What Baseball Data Still Can’t See

Knowledge at Wharton

Ask about Knowledge at Wharton

Episode Description

ABOUT THE EPISODE Baseball has more data than ever, but some of the game’s biggest decisions still cannot be reduced to a single model.

Dan Szymborski, Senior Writer at FanGraphs and Eric Babitz, Director of Player Acquisition for the Milwaukee Brewers, explore how teams use projections, scouting, and player development to make better decisions.

They discuss the changing standards for Hall of Fame pitchers, the limits of forecasting, why some teams outperform projections year after year, and how front offices balance data with human judgment to construct winning rosters.

ABOUT THE PODCAST Sports is a game of numbers. Wharton experts Eric Bradlow, Shane Jensen, Cade Massey, and Adi Wyner team up to tackle the world of sports, from current events to longstanding issues such as: What sports streaks are the most impressive? How do you rank the best players? Can athletes be compared across sports? Moneyball Highlights explains how decision-makers in the game can avoid the common mistakes and embrace the data. Episodes are recorded at the Wharton School.

#BaseballAnalytics #MLBFrontOffice #brewers #HallOfFame ----- The Wharton School of the University of Pennsylvania is committed to sharing its intellectual capital through Knowledge at Wharton, the school’s online business analysis journal.

Launched in 1999, Knowledge at Wharton offers free access to:

- Articles, podcasts, and videos highlighting Wharton faculty research and analysis of current business trends - Interviews with book authors from Wharton School Press - In-depth series of curated content like Wharton Executive Education’s Nano Tools for Leaders - A searchable database of more than 10,000 articles covering all aspects of business - A weekly newsletter that delivers Knowledge at Wharton insights directly to your inbox

Visit the Knowledge at Wharton homepage: https://whr.tn/3qUpSBE Subscribe to the Knowledge at Wharton YouTube channel: https://whr.tn/3DEAgAU

