Christopher J. Phillips is a historian of science at Carnegie Mellon University. His research is on the history of statistics and mathematics, particularly the claimed benefits of introducing mathematical tools and models into new fields. He is the author of "Scouting and Scoring: How We Know What We Know about Baseball" and "The New Math: A Political History," and his work has been featured in the New York Times, Time.com, New England Journal of Medicine, Science, and Nature. He received his Ph.D. in History of Science from Harvard University.
Episode Description
Much of the United States is buried under snow and ice, leaving many dreaming of spring. For some – that dream of spring brings with it a longing to hear the crack of a ball on a bat or the taste of peanuts in a ballpark. With the spring thaw comes baseball season and, with it, the inevitable number crunching associated with the sport. Data and baseball is the focus of this episode of Stats and Stories.
+Full Transcript
Rosemary Pennington: Much of the United States is buried under snow and ice, leaving many dreaming of spring. For some that dream of spring brings with it a longing to hear the crack of a ball on a bat or the taste of peanuts in a ballpark. With the spring thaw comes baseball season and, with it, the inevitable number crunching associated with the sport. Data and baseball is the focus of this episode of Stats and Stories, where we explore the statistics behind the stories and the stories behind the statistics, I’m Rosemary Pennington. Stats and Stories is a production of Miami University’s Departments of Statistics and Media, Journalism and Film as well as the American Statistical Association. Joining me is panelist John Bailer, chair of Miami’s statistics department Richard Campbell is away today. Our guest today Christopher Phillips. Phillips is an historian of science at Carnegie Mellon University. He is the author of Scouting and Scoring: How We Know What We Know About Baseball and The New Math: A Political History. His work has been featured in the New York Times, Time.com, New England Journal of Medicine, Science, and Nature. Phillips is also the General Editor of the Encyclopedia of the History of Science and an Associate Editor for the Harvard Data Science Review. Chris, thanks so much for being here today.
Phillips: Thanks so much for having me.
Pennington: Your book scouting and scoring looks at the history of the use of of data numerical analysis and baseball, why did you feel compelled to write that book.
Phillips: Well I think for many of us baseball has become kind of the paradigmatic example of a new way of knowing that it's the same kind of database way of knowing, taking over from the way things used to be done, which of course the new folks thinks is superstitious or religious in nature. And so we think of these as two very different cases but what I was really interested in is that when I started looking at baseball. It seems like the folks who did it the old way actually had a lot in common with the people who did it the new way. That is to say, they tried to reduce things to numbers. They tried to make reliable inferences on the basis of data. They tried to think carefully about what would be good predictors of the future. And so part of what I was just interested in is why do we think of these as two really different ways of knowing. And actually what do they look like on the ground when you're trying to predict who's going to be a great baseball player.
John Bailer: I liked how you framed it in your, in your writing as scouting versus scoring, you know, could you just define those for us and talk a little bit about why did you sort of glom on to those ideas.
Phillips: Exactly. So those are the, the two groups that are supposed to be enemies or the folks who come in, using the data that is to say the scores who traditionally keep score many people might keep score in baseball games or be familiar with all the stats that are on the back of baseball cards, or you know kind of that they are familiar with from fantasy leagues, and then scores are supposed to be the other end of the spectrum right this kind of almost always male old guys who sit on the stadium and they wear hats and they watch the game and then they choose to gars, and they try and pick out which of the players are going to be successful in the future, and couldn't be different ideas to groups. And so what I'm interested in are the ways in which they're different, but also the ways in which they're similar, you know, the kind of times that they're both trying to find reliable ways of taking past data and predicting the future on that basis.
So, good enough, just a quick follow up to that. So I just was thinking back on all the times I used to do scoring at soccer games. And, and there was there was the subjective components, you know, because there was, there were lobbying by players like I was sure I had an assist on that. Yeah, just the the subjectivity that components there that could play out in terms of how you do scoring. I hadn't really been thinking as much about it until I was reading your article and thinking about the the idea of errors versus hits, and the consequences of some of the subjectivity of it so I was wondering if you could just weigh in a little bit on on that aspect of the subjectivity in the scoring process.
