acknowledged that hitters on a team are going to have various sample sizes (statistical confidence levels) for situations.

    Any output of a system/program is somewhat dependent on the quality of input, and at least for some of the players the input may be lacking.

    Then, there’s luck and noise which complicates it further.

    I’m not sure if we could know the optimal lineup for any given situation. Since player fatigue/health is never/rarely 100 or 0, we also have that difficulty.

    Vote -1 Vote +1

  19. Shaun says:

    Here’s a perfect example of why batting order is what most of us would call significant: Today I did some very basic research into Jason Heyward batting sixth (because that’s what Fredi Gonzalez seems to be planning) instead of batting second.

    Last season Braves’ number two hitters came up 79 more times than their number six hitters. Let’s say a hitter averages around 5 plate appearances per game (which may actually be generous and harm the argument that Heyward should be batting second). Well, that’s 15.8 games worth of plate appearances that Heyward is losing if he moves to sixth.

    We can’t assume Heyward will play every game, so perhaps the Braves only lose 10 games or 5 games worth of plate appearances from Heyward if they hit him sixth versus second. It’s really not necessary for them to lose any games worth of plate appearances, because all Gonzalez needs to do to fix this is bat him at least second in the order.

    Vote -1 Vote +1

  20. Anthony says:

    Wouldn’t how many wins runs are worth be relative to the team? If you have a studly pitching staff, 5 runs is probably a win. If you allow an average of 4.5 runs per game, a 4.50 team ERA, then 5 runs would be a win and optimum lineup would produce 1-3 more wins.

    After that, you have to ask “are 1-3 wins significant”? Well, if you’re in a playoff race then yes, if you’re not then no.

    So the only thing I can conclude is that optimum lineup is important to the Giants, Phillies, and any other contender that’s strength is it’s pitching. To the Yankees 15 runs probably means .5 wins with anyone but CC pitching. It’s all relative.

    Vote -1 Vote +1

  21. Bruce says:

    Since I was a little boy I couldn’t fathom why your best hitter would not hit leadoff, just from the standpoint of amount of at bats.

    Vote -1 Vote +1

    • pft says:

      Most folks like their best hitter to hit with men on base. That way they drive in more runs. RBI’s always change the scoreboard you know.

      Vote -1 Vote +1

  22. Anthony says:

    I get not having your best overall hitter bad first as long as he doesn’t have the highest average. The whole “his average is .250 but he’s fast” leadoff thing is what bugs me. I really do think lineups matter, especially, to pitching strong playoff contenders.

    Vote -1 Vote +1

  23. TLD says:

    Are you guys like, under contract to mention Jonah Keri’s book in every post on this site? It’s getting a little ridiculous now.

    Vote -1 Vote +1

  24. Sox27 says:

    So are you telling me I should expect to see Mark Teahen playing in person this year (sarcastic tone of course)?

    Vote -1 Vote +1

  25. cardhorn says:

    None of the statistical models and lineup run generators can account for protection in the lineup. Let’s say, for example, that the Rangers decide to flip their 2-3 hitters and hit Hamilton 2nd and Andrus 3rd (following the generally-held belief that you should hit your best hitter second). How often is Hamilton going to come up in a situation where it doesn’t make any sense to throw him a strike? A lot. 2 outs, first base open. Perhaps even 1 out, first base open. I’m throwing Hamilton junk off the plate and hoping he will reach for something and get himself out (often he will). However, if Adrian Beltre or Nelson Cruz are hitting behind Hamilton, I’m going to throw Hamilton more strikes, even if it means I have to throw more fastballs.

    Just plugging in stats doesn’t account for the beauty of situational baseball, which helped create those stats in the first place.

    Vote -1 Vote +1

    • Jono411 says:

      there is pretty good evidence that “protection” makes no difference in the results of at-bats. maybe the pitcher does change his approach (or maybe not, have you ever seen any evidence for that?), but the end result is quality of following batter has zero/insignificant impact on how well a batter does.

      Vote -1 Vote +1

      • cardhorn says:

        How is my post evidence of the opposite of what I was arguing?

        I don’t like being “that guy,” but I played a lot of baseball (not quite a major leaguer…), and I think protection definitely has an impact on how a pitcher pitches to the batter…and for most batters, that affects how well the batter does. Generally, great hitters punish mistakes and just do OK on pitches off the plate or nipping the edges of the zone. If a great hitter has no protection, he sees much fewer mistakes and much more pitches outside the zone or on the edges of the plate. Very few are patient enough and have a good enough eye (Bonds) to wait out the junk. Most end up swinging, and the result is either a single or they get themselves out. However, if they had protection, they would get more pitches to drive.

        Vote -1 Vote +1

      • Jono411 says:

        and i’m saying that while your line of reasoning makes sense in theory, in reality the quality of the following batter has essentially no impact on how well a given batter performs – ie you’d expect Pujols to put up a ~.420 wOBA whether he’s being followed by holliday or by molina. if you don’t believe me, there’s a good chapter in Baseball Between the Numbers on it.