Read the full transcriptHide the full transcript
  1. Welcome, welcome everyone to this week's edition of Wharton Moneyball. I'm Eric Bradwood, Professor of Marketing Statistics and
  2. Data Science here at the Wharton School. I'm joined today by my two long-time collaborators, friends, and co-hosts, Professor Adi Weiner,
  3. Professor Shane Jensen, some combination of the three of us, and our fourth collaborator, Cade Massey
  4. are here every week on the Wharton Podcast Network. Guys, as far as I'm concerned, it's always
  5. baseball season, and we're lucky today to be joined. Actually, we have two guests today, which is
  6. an even special day for us, but in the first half of our show today, we're rejoined because Dan's been on our show many
  7. times. We're joined by Dan Cymborski. Dan's a senior writer at Fangrass. We can see his nice Fangrass logo on
  8. his screen. He's certainly, everyone would agree, a baseball analytics expert, a creator, as everyone remembers, of the
  9. Zips projection system, which forecasts player performance and team success. I also learned from Dan off air, but
  10. it's in the rundown. He's now a first-time voter of the Baseball Hall of Fame, which, you know, Adi
  11. and I and Shane, we have lots of opinions on that. And of course, as I mentioned, he has
  12. appeared on the show many times. Dan, first of all, welcome back to the Wharton Podcast Network, and welcome back to Wharton
  13. Moneyball. Thanks for having me on again, guys. Well, let's just get right from the beginning.
  14. Actually, you know, I'm sure Shane's going to have a lot of questions because, you know,
  15. we've had lots of discussion over the last 12-plus years, but especially in the last
  16. couple of years about how the standards for the Hall of Fame are just going to have to change.
  17. Like, you know, in Adi and my old days, 300 wins, I don't know, maybe 3 ,000 strikeouts, maybe 3,000 hits, you know,
  18. those seem, certainly the 300 wins seem just unobtainable. And Shane did an interesting calculation a number
  19. of years ago, maybe a couple of years ago, about, like, what would be the equivalent
  20. that you would need today to put you in the same, basically, percentile as people in
  21. the Hall of Fame. Like, it's not 3,000 hits anymore, it's probably 2,500, and it's not 300 wins,
  22. it's like 220, and it's not, you know, et cetera, et cetera. So, Dan, we'd love to hear from you,
  23. both as a voter, but as someone that, you know, senior writer at Fangraphs, how do
  24. you think about the Hall of Fame and the changing statistics and the criterion for getting
  25. in? Well, one of the things that people have to consider is when they talk about milestones,
  26. I like to take an approach, and a lot of people do, is that you want to know what the numbers mean, not what
  27. they are. 300 isn't special because it's a number 300 that means the same thing every point in
  28. history, it's because of what it represents. And if we're going to have this standard
  29. of 300 wins gets you into the Hall of Fame, fewer, lets you linger for a while, maybe get in, and you'll have very,
  30. very, very, very, very few pictures starting to get in. And the thing is, 300 wins is not
  31. really a long-term historical standard for the Hall of Fame. You look at it, I mean, a large
  32. percentage of 300 win players were from the 80s and 90s. I think it's between like 1930 and 1967,
  33. there were no 300 win pictures born, but I might be making that up out of the top of my head.
  34. It's a similar number that I can't pull up on demand. And when we look at it, I think
  35. even 200 wins isn't going to be a very easy thing to surpass. I projected this a few years ago, and
  36. at the time, and it's still true, if you projected with the Zips projections, at least,
  37. players' pictures the rest of their career, no one had better than a 50% chance of hitting 200 wins that hadn't already passed
  38. that mark. Shane, please. Well, I guess, because this whole discussion about wins is obviously framed towards, that's a statistic
  39. entirely about starting pitching. And then, of course, you've got the requisite sort of saves kind of metric for relief
  40. pitchers that everybody bases their landmarks on. But obviously, I think there's not going to
  41. be as much of it seems like we're going towards a pitching future where there's less
  42. of a distinction between the starter and the reliever. Is the future just everybody throwing their best
  43. three innings in a game or something like that? And I guess, do you sort of see whether or not, do you kind of, in
  44. your own mind, think about when you think about a Hall of Fame pitcher, do you think the Hall of Fame will change their
  45. kind of view on, could we have a future where some elite middle reliever at some
  46. point gets in the Hall of Fame? And let me just add to Shane's question, Dan, before you answer, does that imply a
  47. different set of metrics that we'll be looking at? Like, maybe it's WIP, or maybe, who knows
  48. what it is? I'm just saying, if the answer is yes to that, which I think it's likely to
  49. be, what are the metrics that we should be thinking about in your mind? I think it's very possible, because one of
  50. the things that was keeping inning totals up for a while was that five-inning requirement
  51. for wins was kind of this magical threshold that was almost sacrosanct in baseball.
  52. And once that's kind of been breached and isn't considered the same, then all of a
  53. sudden there's no problem having pitchers throw three innings, four innings, four and two-thirds innings
  54. per start. And over time, I mean, our metrics are going to have to reflect this.
  55. I think we'll start to go farther away from wins. I don't think that we're going to get
  56. to a point where we say wins are three innings, because then you have to go back retroactively, and it's a whole nightmare.
  57. I do think that we will adjust over time. Generally speaking, for the Hall of Fame, you
  58. can usually get a preview when you look at the year-end Cy Young and MVP ballots, because the way the writers' vote works,
  59. you can vote on year-end awards your very first year as a member, but you need 10 years to vote for the Hall
  60. of Fame. So younger writers, when they join, it takes a while. There's a kind of a lagging indicator of
  61. writers. I mean, I voted my first year in 2016 for Rookie of the Year, and I didn't vote my first Hall of Fame until
  62. last year. It's you'll start to see that happen. Dan, can I just, I mean, one of
  63. the questions that I have relates to a presumption that somehow we should be continuing to
  64. admit pitchers into the Hall of Fame at about the historical rate. That's kind of like maybe a standard.
  65. I'm not sure we should do that anymore. If pitchers are now devaluing themselves as a
  66. role by cutting back on their earnings, by having relievers to do such a narrow role,
  67. maybe we should just get used to the fact that very fewer, far fewer pitchers are
  68. going to be in the Hall of Fame anymore, because they've just divided up their job.
  69. I mean, maybe not. I mean, what do you think? I mean, that's really a question. Well, you're kind of getting into philosophy on
  70. that point, and that's going to differ in the eyes of the beholder. I don't think it's beneficial for the Hall
  71. of Baseball to necessarily have fewer Hall of Famers, because that kind of affects legacy, and
  72. baseball's a very history-based sport. And if you're telling fans today that all
  73. these great pitchers, only a few of these guys you see are ever going to join them, it kind of devalues the present in
  74. a way. And I think from a certain standpoint, if pitchers are doing more max effort in fewer
  75. innings, they're kind of doing the same job, but the job has changed a bit. There's more expected of it.
  76. You don't have – I mean, Christy Mathewson, when he wrote, what was it, Pitching in
  77. a Pinch, I think the book was called, he talked about going easy with men on base.
  78. You can't do that nowadays. Every batter is max effort. There's no Ray Euler in the lineup who's
  79. not going to be at risk to hit a home run. Like, a bad home run hitter now hits 10 in a year, maybe.
  80. So, Dan, I was going to ask you whether you're a peak performance guy or a longevity guy, but we can do this through
  81. some names. Let me throw out some names here. And I know Adi's going to throw out
  82. his two names that he always talks about with the Hall of Fame. I even remember Kevin Brown and Dave Steed,
  83. but we'll get to those in a second. Let's start with King Felix. So, he had a peak that there's a
  84. five-year stretch where it was King Felix. And so, first, a couple of questions on
  85. that, both Dan Zimborski's personal opinion, but also, can you give us a sense also of
  86. what data is provided to the voters when considering people? Like, is it just the aggregate metrics that
  87. Shane, Eric, and Adi can see? Is it more detailed metrics? Is it zips? I mean, what the hell are provided to
  88. you guys that can help people make these decisions? But also, I would like to talk about
  89. King Felix and whether you're a peak or longevity guy. Well, data-wise, you get an envelope from
  90. the Hall of Fame in the mail, and it's just the ballot. You have nothing to go for.
  91. It's like, here's the ballot. Go for it. You figure it out. And so, naturally, you're going to have a
  92. lot of big differences between how writers approach the situation. I do believe that at the Veterans Committee,