Phillips: Well you know well from when you judge something to be a kind of success for one player it usually means somebody else screwed up. And so, there's always this kind of zero sum nature to most sports scoring. And the way, scoring and baseball was set up is that there, there's always it's the credit to one team as a debit to the other. So when you give a player a hit, what you're saying is that the player deserved to get to first base right you're saying but it wasn't a fault of the other team, but when you're giving a team an error it's actually the player does not deserve to get on base and so one of the interesting things about statistics is that it's kind of miraculous in baseball that we think of these things as objective at all, because at the heart of almost all these judgment calls is a person who is watching the game and deciding who to give credit to. And then, of course, when they're aggregated and compiled and we analyze them then we forget the origin of just the basic statistics is not a fancy statistic it's basic it's just the head, but it comes from people. It comes from a judgment call.
How do you, how do we get, you know, the casual fan or or someone who's interested in the sport to understand that because I do think there are moments where you know people see these stats and they imagine that they are these things that sort of come out of the ether, that frame what is good performance and what's not. And yet sort of forgetting all this subjective things that went into the construction of it, how do we need to communicate that to the guy who's showing up, you know, Reds stadium on a, on a Friday to watch a game to understand like, you know, that's not as objective maybe as you might think it is.
Phillips: Well I think when you watch a baseball game there's a lot of time to chat. And one of the good things you can do is that you watch a play and you like, how did that happen, you know, who deserves to get credit for that you know so when a ball gets hit hard between two players you can blame the pitcher, you can get a credit to the batter you can blame the players for being out of position. And so a lot of these kind of informal conversations, one of the natural results of them is that you might keep score differently than when my wife and I go, well we always debate you know whether it's credited because she has a very high standard. And so people not playing very well and she thinks, oh no he should have definitely gotten that, so I think sometimes it's just, we're so familiar with separating out these numbers as if they have kind of been God given in a certain way, instead of really thinking about the fallible processes by which all data of any kind, are made, but certainly baseball data, but I find it baseball you just start the conversations didn't look like he tried that hard did he on that one.
Bailer: You, you talked a little bit about the idea that that if you push back in time and talked about the people that were scouts or the scores that there were some similarities and differences that you, you know, could you could you talk a little bit about some of the similarities that that came out that might surprise us, perhaps, and some of the differences that that you saw, for sure.
Phillips: So one of the similarities that I think surprises. A lot of people is that the scouts actually from the 1960s onward would assign every skill a baseball player had a number, and then they'd average them together and add them up and so you get an overall what's called an overall future potential, and ofp, and so the reason why they did this was when you're drafting players and professional baseball, you have to draft players. Every team gets a pick and you go through and so you have to have a rank order list. Everyone you're potentially interested in. And the easiest way to create a ranked list is to actually create a single number for each player. And so in scouting Of course they might talk about it as this impressionistic thing but in the end of the day they put a number on every single player. And in fact, the, the fun thing I think about it is that they are in more audacious than the counters you know the baseball stats are mainly counting, but what they're doing is actually quantifying skill and then averaging into an overall skill. And so one of the kind of ironic things is you could make the claim that actually scouts are way more quantifying than baseball cores, in a certain sense.
Pennington: I was telling some friends that I was, and we were going to do this copy of this conversation, and they immediately went to Moneyball. And so, might want one colleague if you can answer this would like to know if now that everybody's money balling if small market teams are going to be competitive. So if you can answer that that would be great. But my question is, I think, I think that the history is ation of the use of data in baseball is really interesting to me because I think that's a general response. When people talk about data in baseball is to immediately go to Moneyball as though data had never been used in baseball before, so I want Why, why do you think is it just the popularity of Moneyball that is sort of framed that or is it just has the sport obscured the data somehow.
