Notes on Brian Kenny’s Ahead of the Curve

Brian Kenny’s Ahead of the Curve: Inside the Baseball Revolution was published last July, and I had it marked as a pre-order, but I had to put myself on a little restriction from new books in the second half of last year, so I didn’t get to read it then. I’m finally getting to it now, and it’s a fun read. Fun mostly because Kenny is coming out with his dukes up, and he knows what he’s talking about. I thought I’d put a few notes here, and maybe they’ll spark some discussion or maybe they won’t. My notes are mostly so I can keep track of some of the arguments presented in the book, since the book I’m really looking forward to is Keith Law’s Smart Baseball, which comes out in three weeks.

There’s a chance I’ll change the title of this post later, depending on how things move.

From the publisher’s website:

Most people who resist logical thought in baseball preach “tradition” and “respecting the game.” But many of baseball’s traditions go back to the nineteenth century, when the pitcher’s job was to provide the batter with a ball he could hit and fielders played without gloves. Instead of fearing change, Brian Kenny wants fans to think critically, reject outmoded groupthink, and embrace the changes that have come with the sabermetric era.

Rejecting outmoded groupthink is what I’m all about, so here we go.

28 Responses to “Notes on Brian Kenny’s Ahead of the Curve


  1. Reid

    “Resist logical thought.” Is that really a fair characterization of traditionalists? In my experience, traditionalists and non-traditionalists can behave in illogical ways. If the implication is that using numerical data is logical, while eschewing or not fully embracing such data is illogical, I disagree with that.

  2. Mitchell

    There’s a relevant story in Michael Lewis’s The Undoing Project. Doctors were asked what they look for in diagnosing cancer. Then someone watched their examinations and diagnoses and pointed out that many times, the doctors made their judgments based not on the evidence in front of them, but by instinct or something. In some cases, they pointed to possible explanations that are correct in tiny percentages of cases, revealing a kind of bias in favor of certain subjects they knew more about, or had recently been thinking about.

    So someone asked these doctors to create a checklist of things to look at and consider, and determine a threshold at which a reasonable preliminary diagnosis of cancer could be made. The idea was that this initial checklist would have to be evaluated and refined over time, since cancer seems like such a complicated thing, and a doctor’s expertise is so wide.

    That’s not what happened. The checklist was pretty good in its first draft. And interestingly, you didn’t really have to be a doctor to follow it. As long as you had the patient’s info, you could just go through the checklist and make a reasonable preliminary diagnosis.

    They gave the checklist to a bunch of non-doctors and presented these non-doctors with a bunch of patients’ data. To see if non-doctors with the basic info could use the checklists to flag possible cancer patients.

    These non-doctors, strictly following the checklists, were far more accurate than the doctors themselves — the same doctors who had created the checklist. The doctors were smart, educated, and experienced enough to know what to look for, but then in practice they didn’t follow their own knowledge!

    This is one of the themes of the Lewis book — that even experts in certain fields who knew enough about their subjects to come up with correct answers (when there was such a thing as correct answers) would often be influenced by their biases and give answers that contradicted their own expertise. It happens a lot.

    It’s a relevant thought to keep in mind, since the people who know baseball best, the ones who play it and watch it their whole lives, reject true data in favor of their biases.

  3. Mitchell

    Yes, the accusation Kenny makes is that they resist logical thought, and I’ll get some quotes up about that as I go through the book.

  4. Mitchell

    Also, be careful. Whoever wrote that blurb isn’t saying that people who don’t embrace the new metrics are resisting logical thought (although I think they are). He or she is saying that people who do resist logical thought cite tradition and respecting the game.

  5. Reid

    What I object to is labeling, in a blanket sort of way, those who don’t always accept the conclusions based on data. On the other hand, I think it’s somewhat reasonable to say that those who always reject data-based conclusions are illogical. But it’s also fairly reasonable to say that those who always think data-based conclusions are correct are also being illogical, too.

    I don’t really care for using an logical/illogical dichotomy for this, especially not in a way that labels one group as the logical ones and the other as being illogical.

    He or she is saying that people who do resist logical thought cite tradition and respecting the game.

    I think it depends on what the writer means by “tradition” or “respecting the game.” If “tradition” and “respecting the game,” encompass making decisions that sometimes go against the data, then I don’t think it’s fair or accurate to describe that person as resisting logic.

    I also think that people (at least in football) will embrace statistics/data in a way that seems illogical or not based on sound thinking. As a quick example: in my view, the value of statistics as a way to evaluate performance diminishes significantly with a small sample size and many changing variables. I think most people would agree, but these points are often ignored when football fans use data to make “objective” claims about teams or players.

  6. Mitchell

    On the other hand, I think it’s somewhat reasonable to say that those who always reject data-based conclusions are illogical. But it’s also fairly reasonable to say that those who always think data-based conclusions are correct are also being illogical, too.

    Okay, but read the quote again. It’s not calling anyone a name; it’s saying people who _________ (in this case, resist logical thought) say ___________. I understand where you’re coming from, and there’s a case to be made for it, but you’re aiming it at a quote that isn’t saying what you’re saying it says.