  93. people tend to get more of a dossier for the players they're looking at than we do.
  94. We're just kind of expected. I mean, we're ostensibly supposed to be an expert panel.
  95. We should be able to kind of get this stuff on our own. Yeah. If you can't do the research, then why
  96. are you voting at all? That's kind of my feeling about it. What about King Felix?
  97. Oh, I voted for him for the Hall of Fame, and I'm going to again this year. I'm a big peak guy because we talk
  98. about war a lot when we're evaluating, but I'm not sure war is necessarily the right
  99. framework when we're talking Hall of Fame, because Hall of Fame, we're talking more just philosophical
  100. greatness. And even if average players are really valuable, is an average player for 23 years a
  101. Hall of Famer? I would say it's hard to argue in that case. Well, average, no, but maybe better than, I
  102. mean, Andy Pettit would come to mind. Well, I mean, it depends. Well, I did vote for Pettit.
  103. Dan, let's go straight to Andy Pettit now. Oh, I voted for Pettit too. I would have voted for 12 guys on
  104. the list. When you think about peak, are you allowed to sort of stitch together kind of seasons
  105. across? Like, when you think about kind of a player's peak performance, are you thinking about kind
  106. of a continuous block, like Felix's best continuous five years? Or can you kind of pick and choose
  107. across different years? I'm thinking of somebody like Jacob DeGrom when he is eligible.
  108. That guy probably had five years of absolutely peak performance kind of stitched together over 15
  109. or whatever. I don't really have personally as big of a deal as sustained. I think teams have an interest in that,
  110. but I'm not sure it necessarily involves greatness. Like think of your favorite bands and your
  111. favorite actors. It doesn't really matter if they have a bob in between. I mean, you look at the Rolling Stones
  112. discography, there's some pretty, there's some- By the way, you could make an argument, Dan,
  113. to Shane's point. What Jacob, let's pretend that the numbers between DeGrom and King Felix are similar.
  114. I haven't looked at them, but let's pretend they are. You could make an argument that having five
  115. great years spread over 13, 14 years is more impressive than having five great consecutive years.
  116. So, you know, if one thinks about that, from that point of view, which one's rarer?
  117. Probably what DeGrom did, as opposed to what King Felix did. They're both exceptional, but in a different way.
  118. Yeah, right. So, you're basically pointing out that you, how do you define peak?
  119. Max five among your career? Or can you take like a half a season? Like, does it even have to be max
  120. five seasons? Can it be like a half a season, you know, of excellence before they got in
  121. there, like at 23? You know, et cetera, et cetera. I think that's going to come, you know,
  122. to the eyes of the voter. You're not going to get magic bright lines that really tell you five years instead of
  123. seven years or seven years instead of six years. But it's just a general philosophy.
  124. I think about Miguel Cabrera. If you look at his career to 2016, he didn't have all the milestones then that
  125. he did now. But did he really do anything in the last stretch of his career after 2016 that
  126. made him a greater player than he was? Was he the one that won the Triple Crown?
  127. Yes. Yeah. Didn't he do that since 2016? No, that was 2012. Oh, yeah. That was a while ago.
  128. Okay. Okay. This was after he hit the 30s and he was a below average DH for the last five or six years of his career.
  129. And the question you have is, did that really enhance his greatness? Would he have been just as great if
  130. he had retired after 2016? And I have to say, yes. I don't think, you know, playing like Dan
  131. Vogelbach, just to pull out a random name, makes you a greater player than you would
  132. be otherwise. I think I can absolutely buy that. Let me ask you a question. You mentioned that, you know, it's good to
  133. have people going into the Hall of Fame. I think every listener here on Morton Moneyball
  134. knows that I go to the Hall of Fame ceremony every year. Am I going to be by myself the
  135. next couple of years? I don't know. I don't think you'll be by yourself. It depends what happens with the baseball lockout.
  136. I think that's going to create a lot of less interest for baseball. You'll still get people inducted, but the ceremonies
  137. are going to be less. I don't really don't want nothing personally, because look, I'm wearing a Phillies shirt today, if
  138. people that can't see, you know, I really don't want to be there with Chase Utley
  139. and Jimmy Rollins. And I mean, you know, my kids would find that enjoyable, but they're not really Hall
  140. of Famers. Chase Utley is actually, I think as Dan would know, this may be Chase Utley's last
  141. year. I mean, he's been, no, no, no. He's one of those people, he's creeping up on 75%.
  142. No, he's, I think he got the highest percentage of anyone who's on the ballot, who's
  143. returning this year. He did, that's correct. But he's pretty early in his ballot.
  144. Maybe fifth or sixth year. He's got a while to go. He's still, yeah, maybe if you add five
  145. years to, you can't come up for five years, then he's maybe in his fifth or sixth year, because he probably retired 10, 11
  146. years ago, something like that. But it's fun to have exclusive access to Chase Utley.
  147. There's a whole, It's Always Sunny in Philadelphia storyline about one of the characters kind of
  148. stalking Chase Utley. So there's some meme benefits to that. Well, the next question I would ask you,
  149. Dan, is going to tie in well to our guest who's coming in the second half hour of our show, which is Eric Babitz
  150. from the Brewers. You've studied, you came up with the Zips projection system. How did the Brewers do it?
  151. Analytically, what's the magic that you see? Is this something like when you evaluated players
  152. and teams at the beginning of the season, did you have the Brewers as high as they are right now?
  153. I mean, it's not their first year, but how did you think about it? Oh, no, the projections for the Brewers have
  154. been absolutely terrible for the last six years. One of the things I've discovered is that
  155. it's not so much the projections that have underperformed, it's my, it's not that Zips is
  156. dumb, it's that I'm dumb, it apparently seems, because I regularly underestimate some of their better,
  157. more interesting pickups, how much playing time they will get. And they have such a knack for finding
  158. the guys who are outperforming in the first few months of the season, and they know
  159. which guys are going to stick and which ones aren't. And Zips hasn't figured this out, and no
  160. other team has really figured this out, because no other team's doing the same thing to
  161. the projections that the Brewers are. I'd love to say I knew the magic key to the Brewers.
  162. Let me just get this straight. You're basically saying that some of the great Brewer overperformance is due to their ability to
  163. acquire a player mid-season that will continue to do really well, so they're essentially buying
  164. them low? I mean, can you give me an example? Well, I think Jake Bowers is a good
  165. example of this. Not necessarily that they acquire the player in season, it's that they know, they seem to
  166. have a knack for the guys who hit those high there's high percentiles in their projections.
  167. And they rarely make a mistake from small sample size. They seem to know if you look from,
  168. I've done some research in this, it's not as complete as I'd like it to be, but if you look at players who have
  169. like Big Aprils and Mays in Milwaukee, they have a really good knack of knowing which
  170. guys will keep on playing like that and which don't. And that's a really hard thing to do.
  171. There's some special sauce in there that I haven't figured out. And if I did figure it out, then
  172. I would make a lot of money consulting, I think. So, tell me, I've mentioned a few times
  173. in this interview so far about the ZIPS projection system. Can you just remind our listeners, because I
  174. think it's been a few years maybe since you've been on, certainly one or two years,
  175. what is the system and what's the data input? What does it output? And how can it be used to evaluate,
  176. as we've discussed, players or teams? And also, as we look towards the 2027 season, who's high on your list?
  177. Well, ZIPS is a computer projection system. It's a whole bunch of algorithms that work
  178. together. The basic principle of ZIPS is it looks at a player's history, looks at their characteristics,
  179. uses the statcast data, compares players to large cohorts of similar players throughout baseball history, about
  180. 160,000 pictures and 180,000 hitter chunks of their career. And it puts together an ensemble model of
  181. what to expect going forward. It's useful, but you know the old saying that all models are wrong, but some are
  182. useful? ZIPS, I think, is useful, but predicting the future is really, really hard, so you're going
  183. to miss a lot, as the Brewers will demonstrate. What do you consider the magic? I always like to ask this question.
  184. What do you consider the magic of ZIPS? Is it the data inputs? Is it the, as you said, you know,