Phillips: That's a there's a lot of lot going on there and that question. So, first of all, actually as Moneyball has spread small market teams are actually worse off, you know, the whole premise of Moneyball is that it's a it's a leveraging move, it's an arbitrage, where if you have more information you can actually beat the richer teams but once everyone does the same thing. There's no advantage so in a funny way it's actually ruined the effect. In terms of the book, though. So first I should say, like, Michael Lewis writing Moneyball he tells a great story, and that's that's the key thing is that it's an amazing story, but the problem is that in order to make it an amazing story you have to have a David figure and you have to have a Goliath fit. Yeah, so there has to be a kind of overthrowing of the traditional ways of knowing. And so, when I, when I first read the book. My first reaction was first of all these are not that separate of categories and second of all, people have been doing this a very very long time, and he acknowledges this quite clearly that it's not as if numbers are printed in 2003, it's just that better numbers are circa 2003. And so, you know, I think part of what I was trying to do as a historian and not say a front office executive or, or even say a kind of baseball journalist tracking the latest thing that I was interested in is where did these come from are these old what's new, if it's new and, and I was just really interested in this long tradition, in which baseball is the sport that quantifies, and there's lots of debates oh boy you can get people started on baseball historians I mean they've put to put forward some kind of absurd ideas about it being quantified pastoral sport you know the the bottom line for me is that baseball was invented at the same time that statistics were flourishing so the American Statistical Association is created in the 1830s, and baseball is formalized as a sport in the 1840s. And so in baseball is created in the age in which statistics are created as the way of knowing rigorously about the world, and so they grow up together. And so baseball is a statistical sport not just because it's easy to count things in baseball, but also because it's created in this moment of a kind of flourishing of statistical knowledge.
Pennington: Oh, nice. I you know you mentioned the Moneyball you've you know we could talk about Bill James you go back to sort of a generation before, but but you know as reading your work you talked about the idea of, you know, Henry. Henry Chadwick was a character from the past that this was a name that I had never heard Can you talk a little bit about why this was a person that you would feature and highlight your account, and you're writing about this.
Phillips: Yeah, so, Henry Chadwick is a newspaper reporter from the 19th century he was an immigrant from England. He starts off thinking about cricket and then he moves to chronically baseball in the New York area, New York Brooklyn where a lot of the baseball is being played at that point. And he wants to make the game more manly and scientific, is how he describes it. And the way to do that he believes he's a reporter and the way he does, he does that is he says I need to collect data, and I need to analyze it and report on it. So he sees his role as a reporter and as a score, and as a baseball kind of reformer as intertwines so the way you would do it is you actually collect more data more reliable data, you aggregate it. And so Henry Chadwick stands is this kind of very early 19th century figure that looks to data to solve all the problems with the game. So, we want to know how to change the rules of the game, or collect data. We want to know whether players are good, collect data, and so he's a kind of figure that that comes across even in the 1860s 1870s very early on, for baseball and for what we think of as the data revolution, but he's really a kind of classic 19th century data head, and his half brother is actually interestingly, Edwin Chadwick, who is a public health reformer in England. There's a famous connection that a lot of the public health epidemiological data in England has been collected at the same time as baseball data in America and they actually write letters, where they say I'm cleaning up the streets of London and you're cleaning up the sports. So, it's like, it's a direct connection with some of these more commonly known data and epidemiological stories.
Bailer: So just a quick follow up, you know, now all of a sudden I'm wondering where you know we have sabermetrics is there is there a cricket equivalent, that's a that's a great question.
Phillips: So, data is collected in cricket it. There is no kind of equivalent use of the expansiveness of the data in baseball and cricket, but cricket is, is the way we kept score and baseball was clearly designed after cricket scorekeeping and the role of the umpire in the score in cricket was similarly exalted in the 19th century and so even though the data analysis in baseball goes way beyond a cricket and people again historians boy you give us an edge, we'll go We'll go to an idea about the democratic nature of America versus. I don't know how much I would follow that particular line of thought but it is certainly true that data and record keeping and cricket was intimately connected to early data and record keeping in baseball.
Pennington: You're listening to stats and stories and today we're talking with Carnegie Mellon's Chris Phillips. So, you are a historian of science. And I wonder, for the people who are listening, like, you know like, my brother or my little nephew listen to the podcast now john you'll be thrilled he's like in sixth grade, I'm probably gonna get in trouble cuz that's probably wrong, but, like, for, for people who are unaware, what is the history of science.
Phillips: So a lot of people think of science as exactly the thing that doesn't have a history. It only has a future and what what historians of science are interested in is treating science as a human creation, albeit a powerful central important one, but a human creation, just like any other cultural creation, art politics. And so we're interested in investigating science with questions of not you know what precisely did so and so do on such and such a day but rather, you know, how, how, how does physics explain the world differently at different points in time. How can very smart people one year disagree with very smart people 10 years later, about the essence of the universe you know what happens to scientific theories change, and of course for us living in 2021, how does science, collect to this position of authority, you know where if you want to know about the world you know all the good thinking people know about the world scientifically, well that's not the way we thought about the world in 1900, or 1800, or 1700 and so part of what historians of science are interested in is this particular place that science has in our current culture and in the culture of the past.