    I think it depends on what the writer means by “tradition” or “respecting the game.” If “tradition” and “respecting the game,” encompass making decisions that sometimes go against the data, then I don’t think it’s fair or accurate to describe that person as resisting logic.

    It doesn’t depend on that at all. He’s saying what these people say. Don’t make connections yet that aren’t quite there. We’re not even in the book yet; that’s just a publicity blurb on a website.

    I also think that people (at least in football) will embrace statistics/data in a way that seems illogical or not based on sound thinking. As a quick example: in my view, the value of statistics as a way to evaluate performance diminishes significantly with a small sample size and many changing variables. I think most people would agree, but these points are often ignored when football fans use data to make “objective” claims about teams or players.

    Yes, I understand this position. I think the book addresses it too, but it hasn’t yet. Let’s find out.

  7. Reid

    OK, I think I misread the blurb you quoted. What the blurbwriter is saying that tradition and respecting the game are some of the reasons people who don’t accept data use to justify this position. Yes? If so, I was misreading it. (I would just note that I don’t think everyone uses these reasons–I actually think they’re the weakest ones, and I think I, personally, use other reasons.)

  8. Mitchell

    Chapter 1 is called “The Herd.” Its focus is on how there is safety in numbers, and how straying from the pack is interpreted by the pack as mocking it. Kenny asserts that this isn’t as true in football, where Bill Walsh was called a genius for his inventive play-calling. He says that Tony LaRussa was also called a genius — as a way to mock him.

    He does acknowledge that the herd mentality still plays a role in football, as with the issue of punting. He cites the 2005 study on punting by David Romer, who concluded that you should go for it on fourth down if (a) inside your opponent’s territory on 4th and 7 or less; (b) inside your opponent’s 33 on fourth and 10 or less; (c) anywhere on the field with fourth and 4 or less.

    In life and in sports, it’s easier to stay within the herd and fail than go out on your own and succeed.

    As long as you win, you can fight off the criticism, but once you start losing, you will not be “one of the guys.”

    Going against the herd is not for the timid.

    Kenny takes this example and applies it to baseball, where he says the equivalent to punting is bunting. I love a well-executed sacrifice bunt. I always have. But I began to realize that my fondness for it (which I still carry) isn’t good baseball when someone — I can’t remember who — explained how Earl Weaver’s big-inning philosophy was better baseball than Whitey Ford’s small-ball. It was sometime in college when I was reading a ton of baseball books.

    Aesthetically, I’m with the small-ball guys. It’s far more fun to watch a guy draw a leadoff walk, then steal second, then get bunted to third, and then be driven in with a sac fly.

    But the data doesn’t support this style of play. I’ll get into Kenny’s explanation later. For now, I’ll just say that it’s probably clear why Kenny’s approach appeals to me. I hate the herd. The herd is right once in a while, and it took a long time for me to be able to accept that and to be mature enough not to reject good wisdom just because it was conventional. My impulse through most of my life has been to reject what everyone else was doing even if it appealed to me.

    But not doing something just because everyone else is doing it is about as blind as doing something just because everyone else is doing it. It’s a lesson I continue to learn.

  9. Reid

    I wrote a post that I suspect will be a counterpoint to some of this. Indeed, I wrote something that I think addresses the claim that there isn’t justification for bunting. In the thread, I talk about the differences between regular season and the post-season in variety of sports. The main difference has to do with pressure–pressure from higher stakes. Additionally, believe that there are some ways of playing that are more resilient to pressure and other ways that are more susceptible adverse effects. I gave examples from different sports, including baseball:

    In baseball, you’d want an offense that doesn’t rely so much on the long ball. Singles are easier to hit in pressure situations than home runs. Also, getting a walk or stealing doesn’t seem to be affected by pressure so much. Moreover, getting walks generally means that you’re willing to let the pitcher throw more pitches, which can lengthen the time a team is at bat—while also wearing out the pitcher (same with hitting a lot of foul balls.) Now, in baseball there is no game clock, but my sense is that a team that stays at bat for a longer period of time can negatively impact the pitcher and opposing defense. The fielding team can get antsy and nervous—as if the hitting team can’t get out. The longer a team is at bat, the more invincible they appear. I suspect this feeling can easily snowball if the team at bat gets several hits. By the way, one of the reasons I’m cautious about using statistics is that it doesn’t capture some of these factors.

    I’d include bunting in this. I suspect bunting is relatively easy at least compared to hitting a home run, and because of this, a strategy that uses bunting may give a team a distinct advantage in the playoffs. For example, compare these two situations: you’re in the later innings in the World Series in a close game. How would a batter feel knowing if winning depended on bunting or hitting a home run? My guess is that there would be way more pressure and stress on the latter. If that’s true, a strategy or style that relied on bunting would have an advantage over a style that relied on hitting the long ball.

    I hate the herd. The herd is right once in a while, and it took a long time for me to be able to accept that and to be mature enough not to reject good wisdom just because it was conventional. My impulse through most of my life has been to reject what everyone else was doing even if it appealed to me.

    Would you feel this even if the herd mentality formed around the data-based approach? (Where I hang out, I definitely think it is part of the herd mentality.)