        Vote -1 Vote +1

      • CircleChange11 says:

        IIRC, the “average” is that “protection doesn’t matter”.

        I agree, as a former pitcher, that protection does affect how one goes about it. But, whether that is able to be illustrated by data is a different situation, and a difficult challenge.

        When I bring up exmaples of Kent doing much better when he and Bonds had their lineup slots reversed, or when Derrek Lee goes into the tank when Aramis is injured, those are viewed as “exceptions”, and met with the “correlation not causation” phrase … and that might be right.

        But, when we have an “average”, that means for some lineups, there was a positive difference and others there were not. I think it requires a deeper look into who was providing the projection, etc.

        I would not use Pujols for ANY model. Seriously what effect is Holliday going to have? Pujols is going to hit .360-50-160? Pujols is going to hit what he hits regardless of the scenario.

        Pujols is Pujols whether Holliday is hitting 4th, or it’s Duncan/Ankiel/Ludwick, although with the latter there are more IBBs to Pujols.

        It would be interesting to see if stat’s guys smarter than I could look at the groups of lineups where protection seemed to be a positive and where it did not and see if there is any correlations among the group.

        In other words, if a conclusion could be made that “a .370 wOBA experiences no uptick in performance, regardless of protection, however a .340 wOBA hitters experiences a 12% increase in performance if a .365 wOBA hits behind them.” Even then we’re looking at small samples.

        I would imagine that this work has been done, leding to the conclusion that either “protection doesn’t matter” or “protection is difficult to detect”, and I just do not recall the details from memory.

        I will say, as a former player (and current coach), this is one of the toughest “things” to accept from sabermetric analysis. I say this from the experience that who is hitting behind a guy influences how I pitch to the current hitter, and it also affects coaching decisions. So, for it to have “no effect” or something along those lines, goes against how we handle the situation in-game.

        I am definitely open to changing my opinion for accuracy sake. But, as I said it just isn;t one of those things that “makes sense” … but then that’s why the book is subtitles “Why everything you know about baseball is wrong.”

        Vote -1 Vote +1

      • Rally says:

        I’ve never seen evidence of lineup protection existing, and it’s been studied for a long time. The first time I remember seeing it was Bill James looking into the notion that Dale Murphy would not hit so well when Bob Horner got hurt. Didn’t work out that way. Murphy hit as well as ever.

        So assume Pujols is an equally great hitter when pitched to whether he has Holliday, Molina, or a pitcher batting behind him. But there is a big difference in how often he’d be pitched to. He’ll see a lot more intentional walks when he’s not followed by Holliday. That might actually be of benefit to the Cardinals – the intentional walk is rarely a good move even when issued to a great hitter.

        So a proper lineup simulator would have to have some kind of AI to look at the ondeck batter, current batter, game situation, and decide whether or not to issue the IBB.

        Vote -1 Vote +1

  26. GiantHusker says:

    Outstanding article! Best I’ve seen on FanGraphs for some time. It sparked good, thoughtful commentary, also.

    Vote -1 Vote +1

  27. shthar says:

    “I won’t offer a detailed argument one way or the other”

    Then what the hell are you wasting our time for?

    The novelty of the articles on this site is quickly being outweighed by the frivolity of them.

    Vote -1 Vote +1

  28. allen schade says:

    1961 NY Yankees.

    Batting leadoff and batting second were Bobby Richardson and Tony Kubek.

    Batting cleanup was Mickey Mantle.

    That was the year Mantle hit 54 HRs, Maris hit 61. And Berra, Skowron, Blanchard and Howard all topped 20 HRs.

    So why did Richardson and Kubek fail to core 100 runs with all those power hitters behind the? Because they both had anemic OBAs.

    Mantle had 54 HRs and a .400 OBA that year.

    Why not leadoff with Mantle???

    He would gain 54 ABs, and probably hit an extra 4 HRs that year. Plus he would have 10-12 extra hits in addition to the HRs.

    Granted, he would have batted after the 7,8 and 9 hitters so his RBI opportunities would have been fewer. But, then again, Richardson and Kubek provided few RBI opportunities so how many fewer runs would Mantle have driven in had he batted leadoff????

    I think that the fact that each spot in the batting order bats 18 times per season more than the next spot in the order is signifigant.

    The leadoff hitter has 18 more ABs than the number 2 hitter, 36 more than the number 3 hitter, 54 more than the cleanup hitter, 72 more than the 5 hitter etc etc.

    If I had Pujols, I’d bat him leadoff.

    Vote -1 Vote +1

  29. Richard says:

    10 runs does not produce 1 win with regard to “optimized” batting orders vs. inferior orders.

    Remember, there are two teams playing. It is unlikely that one team’s order would be considerably more optimized than the other.

    So if assuming the most optimized order compared to the least optimized accounts for the 15 runs or so, comparing the level of optimization of the home team vs. the visiting team, the resulting difference is even more miniscule.

    Making a batting order is the least significant task a manager has to make and is hardly significant whatsoever.

    Vote -1 Vote +1