Bailer: You know Chris you have this reputation of running with a rough crowd, I've heard you even team teach with statisticians. I mean, that could be the end of many reputation, I know that. Rosemary feels damaged as Richard. Brett. I could fill a volume. So I'd like to follow up a little bit with, you know, the course that you're you've been teaching with one of your colleagues at CMU. Absolutely. So
Phillips: One of my colleagues Joel greenhouse and I were talking because my current book project is on the history of statistics and medicine and he's a biostatistician that statistician who works in epidemiology and the Allied fields of medicine and so we were talking about the kind of trajectories and and statistical data in medicine or in health. And one of the things we quickly realized is that we're actually interested in very similar questions we're interested in questions about how you use data to make reliable inferences we're interested in questions about what's the context of data. What is the meaning of data, how does that change over time. We're interested in telling convincing narratives using particular sorts of data. And what what data is not that interesting. And so we realized that actually, we could teach a course for undergraduates and actually first year undergraduates, that would introduce them to statistics and history not as these opposed disciplines where one involves books and essays and the other involves, you know, sort of, long lectures and problem sets, but rather that both of them are disciplines that are fundamentally interested in this question about making inferences from data.
Pennington: Oh, sorry. I'm gonna jump in again I so I in my own work, I feel like I straddle like the humanities and social sciences I was trained as a social scientist. But I increasingly sort of embrace a more humanities, I think bent in my work, and it's always sort of difficult to navigate those spaces, will be depending on what audience I'm talking to and I wonder given that you've been teaching this class now and thinking about these issues. You know, what can we as academics do to try to help. You know, it's a huge question but like try to help do tear down that divide a bit because I think it does make it difficult to sort of see the ways that we are interested in similar questions like what is data, how do we know data, where does it come from like those are questions that you know again a historian and a statistician are interested in right, but may not always know that they're interested in the same things because they're not speaking to each other or even sometimes the same language.
Phillips: Right. And we really spend a lot of time clicking what we find in language that ensures nothing but our experts and our fellow experts could ever read it. And so part of it is is a communication issue and I think one of the ways that we found around that is by finding concepts that crossover really easily so alternate explanation is a concept that in statistics, you know if you're trying to make causal claims alternate explanations are really important right you want to think about what would be other explanations for this data, right why, why are you believing, a particular explanation or particular account of the data, but for historians alternate explanations are, are the kind of bread and butter that you operate off of to because history only runs once it's an N of one. And so if you want to think about how it might have run differently, you have to think about alternate explanations. And so, for, for us concepts like that, which are, are, they're not formal in many cases and so you don't have the kind of formalism or the the the jargon heavy nature of it but rather they're concepts that move very easily between fields. And so we try and emphasize those concepts, when we're teaching together. And so we can each say well what would be an alternative explanation and we can do it from a historical perspective or from a statistics perspective,
Bailer: Yeah that's that's really fascinating. I mean, I love this discussion of the idea of of kind of the, the, the, the fact that there's measurement involved in what we do we just can't take it as a given, and that's been part of you know that's that's a common touchstone for many of our discussions. I had a great collaboration many years ago with a Russian historian colleague friend and he was trying to understand sizes of things like gulag populations. And, and all you could get were kind of these other ancillary measures that were not direct and sort of just just thinking through, you know, what does that mean how does that, how would that inform you about this question that's, that's really incredibly core for what you're doing as historians I, I really find this fascinating what are what are some of the things that that surprises the students the most yeah just curious when they go, I mean first they're going, what the heck are they decide, you know, with all the other classes filled I got stuck with this history and stat guy and a class, you know. But what are they surprised to hear surprised to learn with you in this class.