    Oh, as for the punting issue, one of my objections is based on psychological reasons, and my sense is that statistics-based analysis doesn’t account for this. Once you’re in FG range, failing on a 4th down conversion basically means you’ve given up three points. Now, that’s not entirely true because you could obviously miss the FG. However, I believe many people on your team will interpret the failure in this way. (It could conceivably give a morale boost to your opponents as well.) These things matter to me–and they matter even more in close games, and playoff games. Think about the recent Super Bowl when the Falcons got pushed out of FG range. If they were in FG range, should they still have gone for it? That would seem crazy to me.

  10. Mitchell

    You’re pointing out a difference (in part of this) between playing for one run vs. no runs, as opposed to increasing your chances of scoring at all, and this is addressed in the data I’m going to post soon from the book.

    It’s true that I haven’t yet read anything about the difference between playoff games and regular season games. That might be worth looking at.

    With punting vs. field goals, you are again talking somewhat about the difference between increasing your chances of scoring over the course of a game vs. increasing your chances of scoring at all. I don’t have data on that, but it might be fun to look that up. Obviously, there are times when in baseball you want to play just for one run, as in football there are times when you just want to score any points at all (as in extending a lead late in a game). I wonder what the numbers say.

    If the herd had embraced data-based strategy before I’d considered it, my initial response might have been to resist it, since that’s usually my default. But on the other hand, I’m fascinated by technology and data, so I’m not sure. When I was a kid, I was the sort who would stare at (just to pick an arbitrary example) a bus schedule for an hour just to see what patterns or breaks in patterns there were. I find that stuff super interesting. So I don’t know.

    One thing worth pointing out is that Bill James and Brian Kenny both say it’s not the data they love. It’s the game of baseball. The data helps them see the game better and it makes them enjoy it more, but it was always about the game. This is how I feel, and it’s why I still appreciate a well-executed sac bunt.

    So you didn’t think the Falcons should have gone for the FG? I waver on it from day to day. I don’t think it would have been crazy for them to go for it; I think that would have been safe.

  11. Mitchell

    By the way, walks are considered terrific in the new metrics. That’s why there’s so much emphasis on on-base average vs. batting average. The problem with a bunt is that it gives up an out, bringing you 1/3 of the way to the end of the inning. The problem with a steal (and so far I haven’t read that steals are bad) is that it risks an out. Billy Beane in the movie version of Moneyball hated steal attempts. I can’t remember if he said that in the book.

  12. Reid

    You’re pointing out a difference (in part of this) between playing for one run vs. no runs, as opposed to increasing your chances of scoring at all, and this is addressed in the data I’m going to post soon from the book.

    I suspect thinking that basing decisions on the objective of increasing one’s chances of scoring might be problematic, particularly in the playoffs where playing percentages makes less sense because there are fewer games.

    Also, I’m going to guess that another issue involves the idea that some situation in which points are scored matters, versus the notion that all points are equally valuable.

    So I don’t know.

    OK, just wondering.

    So you didn’t think the Falcons should have gone for the FG? I waver on it from day to day. I don’t think it would have been crazy for them to go for it; I think that would have been safe.

    The game is hazy now, but I believe they didn’t face the decision of kicking a field goal or not. The issue had to do with aggressive or conservative play calling. You’d choose the latter if the priority was coming away with a field goal, but the Falcons decided against that, and they got pushed out of FG range. In the Falcons example, I meant to say that if the Falcons got into FG range and faced a fourth down situation, the Falcons would be crazy to try to convert the 4th down instead of kick the FG, even though, according to the data you cited, they should have tried to go for it on 4th down.

    By the way, walks are considered terrific in the new metrics. That’s why there’s so much emphasis on on-base average vs. batting average.

    Right. I don’t think there’s a conflict in my impression and the stats in this case.

    The problem with a bunt is that it gives up an out, bringing you 1/3 of the way to the end of the inning.

    Right, and the logic makes sense–i.e., you should do you darndest to protect your outs. What I think is being left out is this other consideration regarding pressure–that some things are more resilient to pressure while others things aren’t. If you’re relying on an approach that’s resilient to pressure that’s an advantage and that should be considered, too. This applies to steals as well. To me (and I could be wrong), steals can actually increase the pressure on the opposing team, not to mention possibly distract the pitcher, which diminish his effectiveness. All of this could be heightened in a playoff game in a close game. Steals might be too risky, in relation to giving up an out. But that should be weighed with the pressure/distractions you could put on opposing teams. (By the way, I’m pretty sure Beane does mention hating steals in the book.)

  13. mitchell

    Do you know about Jon Lester, the Cubs pitcher? He’s pretty much their ace. They pitched him in Game 1 of the NLDS (which he was co-MVP of) and Game 1 of the World Series. And he can’t throw to first base. On one hand, I think this puts incredible pressure on him since he can’t really hold a runner (the first baseman does hold the runner, but it’s really kind of a token gesture, because everyone knows Lester is not throwing over there). On the other, maybe since he knows he’s not throwing over there, the steal threat means he has less pressure than another pitcher in that situation.

    I only mention it because he’s a successful regular and post-season pitcher and pretty much anyone who wants to steal on him can just do it.

  14. Reid

    I’ve heard of Lester. You’re saying he literally can’t throw to first base, or that he’s really bad, so the team just doesn’t even bother? This has to be an advantage for the opposing team, especially if the situation is former.