  185. I assume it's maybe an ensemble of various machine learning and other related types of models,
  186. but what is the magic that makes it useful? Is it that you have access to data that other people don't, or is it more
  187. the way you think about modeling and doing an ensemble type of forecast? I wouldn't describe ZIPS as a product of
  188. inspiration, just a lot of work and a lot of study for it. It just takes a lot of time to
  189. build a model that works with so many in the working parts, because I have models
  190. for how players will decline and when they choose to retire. I have models that will change a player's
  191. decision based on if they're nearing a milestone, so a player who's projected to have 398
  192. home runs will actually be projected for a few more appearances. I've worked with a lot of general data
  193. for injuries, so ZIPS does, on a general level, understand Tommy John surgery. It understands thoracic outlet syndrome and general injuries
  194. like that. ZIPS is just, I always say that someone like Sean Foreman, a baseball reference, who's a
  195. smarter man than me, he has no reason he couldn't build the same ZIPS projection system,
  196. but he would spend a lot of time doing it for not a lot of benefit. It's just that I have spent so much
  197. time digging through the data. I mean, ZIPS is the product of thousands of little mini studies I've done.
  198. I tested some of the configurations. I ran ZIPS for a year on a computer, pretty much nonstop.
  199. It's just calibration, I guess, is the key. I am fierce about calibrating. I'm conservative about doing new things.
  200. I don't just throw things in. I make sure I have years of new data. There's only so much you can kind of
  201. cheat out not having new information to test out. I think just the time involved is the
  202. secret sauce to ZIPS. Can you give us a sense of what you, I mean, it says senior writer, but
  203. like, what do you, not what do you do at Fangraphs, but like, what do you write about?
  204. Like, do you do analyses and write about those? Or do you look at output of like,
  205. call it, I'll call it Fangraph prediction models and write about those. Like, can you give us a sense and
  206. our listeners a sense, because I'm sure they're going to go on to Fangraphs and they're
  207. going to type in your name and they're going to start looking for articles that you've
  208. written. Like, how does that work in what you write for Fangraphs? I very rarely actually write about projections.
  209. I tend to find that kind of writing kind of dry and boring. I mean, I worked for ESPN for eight
  210. years and I wrote just analysis pieces, transaction pieces, news updates. Basically, I'm just a normal nerdy baseball analyst
  211. who happened to put together a projection system, which he could use for things.
  212. And so could you give us an example of like, what are you working on now? And you know, what's the next article to
  213. give a preview to our listeners? What's your next article about? And what was the spark and inspiration for
  214. the particular thing you're working on now? And is there anything different, like in the
  215. off season, like what goes on it? Maybe enlighten all of us. Like, what's interesting to you in November?
  216. Oh, well, I mean, when you're talking with free agencies and projecting, like who's the best
  217. fit for a team, all those kinds of things, you know, I like to take kind of standard type pieces that you see and
  218. then apply projections to them and apply various sabermetric actions. Because I think I'm well qualified to do
  219. that. This time of year, I'm starting to project 2027 and we do the ZIPS team by
  220. team rollout. And that is what pretty much takes up every moment of my free time and work
  221. time between usually the end of the season to tomorrow, because we're actually launching the Brewers
  222. on Wednesday to start things off. But, you know, otherwise, it's a pretty traditional
  223. package of writers for a pretty traditional package of pieces for a writer. I wrote gamers on some of the playoff
  224. games. I wrote previews for series, things like that. Yeah. Adi, please. Yeah. So I have a couple of questions.
  225. The first one, which alludes to something we talked about earlier. So in my estimation, Kevin Brown is the
  226. most robbed home run Hall of Fame, non -Hall of Fame member who's not in because
  227. he's Roger Clemens or et cetera, et cetera. But just a guy who should be in the Hall of Fame, but he's not.
  228. He got eliminated in his first ballot. Is there a path for him into the Hall of Fame?
  229. It's harder to the current Veterans Committee, because generally speaking, the Veterans Committee are not the
  230. type who really look at his numbers. Yeah. I think Brown would do a lot better today.
  231. Yeah. I mean, my God, I mean. I mean, no one I mean, today, no one cares if a if a player is
  232. surly to the writers. I mean, even Curt Schilling almost made it in, and he was considerably sillier than Kevin
  233. Brown was, but we won't go into that. The reason why Brown was robbed was because
  234. he didn't have a good winning record and all those old statistics, but his numbers were
  235. spectacular for poor teams for four or five years with the Padres in the late 90s.
  236. In 2000, he had that years that just used a fantastic peak. And then he played very, very well for
  237. a long time and was very good. He's very obviously a Hall of Famer that I don't know how to correct that.
  238. Do you do you kind of feel like and I'm just I'm just kind of throwing this out there, devil's advocate.
  239. Do you feel like voters in general kind of because, you know, it's the best pictures
  240. park probably in the majors. So is there like almost like an out, like almost anti-Colorado kind of effect going
  241. on here with Kevin? Honestly, I don't think that the writers at the time he was one and out, really
  242. considered park effects too much as a general rule. I think Brown is one of those pictures
  243. who would do a lot better today. I don't think there's any chance Johan Santana gets one and out it again.
  244. I would have absolutely voted for him. Right. I think you're seeing a lot of the
  245. change in the Hall of Fame voting because we now we now call some of the voters who don't cover baseball.
  246. So there's no more instances of a guy who's covered hockey for 22 years and covered
  247. baseball in 1972 still having a vote because there were a lot of a lot of guys like that.
  248. Before we wrap up, let me leave one more question to you. Yeah. Did you vote in the rookie of year
  249. in the American League or in the National League? Not this year. I voted last year.
  250. I voted for the reliever of the war, the new award that we put together for the end.
  251. There's a lot of controversy over, I guess it was Stewart who got who was in for the Reds who got the rookie of
  252. the year who is maybe not even the fourth top four or five. Yeah, he would not have been.
  253. He probably would have been fifth on my ballot, maybe which and I'm not super upset.
  254. I'm not in this because I'm in the Cincinnati chapter of the Baseball Writers Association and
  255. I and they Cincinnati was pretty mad at me because I was the guy who voted for Trevor Rogers over Jonathan India in 2021,
  256. the one writer who did. I mean, he had a lot of RBIs, but it's all he really did.
  257. He didn't have defensive value. He was not a well-rounded offensive player, really.
  258. He had a lot of RBIs, but he also had a guy with a 365 on base percentage in front of him in the
  259. lineup. And I just think that there were better all around players. I think when you look at it, you
  260. have to look at J.J. Weatherholt. You have to look at Carson Benj. I probably would have voted for Benj Weatherholt
  261. because Benj came on real strong at the end of the season. I certainly would have done McLean third and
  262. then probably Logan Henderson fourth. And then for fifth, I think it's between like, you know, Sal Stewart and T.J.
  263. Rumsfeld and Didier Fuentes and Andrew Alvarez, who I keep calling Abe Alvarez, who hasn't pitched
  264. in like 10 years. It is Andrew Alvarez, right? I might've gotten him backwards again.
  265. Let me ask you one last question. What innovation has most impressed you in the world of baseball analytics over the last five
  266. years or what do you see happening? Is it something having to do with the ability of, you know, motion tracking data or
  267. other types of data we didn't have before? I'm just interested, what have you either read
  268. about, used in the last couple of years that you've said, wow, that's really so much
  269. different than when I started out 15 years ago or so? Well, I mean, you're talking generally the StatCast
  270. data is just amazing because when I started writing about baseball, baseball reference was just starting
  271. and baseball reference didn't even update daily. There was a big day in October where
  272. Sean updated the site and baseball reference was updated for the year, which seems so crazy
  273. today. But just what people are doing, some of the approaches they're taking to modeling stuff, you
  274. know, the work that, you know, Sarah's has done, you know, pitch bot, pitching bot, there's
  275. just all sorts of approaches. And I think we probably haven't, we're still kind of at the point where we're kind
  276. of winnowing through all these to see what works and what doesn't, what can be improved