Phillips: Yeah, I'm sure some of our students are a little surprised by the class but it just sticks in data science crowd here at Carnegie Mellon so yeah history than they are by AI that's okay. I'm okay to try and find converts it's always easier for me that way. So the one thing that people tend to be surprised at is how old trying to make sense of data really yeah they're shocked that data predates computers, yeah. Part of this is we do a really bad job of communicating that what computers do in some senses replicates the work that was done before in some senses it's different. And these are complicated historical questions but, you know, we start the class with the bills of mortality. 17th century London. Now these are for historians, these are very you know typical kind of documents you would use, but the students look at them and they say oh, you know, these are interesting questions now you can ask you can ask about the distribution of disease you can ask about seasonal differences because they're collected weekly, you can ask questions about the difference between say measures of christenings and measures of birth, and so you can, all of a sudden a lot of the kinds of material that in a stat class you're going to get the latest or you're going to get these classic data sets, you know from the last 50 years. They're really surprised to see that people in the 1700s or even in the 17th century are really grappling with some of the same questions that they are today. So I think that's one of the most surprising things about it, I think another surprising thing is that statistics is not always confirmatory testing. Yeah. No, and I think one of the great virtues of the clerk course, excuse me is that we spend a lot of time exploring data, and not necessarily jumping right to making assumptions and running tests, and there's so used to and so many statistics courses trying to get to these tests to try to get to a kind of answer that they can give that the fact that we are not interested and will never be interested and then kind of getting to that answer I think is really surprising to them.
PenningtonSo you've written a book on books. Books stats baseball stats, a book on on the new math, and you mentioned this other project that you are working on that seemed like it had an epidemiological connection. How do you when you're deciding what you're going to work on. Choose your path.
Phillips: That's a good question. I'm glad you didn't see some of the other stuff like wine tasting. On one hand I write what I'm interested in i think that's that's a safe thing that most historians do, but the. When I think of the overarching questions that I'm interested in. One of them is, how does a field that defines itself as non mathematical or non quantitative quantitative. And so these are really interesting moments in the history of science. When a certain way of knowing about the world changes, and all of a sudden conceptions of what counts as rigor or what counts viable change. And so historians of science are often interested in these moments when you have a transformation, and what counts as reliable and rigorous knowledge and and for me, I think, you know, growing up in the late 19th early, excuse me the late 19th 20th and early 21st century. You know, I think, to me, or thinking about the last six months in the pandemic and data about the pandemic I mean we live in a world where data is put forward as a reliable way to go about things and so for me I was just interested in these fields that for so long define themselves as not numerical or as not mathematical, and some of them are fields now that we forget like experimentation, for instance, yeah, the kind of non mathematical field for a long time, or thinking about literary analysis or archeology, there's lots of fields where mathematics is now commonplace but in the past was not so those are where I kind of hone in on.
Bailer: I think that's a I had an epiphany when I was in grad school reading things like Stephen Jay Gould or, you know, to and thinking about kind of the construction of what what we think of as certain concepts that there's there's a cultural context in which work is done, and then our own values and perspectives are often layered on that, and you know it's it's very different to think about where, where information comes from than the think mechanically about how you process the the outcome of such measurement. And I really love that, as you were describing your course that this focus this emphasis on conceptual statistical thinking is as really being sharply in contrast with procedural mechanics. And what's really a sort of an aspiration that sounds like it sounds like a great class I wish I could take your class.
Phillips: We'd love to have you in it sometime. You know, I think it's exactly right that we, we tend, as humans, in many cases to want to move to something mechanical something that we can just follow the steps on right something that if we do what we're told to do we'll get the right answer. And, of course, none of the interesting questions or meaningful questions are important questions are ever answerable in that way. And so one of the kind of great virtues I think of having people come from very different disciplinary backgrounds, is that we absolutely agree that if it's an important enough question it's not going to be able to be answered by running an algorithm.
Pennington: Well that's all the time we have for this episode Chris, thank you so much for being here today. Thanks, Chris.
Phillips: What a great conversation. Oh my pleasure fun
Pennington: Stats and Stories is a partnership between Miami University’s Departments of Statistics, and Media, Journalism and Film, and the American Statistical Association. You can follow us on Twitter, Apple podcasts, or other places you can find podcasts. If you’d like to share your thoughts on the program send your email to statsandstories@miamioh.edu or check us out at statsandstories.net, and be sure to listen for future editions of Stats and Stories, where we discuss the statistics behind the stories and the stories behind the statistics.