    On the other, maybe since he knows he’s not throwing over there, the steal threat means he has less pressure than another pitcher in that situation.

    But there would probably be more pressure placed on his pitching. Also, if the pitcher and fielders perceive that the hitting team has a pretty big advantage–not only in terms of giving up stolen bases, but that the base runner will have an advantage even if he doesn’t steal–then I feel like this make them anxious as well. It puts pressure on the pitcher to pitch even better. Same with the fielders.

    By the way, my points shouldn’t be taken in an absolute sense. For example, relying heavily on home runs, as an offensive strategy, may be vulnerable to the adverse effects of pressure. This doesn’t mean that a team or player never has any success with this approach. Exceptions exist, but do they prove or disprove the rule?

  15. Mitchell

    I think it’s both. He’s lost his ability to throw the ball reliably to first, so now he doesn’t throw there at all.

  16. Reid

    I would think knowing that you can get a lead or have a good chance of stealing is an advantage–unless Lester is so dominant that this doesn’t matter. But if dominance neutralizes any disadvantage, I would think that puts pressure on him to be dominant. This a disadvantage, especially when the stakes are high, as they are in a the playoffs, and even more in the World Series.

    By the way, would you agree that using statistics, at least in terms of making strategic or tactical decisions, basically involves playing the percentages? That’s the impression I get. If so, I think that’s one of the weaknesses of the approach–especially in the playoffs. The approach works if you have may opportunities–say, a long season like baseball. I would think it wouldn’t work so well in a football, which has a short season. Playoffs are only a few games.

    Another key feature is that playoff teams are generally all good. In the regular season, the range of quality for opponents can vary quite a bit. And ultimately, in the playoffs, you have to beat one team–and that team usually is really good. I also think they tend to have certain types of characteristics. When making decisions about what kind of team to build, what type of strategy to use, my approach has to been to think about a) the type of team one will likely face in the championship game; b) the strategy and type of team that gives you the best chance of beating that team.

    My sense is that this thinking is outside the data-analysis guys. Would say that’s accurate?

  17. Mitchell

    b) the strategy and type of team that gives you the best chance of beating that team.

    I would agree that metrics-driven strategy is largely about playing the percentages, but you offer two considerations here for building a playoff team, and the second one is about playing the percentages. So no, I totally do not think this kind of thinking is outside the data-analysis guys. In fact, it is in these situations where I think the temptation is to move away from the data, when teams should be sticking to it.

    In the example I cite from The Undoing Project, I give an example of how physicians are less accurate with their preliminary diagnoses when they disregard their own data-driven checklists. The book gives multiple other examples of how people’s biases cause them to make decisions against what they already know are the most likely outcomes. They may be right a few times, but they’re wrong far more often.

    Anyway, I haven’t gotten to anything in the book yet (I’m on chapter 7) that addresses the differences between playoff baseball and regular-season baseball. I’m hoping there will be something, because I’ve always been sad that the 90s Atlanta Braves were dominant in the regular season but won “only” one World Series.

    ——-

    Chapter 1 (continued; still on the topic of bunting)

    Based on numbers from 1993 to 2010, with a man on first and nobody out, you can expect to score .94 runs.

    With a man on second and one out, you can expect to score .72 runs.

    Bunting the runner to second gives up an out and decreases the number of runs you can expect.

    —–

    To address Reid’s question (playing to get just one run):

    In the same scoring environment, with a runner on first and no out, you will score 44.1% of the time.

    With a runner on second and one out, you will score 41.8% of the time.

    This means that even if you just want that one baserunner to score, you decrease your chances (slightly) by bunting him over.

    This does not take into consideration other things, like putting pressure on the pitcher, making the pitcher throw more pitches, making the team feel like it has momentum on its side, or capitalizing on the batter’s strengths, so if you think of the difference as negligible, I can see why you might want to go ahead and bunt anyway, if you were playing just for the one run. I kind of think these things even out with a large volume of data, but I can also see what I think would be Reid’s position, especially in a game of high stakes and a just-win-this-one-game situation.

    You might even argue that in a regular season game, you want to sacrifice a guy over so that the team is confident in its ability to do it in the post season, but I’d say that because you decrease (significantly!) your run expectancy, it’s better not to bunt.

    You also have to consider that bunts often fail, so maybe you have to look at Player X’s bunting success vs. Pitcher Y or something like that, but sample sizes are probably way too small to make good decisions that way.

    Since you score 20% more runs when you don’t bunt, in almost any situation I’d say don’t bunt.

    Kenny then shares some stuff about the 2013 post-season Angels, Dodgers, Nationals, and Royals. The Dodgers, Nats, and Royals managers (Mattingly, Williams, and Yost) made some questionable moves for which they were justly taken to task, but Scioscia for the Angels wasn’t, even though in a Game 5, he got the leadoff man on in innings 7, 8, and 9, bunted in each case, and scored zero runs. Kenny’s point (going back to safety in the herd) is that because Scioscia was managing the way the herd would manage, he wasn’t questioned.