  277. on. And it's fun to look at those things because it's something new, like explaining how offense
  278. works in baseball, that's pretty much solved. So it's not quite as much fun. Questions are fun.
  279. I still want to be like in the room with the team that really figures out how to predict picture injuries through biomechanical data
  280. or whatever they do, because you figure that out, you are printing money in baseball, I
  281. think. Well, Dan, we'd like to thank you for joining us on Morton Moneyball. We've been joined by Dan Szymborski, senior writer
  282. at Frangraphs, creator of the Zips Projection System, and a voter in the Baseball Hall of
  283. Fame. So Dan, thank you for joining us here on Morton Moneyball. Thanks, guys. Always fun.
  284. So that's been the first half of the show. Please stay with us and join us right
  285. after the break. Welcome back. Welcome back to Morton Moneyball here on the Morton Podcast Network.
  286. I'm Eric Bradlow, professor of marketing statistics and data science. And I'm joined by my two co-hosts
  287. today, Adi Weiner and Shane Jensen, some combination of the three of us. And Cade Massey are here pretty much every
  288. week, like we have been for the last 12 and a half years here on Morton Moneyball.
  289. Today, guys, is a very special day. We just had Dan Szymborski from Frangraphs talking
  290. us about baseball analytics. And today we have two guests. So without further ado, since Adi knows him
  291. well and he's been kind of very much engaged with many baseball teams, and certainly Eric,
  292. I'll turn things over to Adi, who's going to run the next interview. Thank you, Eric.
  293. Eric Bradlow, it's my pleasure to introduce Eric Babitz before we turn it over to our
  294. interview section. I'm in my office now. Eric was one of the students who was in my office all the time.
  295. And it's before we actually had a formal Wasabi, before we had our research seminar, before
  296. we had all this institutional stuff that we have here at Wharton now. Eric was part of that founding class of
  297. students that grew us from just... It was Sam in my office before him, and then it was his friend Sam also.
  298. And then we grew up into what we are now. So Eric graduated 2018. He was a student at Wharton.
  299. And he went a couple of summers. I think, Eric, you were at the Phillies for one summer, and I think you were
  300. at the Cleveland... I think they were the Indians at that time for one summer. Baltimore.
  301. Oh, Baltimore. It was Baltimore. And then what else? And then from there, when you graduated in
  302. 2018, you went right to the Brewers, which was... And you've been there ever since.
  303. He's now... His official title is Director of Player Acquisition, which we want to talk about that specifically.
  304. And it's been, I guess, eight years there, and you now have a World Series title
  305. in your belt, which we want to wish you a tremendous congratulations for doing that.
  306. So welcome to the show, Eric Babitz. Great to have you. Yeah, great to be here. Thanks for having me.
  307. Right. So I guess the first thing we want to talk about a little bit is your path.
  308. So tell us... You started here at Wharton, and you were pretty determined to be in baseball.
  309. What are the things that you think worked and would be good for our listeners to
  310. know that? How'd you get to be in baseball? Because I know you studied statistics with me,
  311. but that's not the only thing you did. Yeah. So first off, thanks for having me.
  312. And like you mentioned, I tried to bother you and be in your office as often as possible when I was a student.
  313. When I interned with Baltimore, my boss there had actually read you and Professor Jensen's research
  314. on defense really early on. And had recommended that I reach out to you guys. And I also heard you mentioned Sam, who's
  315. in Miami now, used to have baseball office hours with you. And I decided I just want to copy
  316. that. I wanted to work in baseball. I was lucky to be on a campus where maybe things weren't as formally built out,
  317. like you mentioned, as you guys have now. But there were still a lot of people like you guys who have research interests that
  318. aligned with sports. And a lot of my classmates at the time are now working in sports.
  319. And we all just wanted to try to sneak as much baseball and as much sports into our school and education as possible.
  320. And I think we did that. So yeah, I mean, tried to blend the academic side of things with sports as much
  321. as possible. But there was also a lot of networking. I think it was freshman year, I sent
  322. emails to all 30 teams. I was asking people who worked in baseball just for advice.
  323. You quickly learn how small the industry is. It's grown a lot over the past decade
  324. or so, but it's still relative to most other industries. It's very small. There's 30 clubs.
  325. There's one employer per city. If that, maybe there's a couple with two. But everyone kind of knows each other.
  326. And so some of those same folks that I was able to get on the phone with for like 15 minutes asking for advice
  327. when I was a freshman at Penn, I now see at the winter meetings every year or intern for a couple of them.
  328. And that type of thing was just really helpful in terms of trying to break in initially.
  329. So I had a couple of summer internships with the Orioles. I didn't have a serious playing background.
  330. I played growing up. I love baseball, but I wasn't very good. That was never going to be my path.
  331. And that's fine. But I think in my current role, I have some statistics and data science education from
  332. you guys. But I think coming into the game, the gap was that I didn't play at a serious level.
  333. And so a lot of people who are on the conventional scouting side of things have a much higher level of playing background than
  334. me. And so when I was in Baltimore at the time, this was the previous leadership there.
  335. And it was still a relatively traditional old school type of approach to the game.
  336. And that was great because I got to learn so much like pure baseball, talking to
  337. baseball lifers, learning the scouting side of things, learning that language, and just kind of getting
  338. to understand a lot of the things that I wouldn't have otherwise been exposed to.
  339. And that was really helpful. I mean, for as great as all of the academic resources at Penn and at Wharton
  340. were, they're not going to teach you what a good slider looks like in class. That hasn't gotten into the curriculum yet.
  341. I keep proposing the course, but it's turned down time after time. And so I kind of used my first
  342. couple of baseball internships to try to soak up as much on the pure baseball side
  343. of things as possible, and then try to learn the quantitative side of things and just
  344. the idea of thinking probabilistically and thinking in terms of bets and process versus outcomes and
  345. everything that we could be trained to do from the lens that you guys approach things
  346. through. And that was kind of the skill set. And then talk about like job progression and
  347. stuff. So I just want to interrupt just to make it clear. Eric, we've sent a lot of our students
  348. to teams over the years. You were not the first and you won't be the last, but there's many, many, many.
  349. But one thing that you were substantially different was in that approach. Most of our students, like our most recent
  350. PhD graduate, Kai Xu, just went to the Mets. He's an analyst. We're analysts. We put out analysts for the most part.
  351. And Eric, what you wanted to do is, I'm going to learn all the analyst stuff from you, but I'm going to go into
  352. baseball and try to learn all the other things. And I'm going to probably take a job
  353. that's not an analyst, but I'll know the analyst's playbook and that'll help me a lot.
  354. So I guess that was your plan. You've been now with the Brewers for a while. How'd that plan work out?
  355. Yeah, it's worked out great. I think I was fortunate that at the time that I started, I was a baseball
  356. ops intern. We only had a couple of interns, like one in each department, if that.
  357. And I got to be exposed right away to a little bit of everything. It really aligned with what you said, my
  358. plan, I guess, was. I don't know how calculated it was, but in hindsight, I think it was great.
  359. I've been able to have a generalist background in any meeting or department conversation that I'm
  360. having with anyone in any department or any vertical within the organization. I'm never going to be the most knowledgeable
  361. or the foremost expert on something, but there are very few groups where I would be
  362. completely naive to. I get to be able to speak a lot of different languages throughout the organization.
  363. That's just been really helpful for me. I think even within the last 10 years or so in baseball, the amount of head
  364. count that has come into the game has led to a lot more specialization. And I got to be just ahead of
  365. that, I think, where I got to maintain a generalist approach to things. That's been really, really beneficial for me.
  366. And it's fun. I joke that the scouts probably think I'm too analytical and the analysts probably think I'm
  367. too scouty, but there's a positive way of saying that too. And that's kind of what I've tried to
  368. be. So you guys were very helpful with that upfront. So let me just clarify the jobs you've