    Next up (still in chapter 1) is one of my favorite, favorite topics (as has been discussed here multiple times): closer by committee, and the 2003 Red Sox. 🙂

  18. Reid

    I would agree that metrics-driven strategy is largely about playing the percentages, but you offer two considerations here for building a playoff team, and the second one is about playing the percentages.

    How can that be? Playing the percentages doesn’t make sense in one game or best of seven series. Now, if there are no playoffs, and the best regular season record determines the champion, and you have a relatively long season like baseball, then playing the percentages makes more sense. But obviously in the playoffs, there aren’t many games, so playing the percentages makes less sense. Or am I misunderstanding you?

  19. don

    Game flow needs to be taken into account too. So for example if the game is tied in the bottom of the ninth, looking at the percentages whether to bunt or not should almost be taken at face value. But if it’s the top of the eight, for example, and the ninth hitter is up and it’s a player you want to keep in the game whether it’s a good pitcher or good fielder because you may have to take the field two more innings, then bunting would obviously be a better option.

    I think the other thing to keep in mind is what the percentages are based upon. My guess is the percentages are based solely upon what happen if a player is on first with no outs versus what happens if the player is on second with one out. However depending on the situation the manager of the team in the field may choose different strategies like playing fielders closer to line to take away any doubles (but this opens the field to more singles), play at double play depth, or pitching the batter a certain way. All of this would then skew the data because the data is based on all situations not just the present one.

    Bochy’s championship Giant teams loved to move the runner over. He didn’t really have power hitters and he had great pitching and his team seemed mentally strong. These traits seem to work in favor of that strategy for a team that played well in close games.

  20. Reid

    I kind of think these things even out with a large volume of data,…

    Let me see if I we’re on the same page. If this is true–if those other factors, over many time, become negligible–that is, ignore them and play the percentages, wouldn’t playing the percentages only make sense if you have many attempts in that situation? Or am I wrong about this?

    You might even argue that in a regular season game, you want to sacrifice a guy over so that the team is confident in its ability to do it in the post season, but I’d say that because you decrease (significantly!) your run expectancy, it’s better not to bunt.

    I think you’re dismissing the first point–i.e., being confident in being able to bunt–too quickly. This isn’t a small detail. As a coach, I don’t want to play too differently from game to game. This prevents the players from perfecting what they’re doing, and it can hurt their confidence as you alluded to. Basically, I get my team to play in a way that allows them the most success in the post season. What I wouldn’t want is to have a team play one way in the regular season, and then switch to a different approach.

    You also have to consider that bunts often fail, so maybe you have to look at Player X’s bunting success vs. Pitcher Y or something like that, but sample sizes are probably way too small to make good decisions that way.

    But instead of looking at sample size, couldn’t the manager use their judgment about this?

    One other thing I’d say. In the World Series, a team is likely to face great pitching–starters, middle relievers, and a really good closer. If this is true, utilizing bunting may be a sound approach, giving the team an advantage. My thinking is that you if you’re player are used to and fairly good at bunting, bunting is easier to make contact and put the ball in play than hitting it. (I would think something similar applies to if you’re mostly a contact hitting team, versus a team that emphasizes hitting for power.) So against a great pitching team, an approach that uses bunting and relies more on singles and walks may have an advantage to a team that not only relies more on power, but isn’t as comfortable bunting.

  21. Mitchell

    How can that be? Playing the percentages doesn’t make sense in one game or best of seven series.

    In describing your approach to winning a playoff game (or building a playoff team), you say “the strategy and type of team that gives you the best chance of beating that team.”

    Do you not agree that “the best chance” is a statement of probability, which is playing the percentages? That’s all I’m saying. What I’m saying is true, right?

    But if it’s the top of the eight, for example, and the ninth hitter is up and it’s a player you want to keep in the game whether it’s a good pitcher or good fielder because you may have to take the field two more innings, then bunting would obviously be a better option.

    Yes, this all makes strategic sense, but you have a lot of things to take into consideration that the numbers can help you with. If you need more than one run, for example, giving up one out is horribly not in your favor. If you have someone in the pen who has a decent shot to get out the next three guys, even if you’d like to keep the pitcher in, you should probably do it. Weighing all the probabilities together should give you a better idea of which move you want to make.

    All of this would then skew the data because the data is based on all situations not just the present one.

    Right, but what if you have fielding data, and data on the hitter at the plate? You can take all that into consideration and determine, given this specific situation, what gives you the best chance of success.

    Bochy’s championship Giant teams loved to move the runner over. He didn’t really have power hitters and he had great pitching and his team seemed mentally strong. These traits seem to work in favor of that strategy for a team that played well in close games.

    It’s tough to argue against results. The Giants won, so whatever Bochy did or didn’t do clearly worked. But I would say that you have to be careful of words like “seem.” Appearances are deceiving, as when a ride home from work “seems” to be really long but then you get home and realize you somehow got home earlier and in less time. Baseball has been operating on a lot of “seems like” ideas, but the data is often proving these ideas wrong. Going back to the doctors example above, you wouldn’t want your doctor to make decisions about what might be cancer based on what things seemed like. I think you’d have a lot more faith in what the data seemed to indicate, and you’d want your doctor to go by that. Too many doctors are going with what seems right, not what they know to be more likely.

    if those other factors, over many time, become negligible–that is, ignore them and play the percentages, wouldn’t playing the percentages only make sense if you have many attempts in that situation? Or am I wrong about this?

    It makes more sense over many attempts, because of the regression to the mean thing. But — and here’s where I may never be able to convince you — in an overall large body of data, each individual instance carries the same percentage of results as all the instances combined.

    Quick example from the Lewis book. The odds of your pregnant wife having a boy is 50%. We know this, because worldwide data supports this.

    Which is the more likely birth order in families with six children: B B B G G G , or G B B G B G?

    Most people think the second birth order is more likely, but the truth is that the odds for either birth order are exactly the same. The data verifies this, and the rules of probability verify this. The probability of your next child being a boy is absolutely independent of however many children came before, and how many of the recent children were boys.

    Yes, I know that there are lot of variables, including the health of the players and their psychological conditions and whatever, but I suspect (this is only a suspicion, not yet confirmed by data or testimony I’ve looked at) that the idea would be to start with the percentages as a baseline, and move that line up or down depending on variables that have made a demonstrated difference. But I would highlight that “demonstrated” part.

    As a coach, I don’t want to play too differently from game to game. This prevents the players from perfecting what they’re doing, and it can hurt their confidence as you alluded to. Basically, I get my team to play in a way that allows them the most success in the post season. What I wouldn’t want is to have a team play one way in the regular season, and then switch to a different approach.

    I totally feel the same way, actually. I was trying to make an allowance for seeing things your way, which is to have a team that is capable of making a bunt in the post-season. I personally wouldn’t bother, because the numbers don’t support bunting at all. I wouldn’t bunt in the regular season because it doesn’t improve your chances of winning. I wouldn’t bunt in the post season for the same reason (with the caveat that so far, I haven’t seen anything that says the numbers are different for the post season).

    As for playing against good teams in the post-season, what you say makes sense. But because you’re playing a good team with good pitchers, why would you want to give up an out? Giving an out to Pedro Martinez is insane.

    I hear what you’re saying about the safety of bunts, singles, and walks. I have a feeling (and I can’t back it up) that when you put together a team with a good on-base percentage, you’re not exactly playing for power. You’re playing for getting more guys on the bases. The batter at the plate may have an on-base percentage of .350 when he’s swinging the bat, but what do you think it is when he’s laying down a bunt?

  22. Reid

    Do you not agree that “the best chance” is a statement of probability, which is playing the percentages? That’s all I’m saying. What I’m saying is true, right?

    Hmm, I don’t think so. “Best chance” is a figure of speech, and I’m using it that way. In your opinion, does it have to mean that one would figure out the probabilities and then base a decision off of those calculations?

    Appearances are deceiving, as when a ride home from work “seems” to be really long but then you get home and realize you somehow got home earlier and in less time. Baseball has been operating on a lot of “seems like” ideas, but the data is often proving these ideas wrong.

    Something similar happens with data-based analysis as well. I disagree with the implication that assumptions and blindspots are only a problem with human perception and judgment–and that data-based analysis is significantly free from this sort of thing.

    Yes, I know that there are lot of variables, including the health of the players and their psychological conditions and whatever, but I suspect (this is only a suspicion, not yet confirmed by data or testimony I’ve looked at) that the idea would be to start with the percentages as a baseline, and move that line up or down depending on variables that have made a demonstrated difference.

    One question: in order to assess variables, wouldn’t the factors have to remain constant? In other words, if you have more than one variable that is changing from situation to situation, can you accurately assess the affect of one variable?

    I personally wouldn’t bother, because the numbers don’t support bunting at all. I wouldn’t bunt in the regular season because it doesn’t improve your chances of winning. I wouldn’t bunt in the post season for the same reason (with the caveat that so far, I haven’t seen anything that says the numbers are different for the post season).

    Again, if you’re just looking at the post-season numbers, aren’t we talking about a) a small sample size; b) too many variables to rely on the numbers? If both are true, at the very least, we should temper the degree to which we rely on the probabilities to make decisions.

  23. Mitchell

    “Best chance” is a figure of speech, and I’m using it that way. In your opinion, does it have to mean that one would figure out the probabilities and then base a decision off of those calculations?

    I don’t think you’re using it figuratively. You don’t have to calculate a number to be deciding on a probability. If you think a run-first offense is more likely to beat a gunslinging team, that “more likely” is playing the percentages as you intuit them. You don’t need to base it on hard data or a formula to be playing the percentages. In your mind, you’re deciding that factors X and Y put you in a better position to win. You don’t have to use the language of mathematics to be actually doing math, and any time you decide one act gives you a better chance for a desired outcome than another act, you’re quantifying possible outcomes, whether you’re basing that on gut feeling, experience, the alignment of the stars, or which team has the nicer helmet.

    Something similar happens with data-based analysis as well. I disagree with the implication that assumptions and blindspots are only a problem with human perception and judgment–and that data-based analysis is significantly free from this sort of thing.