  369. had. You started out in scouting, right? Let's go currently. You're currently the Director of Player Acquisition, but
  370. that was not your original role. So just trace us through the arc of your responsibilities.
  371. Sure. And Eric, if you could also tell our listeners here on Morton Moneyball, what does it
  372. mean to be the Director of Player Acquisition also? It would be great. Absolutely.
  373. So I started on all of my titles were just general baseball ops. So it's like baseball ops intern, baseball ops
  374. coordinator, manager, and then we shifted over to building out a formal player acquisition department, so
  375. to speak, over the past couple of years. But I would say throughout my time with
  376. the Brewers, half of my job at least has always had to do with player analysis. I would say that's a combination of traditional
  377. scouting and what some people might think of as analytics. Although me looking at a player's chase rate
  378. or something like that is I'm looking at data, but I've never written code or built
  379. models the way that the much more technical people in the game do now. But I think the player analysis side of
  380. things from my standpoint has always been trying to use all of the really cool tools
  381. that people much smarter than me are building and then try to translate that and complement
  382. anything that might be a gap there through video or through reading other people's scouting reports
  383. and synthesizing all of that information. So that's always been part of my role.
  384. And over the past couple of years, we've built that out where we're bringing more and
  385. more people in to kind of have that hybrid approach to player evaluation while still having
  386. really great analysts and really great scouting staff as well. It's just kind of like a full 360
  387. approach to player analysis. That's always been my approach, and it's always been part of my role.
  388. I've also had responsibilities on the general baseball ops side of things, contracts, a little bit
  389. of salary arbitration rules, transaction planning, things like that, a little bit of exposure to player
  390. development. So can I ask you, without giving away, I'll call it the secret sauce, you talked
  391. about a hybrid approach, which implies to me, I'll make it simple because I only understand
  392. it at a simple level. Let's imagine you have two input sources, although I'm sure you have much more than that.
  393. Let's imagine you have scouts, and then you have, I'll call it the data. And so let's imagine you now have, you
  394. know, whether they, let's imagine they disagree. The scout says X, the data might say
  395. Y. How do you think holistically about, you know, as statisticians, we might say, how much weight
  396. do you want to put on each in coming up with some overall valuation? Without giving away any secret sauce that the
  397. brewers aren't going to be happy about you talking about, how do you think about hybrid
  398. decisions or making decisions with, I'll call it disparate data sources and information?
  399. So I think, I mean, you guys would, would probably agree with this. I think there's a lot of research in
  400. other fields where some of the best decision making frameworks are, you know, you start with,
  401. with an objective approach or a model, and then layer on some subjective human interpretation or
  402. decision making on, on the back end of that. And so I think like, you know, the Moneyball movie and book framed it as like
  403. scouts versus analysts. And that does exist throughout the industry, but it's gotten a lot better.
  404. But I think something that the brewers have, have always been really great at for as
  405. long as, as, as I've seen, I've been here for, for eight years or so is, is that everything's just so collaborative.
  406. Like this, the people who are doing subjective evaluation are very open to, to like learning
  407. from talking about and using the objective information and tooling that we have and then vice
  408. versa. And so we're just trying to get the player right. And so I think it's a non-answer
  409. to your question, but it depends on the player too. Like there's, you know, I, like, I don't
  410. like, like Bryce Harper with, with the Phillies, for example, like I don't need to scout
  411. Bryce Harper to tell you what kind of a player Bryce Harper is. Like he's got so many seasons of, of
  412. data and track record. And we know what he is. We know how players tend to age. We know what he specifically has, has accomplished
  413. over the course of his career. There's other players with, you know, if you take a player who just signed internationally and
  414. is still a teenager, maybe has, maybe he debuted in the Dominican summer league last year
  415. and had like 60 plate appearances. Like, what do we know quantitatively about that
  416. player? We know what he did in 60 plate appearances in the DSL. We know when he was born and we
  417. know what he signed for and like, that's about it. And so you're going to want to lean
  418. on some subjective information a little bit more for a player like that. And those are two extremes.
  419. And then players, there's players across that whole spectrum where you're going to lean on different
  420. inputs and weigh it depending on who the player is. Well, I guess just kind of talking about
  421. that, it does seem like the Brewers from the outside have a secret sauce, which is
  422. separating better than other teams do kind of signal from noise, an emergent player that other
  423. teams have given up on, or, or maybe, or there's tons of examples on the pitching
  424. side of things. Is that, do you kind of feel, and again, you can't give away the secret sauce,
  425. but is, do you think that that represents sort of an excellence or like an edge above other teams on the analytical modeling side
  426. of things or more the subjective kind of, you guys just have the best scouts in the industry side of things?
  427. I mean, I think that the organization is just full of like great people and great
  428. processes like it in all groups. And it's not just acquisition, like it's not just either using analytics or scouts to like
  429. pick the right players. There's also incredible player development accomplishments that the
  430. organization has had over the years. I think, I think we've been very good at letting players do what they're best at
  431. and like amplify that. And sometimes it's, it's not as complicated as people might speculate.
  432. It's, it's just encouraging guys to be the best versions of themselves and empowering them to
  433. do that. And I think we've had a ton of like organizational player development wins in that way
  434. too. But then, yeah, I mean, I think it's both, like, I think it's been good subjective
  435. evaluations. There's been good, you know, analytical and, and modeling based decisions.
  436. And I think there's been a lot of just strong collaborative wins between all of those
  437. things and player development and everything in between that has just led to a lot of
  438. players, the right players getting here and then those players becoming the best versions of themselves.
  439. So I'm going to, I'm just going to follow that up because Dan Zimborski was on our show just before you came on.
  440. And, and we were talking about his ZIPS projection system, which he just flat out admitted
  441. has for six years in a row, greatly under predicted Milwaukee. And we asked him, well, why, why are
  442. you doing that? And how come you haven't corrected it? He says, because I just keep, they keep
  443. signing players who keep performing in the upper quartile of their projection range.
  444. So it's not, I mean, you're supposed to be plus or minus. And on average, you're supposed to be in
  445. the middle, but the Brewers seem to just being in the top quartile and, and they
  446. just do that over and over again. And he said, if I can figure that out, then I would change my, my own
  447. projection. So you're probably not going to tell us what you're doing. And you may not even know what it
  448. is that you are doing. So, so I, and I've spoken to enough people with teams to know that they don't
  449. disclose, but you ever think about, do you ever watch a, a, a decision made by another team and say, oh my God, why
  450. did they do that? Is that, is that something that, that in other words, can you point to mistakes that
  451. get made? And I mean, as I'm, as you know, Eric, I'm still a Yankee fan. I don't work for a team still stuck
  452. on the Yankees. And I'm constantly screaming about what I consider to be bad decisions.
  453. And I just wonder how those decisions are made. And yet your team doesn't seem to make
  454. those. And that, and I just kind of wonder why. I mean, I think at the time you don't know, right.
  455. But you know, another team might make a decision that, that we might not have made
  456. and vice versa. We make decisions that other teams might not make. Like that's what makes it interesting.
  457. Right? Like if there's no information asymmetry throughout the game, then trades and transactions are, are going
  458. to be few and far between. And, and so, yeah, I mean, there's, there's always going to be decisions that, that other
  459. clubs make from a player evaluation standpoint that don't necessarily line up with the way we
  460. might see things, but I, I'm certain that that's the case. And by no means do we have a
  461. perfect track record either. It's just, you know, I mean, on the, on the field, like if, if, you know,