    My two examples from The Undoing Project acknowledge that data-based analysis is not free from assumptions and blindspots. But the data is the data; the faults come either in the analysis, interpretation, or decision-making that’s based on it. Nobody is saying (nobody has ever said) that making all decisions based on data is free from biases or errors. My dad used to know the guy at Ford Island who published each day’s weather forecast for the Pearl Harbor area. He said that every day, the guy would look at the numbers, the conditions, and the histories, and make his forecast. And then before pressing “send” to everyone in the area, he would look out the window. Because he knew you couldn’t predict the future based only on hard data. If his forecast was for a sunny day and it poured, the problem wasn’t with the data, but with the forecaster and with the inherent unpredictability of weather.

    One question: in order to assess variables, wouldn’t the factors have to remain constant? In other words, if you have more than one variable that is changing from situation to situation, can you accurately assess the affect of one variable?

    No, and this applies especially when you’re dealing with human beings who are notoriously impossible to predict. But the more you try to isolate and quantify individual variables, the better chance you have of predicting outcomes, don’t you think? A runner on first with Jon Lester on the mound has a lot of things to consider, but wouldn’t you especially want him to know that Lester hasn’t thrown the ball to first base in the past two seasons? The more info you have about the situation, the better chances you have of making a good decision.

    Again, if you’re just looking at the post-season numbers, aren’t we talking about a) a small sample size; b) too many variables to rely on the numbers? If both are true, at the very least, we should temper the degree to which we rely on the probabilities to make decisions.

    Assuming post-season game situations and results are different, sure. Although it’s a minimum of 30 games per post-season (if every series is a sweep), or 300 games over ten years.

    And again, nobody’s saying not to temper how much one relies on probabilities. I might never want to steal bases, but if Jon Lester’s on the mound, I’m going to consider at least taking a big lead and faking my intentions.

    I’ll say this again, in case it’s unclear: I have yet to hear anyone say that all decisions should be made strictly on percentages.

  24. Mitchell

    I’m eventually going to get to the rest of Chapter 1 in my notes. Hopefully today.

  25. Reid

    You don’t have to use the language of mathematics to be actually doing math, and any time you decide one act gives you a better chance for a desired outcome than another act, you’re quantifying possible outcomes, whether you’re basing that on gut feeling, experience, the alignment of the stars, or which team has the nicer helmet.

    So if I make decisions based on gut-feeling that’s the same as basing it on mathematical probabilities? That can’t be right.

    But the data is the data; the faults come either in the analysis, interpretation, or decision-making that’s based on it. Nobody is saying (nobody has ever said) that making all decisions based on data is free from biases or errors.

    No one says this–and everyone I hear will explicitly acknowledge the flaws and limitations in the use of data–especially when they’re asked a direct question about this. But in my experience people behave much differently, often talking about and using the data as if these limitations don’t exist. It reminds of standardized tests scores. If you ask experts about the appropriate use of these scores and their limitations. People will openly acknowledge the limitations and also express a relatively narrow use of the scores. In practice, however, it’s a totally different story.

    No, and this applies especially when you’re dealing with human beings who are notoriously impossible to predict.

    Humans are notoriously hard to predict, which is a big reasons probabilities should be taken with a huge grain of salt. No?

    I also don’t understand why you’re disagreeing with me. Aren’t we basically assessing cause and effect, or something close to it? In such a situation, if more than one variable exists, how can you determine the cause?

    But the more you try to isolate and quantify individual variables, the better chance you have of predicting outcomes, don’t you think?

    Right, the attempt and objective makes sense, but if, in the end, there are two many other variables, then how meaningful will the data be?

    Assuming post-season game situations and results are different, sure.

    Just to be clear: do you think they’re significantly different?

    Although it’s a minimum of 30 games per post-season (if every series is a sweep), or 300 games over ten years.

    Against different teams, with unique players. Game conditions, player injuries are also other variables. In football, you’re talking about 3 or 4 games as well.

  26. Mitchell

    So if I make decisions based on gut-feeling that’s the same as basing it on mathematical probabilities?

    It’s not the gut feeling part. It’s the part where you decide what gives you the best chance to win. If you’re looking at, say, five different possible approaches to beating the Falcons in the Super Bowl, and you decide approach A gives you the best chance, that’s playing the percentages, even if you’re making the percentages up. You’re making a judgment based on what you think is the more likely outcome, and that gut feeling is based on (most likely) an internal sense of what’s worked in the past, or what your players are likely to do. I suppose that if you just rely on some random decision-maker, such as flipping a coin or throwing a dart at a dartboard, that’s different. But isn’t “gut feeling” a way of saying “based on what I know of the game?”

    In practice, however, it’s a totally different story.

    Is there an example of a professional sports team doing this?

    Humans are notoriously hard to predict, which is a big reasons probabilities should be taken with a huge grain of salt. No?

    Yes. We’re not disagreeing here.

    Right, the attempt and objective makes sense, but if, in the end, there are two many other variables, then how meaningful will the data be?

    I suspect (and it’s only a suspicion) that having some info is better than having no info. There’s a stat called FIP, or Fielding-Independent Pitching. A pitcher’s result is hugely affected by how good the fielders behind him are, a variable that seems impossible to factor out. If I understand it correctly (and I haven’t looked at the explanations, so it’s possible I don’t), FIP is calculated using only the pitches that fielders don’t touch. Honestly, I don’t know how useful this stat is — and the commenters admit that there are a lot of stats about which this is true; it’s data, but how useful it is remains to be seen — but it’s one way you can evaluate what a pitcher is doing without the influence of fielders. Obviously, there are a whole bunch of pitchers for whom this stat is meaningless, since their whole pitching game is about “pitching to contact.” But still, it’s an attempt, and it may be worth thinking about.

    Just to be clear: do you think they’re significantly different?

    I honestly don’t know, but if I had to guess and if I had to be right, I’d lean in favor of not significantly different, especially in baseball. When you say “significantly” are you using it in scientific terms? Or do you just mean meaningful enough for the way you’d manage a team?

    Against different teams, with unique players. Game conditions, player injuries are also other variables. In football, you’re talking about 3 or 4 games as well.

    The different teams thing probably evens out over the long haul. I’ve only seen two players I would consider unique in my experience. Barry Bonds and Mike Trout. There are very few unique players in baseball. And there are always player injuries in baseball in the post-season. Again, I’d guess that those things even out, but I’ll remind you that I’m brand-spanking new at thinking about these things — it’s why I’m reading these books, so I can try and make an informed opinion.

  27. Reid

    It’s not the gut feeling part. It’s the part where you decide what gives you the best chance to win.

    Again, I have hard time seeing how this could be accurate. If what you’re saying is correct, then the distinction between the data-based people and the judgment people aren’t really that different. When the scouts in Money Ball evaluated players based on the way they looked, I think we can assume that they associated certain physiques with successful players. That is, if player X looks a certain way, this increases the chances that player X will turn out to be a good player. The data-driven advocates would consider this a totally different way of analyzing the player, right?

    Is there an example of a professional sports team doing this?

    Not that come to my mind, off the top of my head, but I would be a little surprised if we couldn’t find any examples. Also, I’ve seen this sort of thing from fans or even analysts.

    I suspect (and it’s only a suspicion) that having some info is better than having no info.

    Yes–unless people give more weight to the information than the information warrants. Indeed, would you agree that having such information would actually be worse if having the information often lead to its misuse?

    I actually this occurs. Even in this thread, you’ve made the remark–“the data doesn’t support this.” OK, that may be, but if there are too many variables or too small a sample size, then whether the data supports something shouldn’t carry that much weight. And yet, I think people use the data as if it does carry more weight than it deserves (and it sounds like that’s what you were doing as well).

    I honestly don’t know, but if I had to guess and if I had to be right, I’d lean in favor of not significantly different, especially in baseball. When you say “significantly” are you using it in scientific terms? Or do you just mean meaningful enough for the way you’d manage a team?

    I’m not using “significant” in a scientific sense (at least, I don’t think I am). By significant I mean that the differences are meaningful versus negligible.

    If I made a case for this position, I would cite two things, right off the bat

    1. The opponent in that game will likely be one of the best teams. This seems like a big difference between the regular season as many teams are often far from the best. (Also, statistics derived from the regular season are basically an aggregate of teams with varying levels of quality.)

    2. Because of higher stakes, the pressure (and probably the intensity) would be greater in a playoff game than the regular season. To illustrate this, consider the difference between playing a game of catch versus playing a game of catch where someone dies for every ball you drop. Is there a significant difference between the two situations?

    The different teams thing probably evens out over the long haul.

    That might be true, but this point doesn’t seem meaningful to a manager or coach because the playoffs don’t have enough games to be considered a long haul.

    I’ve only seen two players I would consider unique in my experience. Barry Bonds and Mike Trout.

    I’m not using “unique” in this way. You seem to be using unique to mean not only one-of-a-kind, but also one-of-a-kind performer/talent. I’m using unique to mean simple one-of-a-kind. In this way, every player (and situation, if we get technical) is literally unique, novel. No two players or situations are exactly alike. These differences can be negligible, especially if we repeat these situations over and over again. If we’re facing a situation over and over again, then we could ignore the differences. But my feeling is that, in the playoffs, this approach doesn’t really work because there aren’t enough games where one should basically ignore the differences. I think considering the percentages makes sense. But I also think factoring the novel elements of a situation also makes sense, too.

  28. burgess

    Finished Kenny’s book a while ago. Took a couple of weeks to read it, which is indicative of how much I enjoyed it. He does a pretty good job of explaining some of the advanced metrics. I wished he’d gone into more detail of how these numbers are derived, though I’m not sure if I’d have understood it.
    in his chapter on “following the herd,” He makes a good point, for managers, it is better to lose conventionally then to try to win “unconventionally,”but what happens when the unconventional becomes the new convention? Not that advanced sabermetrics is unconventional, because it is an integral part of the game now, I mean every club has adopted advanced metrics in some form, but what happens when playing statistical percentages becomes the norm?
    That said, I agree with Kenny, regarding the sacrifice bunt. Don’t trade an out for advancing a runner one base. If a pitcher is at the plate, i would always bunt, unless that pitcher is nicknamed madbum.
    The only issue I took with the percentages with a runner on first with no outs scoring more than with a runner on second with two outs, is that Kenny never really explains if the runner on second with one out is a result of a sac bunt with a runner on first with no outs.

You can add images to your comment by clicking here.