  462. the old saying is like one out of three or three out of 10 gets you in the hall of fame.
  463. And it's, you gotta be a little bit better than that from a, from a transactional
  464. decision-making standpoint. But no, I mean, it's, it's we've certainly made mistakes and, and other clubs have, have
  465. made great moves that, that we probably wish we could have made, or they were right
  466. on a player before we would have realized. Also like there's, it's not just player evaluation.
  467. There's also a strategic component of identifying when there's an opportunity to acquire a player, when
  468. a player might be available, when another team might have a need where, where we have
  469. a sort of like, there's all kinds of strategic decision-making things that we do.
  470. And I think there's other organizations in smaller markets that have had success who for a
  471. long time, like, like Cleveland and Tampa are the classic examples. And hopefully we're, we're starting to, to, to
  472. join that conversation as, as well. But I mean, there's some creative like ways of, of approaching roster construction that we just
  473. have to employ. And it's obviously an advantage to have more resources, but some of those, like, like you
  474. mentioned the Yankees and like some of those like more conventional, like payroll powerhouse type clubs
  475. have more of a traditional like approach to roster construction because they're accessing a certain market
  476. of, of players. And, you know, it's, it's kind of up to us to, to find where the opportunities
  477. are for, for the rest of the players where, that we are able to access. Cause you know, we, we weren't strocker, like
  478. we weren't in on the Otani type free agent bidding a couple of years ago. But there's, there's a whole section of the,
  479. of the player universe that, you know, we can, we can spend our attention on and
  480. we've just been able to have some, some wins there. So Eric, let me ask you a follow
  481. -up to that perfect segue to that. So when you evaluate a player, I think you talked about it, Merrick used a great
  482. term, which I like to use, which is portfolio construction. How much, when you evaluate a particular player,
  483. are you like, is this, I assume the answer is all the time. In fact, it's probably your job.
  484. Like, is this player right for us? In other words, like I could, you could project out someone's stats.
  485. And of course, player A may have better stats than player B, but it's like, it doesn't seem to work.
  486. So how much do you let, let's use the language of statistics. How much do you consider interaction effects between
  487. players? And how do you think about that in terms of just going for, I'll call it
  488. from player evaluation to I'll use your word, Eric, portfolio construction? Yeah.
  489. So I think, I think when you think about like interactions between players, it's easy to,
  490. you're going to just jump right to major league roster construction. But there's actually hundreds of players in each
  491. organization and in each draft, there's 20 rounds in the draft. You have an entire international amateur signing class
  492. each year. You make several trades and free agent signings at the major league level each year.
  493. There's a lot more minor league free agents and non-roster invites that we'll bring in.
  494. And so, yeah, there's always like a set of investments in players that you make at
  495. all times throughout the baseball calendar. And yeah, I mean, from a like, I know that in a portfolio analysis, like framework,
  496. like there's, you're going to want a mix of some higher variance assets and some more
  497. stable assets. And I think athletes in any sport, like in baseball, like we tend to have more
  498. players in the organization. So you have to think about it a little bit more that way.
  499. And yeah, I mean, if, you know, some of our highest rated prospects were taken in
  500. later rounds of the draft, but you're going to take some shots later on guys that,
  501. you know, if in rounds like seven through 12, if you get the right tail outcome of one of those guys, you're probably going
  502. to beat the entire expected value of all of those picks on average over the course
  503. of a draft. And so, yeah, I think sometimes we do think of players, of processes for acquiring players
  504. through that lens. I don't think we're necessarily looking at a specific player and saying, Hey, this guy fits
  505. our portfolio really well. But we are trying to acquire sets of players where hopefully a couple of them end
  506. up getting to the right tail of their outcome. And like I said, that's then when player
  507. development comes in and just we're trying to help every single player get to that, you
  508. know, upper range. And they're not all going to, but we try our best to empower them to.
  509. So Eric, I have a final question. I think we're coming close to the end, but Oakland, the athletics under Billy Bean for
  510. 25 years, they were consistently at the top of their division, particularly normalized for their payrolls,
  511. but they never won a world series. I've always, Eric and Bradlow and I have been criticizing the Yankees for creating a team
  512. that's slam dunk to make to the playoffs, but never get all the way. And hey, we have won in 25 years.
  513. Yes, we have won. Same number as the Brewers, by the way, but go ahead. We're the same as the Brewers.
  514. So you guys did it. So, you know, congratulations Mazel Tov to you as personally, Eric, for winning that world series.
  515. But was there any thought in the roster construction, the way that, you know, it's not
  516. sufficient just to be in the playoffs. We need to to have the ability to win when it's really down, or are you
  517. just saying, just create the best team we can get. And then that's it. The rest is dice roll from then on.
  518. I think you have to balance both. I think, I mean, if you look at like public projections and odds, like whoever the
  519. favorite is each year is still going to, you know, the field is going to be favored over that club.
  520. So of course there's variance involved, but I also, I mean, yeah, like there is a
  521. difference between the type of club that succeeds over 162 games and the rosters, at least
  522. there's a narrative that there's a difference between the rosters that then go on to run
  523. in October. And so I think you have to blend both. It's just like how you have to blend
  524. present wins with future wins when you're, you know, making transactions. But I think, I mean, at the trade
  525. deadline, you might be thinking more about October than you are in the off season when
  526. you're preparing for an upcoming season. I also think there's a lot of alignment
  527. between, like, I think we were fortunate to have a roster that was, you know, very
  528. successful over the course of the regular season and then matched up pretty well in October
  529. as well. And I think like pitching and leverage relievers who perform really well, like tend to be
  530. really valuable in October. But those same guys were really helpful for us over the course of the season.
  531. I think it's easy in retrospect when a team falls short in the post-season to construct the narrative that that team wasn't built
  532. for October. But of course, like that sample is small as well. You know, there was so much noise around
  533. not performing well. If he did that for two weeks during a regular season, it would be a non
  534. -story. He's one of, if not the greatest player of all time. And so I wouldn't say that they weren't,
  535. I mean, they kicked our butts a year ago in the post-season with a very similar roster.
  536. And so it's tough to like exactly know, but you balance both. And I think in the micro, like at
  537. the trade deadline, you know, if we have a certain hole or there's a certain matchup
  538. against some post-season clubs where a reliever might be helpful, like you think about that,
  539. but for the most part, you're just trying to construct the best roster and tends to
  540. be pretty similar in the regular season and the post-season. Well, Eric, we'd like to thank you for
  541. joining us here on Morton Money Bowl. First, as Adi said, congratulations. You know, it's not often that we have
  542. a Penn alum as a world champion, but you're a world champion. And also, I think the way Adi described
  543. it was we tend, you know, because of Adi and Shane's work and many others, we tend to put out analysts and you created
  544. your own path. And I'll always remember for the rest of my life, the way you described it, the
  545. scouts would say, maybe you're not scouting enough. And the analysts would say you're not analytics
  546. enough, but you're the director of player acquisition for the world champion Milwaukee Brewers.
  547. And so you're doing something right. We'd like to congratulate you for that. And thank you for coming on Morton Money
  548. Bowl. Yeah, thanks so much for having me. So guys, this has been another hour here
  549. of Morton Money Bowl on the Morton Podcast Network. On behalf of myself, Adi Weiner, Shane Jensen,
  550. in absentia, Cade Massey. Thanks to our producer, Jacob Brodnick. Thanks to our executive producer, if you'd like,
  551. Dee Patel. Between now and next week, enjoy your sports. Always enjoy your statistics.
  552. We got a lot going on in the world of sports now. So we'll see you next week here on
  553. Morton Money Bowl.

Brief Summary

Dan Szymborski and Brewers executive Eric Babitz explain why baseball decisions still need scouting and judgment alongside projections and data.

Key points

  • What Hall of Fame milestones really mean. Dan argues that numbers like 300 wins matter because of what they represent, not because the same threshold has always meant the same thing. He also notes that even 200 wins may be increasingly difficult to reach.
  • The Hall of Fame’s pitching problem. As starters throw fewer innings and roles blur, traditional pitching milestones may lose relevance. Dan expects Hall of Fame standards and metrics to adjust over time.
  • Dan’s case for peak performance. Dan says he voted for King Felix and favors peak greatness over simply accumulating years of average performance. He questions whether a long late-career decline adds to a player’s greatness.
  • Useful models, uncertain futures. ZIPS combines player history, characteristics, Statcast data, and comparisons with historical cohorts. But even a useful model will miss, because forecasting baseball remains difficult.
  • The secret sauce behind ZIPS. Dan Szymborski says ZIPS is built on thousands of small studies, years of data, and relentless calibration. The time spent testing and refining it is what makes the system work.
  • Predicting injuries could change baseball. Szymborski is excited by the possibility of using biomechanical data to predict pitcher injuries. Solving that problem, he says, would be enormously valuable to a team.

Episode highlights

Tension

18/ 100Calm

Related episodes