Can We Measure Skill vs. Luck in Sports?

Knowledge at Wharton

Ask about Knowledge at Wharton

Episode Description

ABOUT THE EPISODE What separates skill from luck, and can we actually measure the difference? Gregory Bond, Chief Investment Officer and Head of Americas at Man Group, joins Wharton Moneyball to discuss his research on competitive balance across sports, measuring skill versus chance, and the surprising parallels between sports analytics and investing. The conversation explores everything from the NFL and NBA to European soccer, along with how investors can think about performance, prediction, and uncertainty.

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.

#sportsanalytics #sportscience #statistics #sports ----- 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 to Wharton Money Ball. Welcome to a full hour of sports analytics
  2. here on the Wharton Podcast Network. This is Cade Massey hosting this week's show
  3. along with two of my longtime collaborators, colleagues and good friends, Adi Weiner and Shane Jensen.
  4. Adi showing up from overseas and very late in the day as he's been doing the last couple of weeks and will for a
  5. couple more. Shane from Philadelphia and Eric is having some technical challenges and he may slide in here.
  6. He may not slide in here. We're gonna run the show the way we have been running it more weeks than not
  7. lately and that is have a guest here for the first half and then we'll do open lines in the second half.
  8. It's a terrifically interesting time of year. This is like championship time of year.
  9. There's no more championship rich moment in the year than probably this week. So we'll dive into that in the second
  10. half of the show. In the first half of the show, we have a guest, Greg Bond is here.
  11. Greg is the Chief Investment Officer at Mangroup. Mangroup is one of the world's largest alternative
  12. asset managers and Greg's been there for some time. I met him first years ago when he
  13. was Head of R&D for one of their subsidiaries, ManNumeric, I think, but he is their CIO now.
  14. He is also Head of Americas for Mangroup and he's the Portfolio Manager for their flagship
  15. multi-strategy fund. He's typically based in Boston. He is based in Boston, typically in Boston,
  16. but he's in London joining us from London. Greg, welcome to the show. It's great to see you guys.
  17. I really enjoy this. This is always find a lot of sort of analogies between sports and investing and sometimes
  18. it's a lot more fun to do the sports part, to be honest, so it's all good. Well, we're impressed that you do anything in
  19. addition to the job that you have with Mangroup, but we're glad you do. It should be said, and we should find
  20. out a little bit about this. My impression is that you've done a little bit with Major League Baseball.
  21. At least a number of years ago, you got involved. I'm not even sure if I'm allowed to
  22. say what team, but you were a little involved with one of the teams. Yeah, with Red Sox, I reached out.
  23. I was working at Numeric at the time, Mangroup, and reached out to Tom Tippett, who
  24. was in the alumni database, wondering if I could do some kind of sabbatical from my
  25. day job. And he said, sure, come on over. And I was an intern, punching my little
  26. card every day, working on their draft strategy. So they had a bunch of these draft
  27. reports from the scouts and could we turn them into something quantitative. And then I didn't quite realize until the
  28. very end at the last day, I showed up to present to everybody and they brought in Bill James.
  29. Apparently they gave him the same data, but unfortunately he hadn't done the work yet.
  30. So we couldn't go head to head, but then he wrote me a memo later telling me how wrong I was on a bunch
  31. of stuff, which is great. We had a really good time, but it was good to just all of a sudden
  32. be in that room and kind of have a little bit of a head-to-head going on. It was good.
  33. That's a ball. Was Theo Epstein running the team in 2011? So 2011 was the famous fried chicken in
  34. the dugout stuff, where when I joined, they had a 99% chance of making the playoffs.
  35. And when I left, they hadn't made the playoffs. So I don't know if there's causal inference
  36. or correlation or whatever you guys want to call it. But that's when Theo then ended up going
  37. to the Cubs and there was kind of a reshuffling over there. Okay. But I was just a lowly intern.
  38. It was fine. You were the only- Laying the seeds for the World Series two years later.
  39. Yes, I take full credit for that. There you go. Now we've found the proper causality.
  40. We're having, Greg's not a typical guest for us. He's not working for a sports team even
  41. though he has that experience. But he is writing interesting papers. So he shared with us a few months
  42. ago, some work that he's done on trying to identify skill versus chance in sports and
  43. across sports. And he explicitly builds on our friend, Michael Mobison's work. Michael wrote the book, The Success Equation some
  44. time ago now, 25 years ago, 20 years ago. And Greg's going to build on that. Michael's been on the show.
  45. He's been a while, but Michael's also in the financial services space. He's also kind of moonlighting as a sports
  46. analyst. Here's Greg writing academic quality papers, multiple. We thought he was going to send us
  47. one paper, sends us two papers, while holding down this ridiculous hedge fund manager job.
  48. I don't know how you do it. No, it's great. And Michael Mobison's work there, I'm trying to
  49. emulate that. I think, again, going back to tying in financial services and invest in sports concepts and
  50. going back and reading that book, that was sort of, everybody had to read that book
  51. that was in investing because it's such a core. And you talk about it here on the
  52. floor. Even people that are much younger than myself will have heard and read about the book.
  53. So really influential. Part of it was just going through that book and looking at it and then getting
  54. re-excited about the space. There was a paper that came out by Gerdie and Newman at the end of 2024.
  55. And they had done some Bayesian stuff around the Bradley Terry model. Paper was kind of cool because it had
  56. sports, the depth of competition in sports on one axis and then like animals and like
  57. dogs and sort of, there's major depth in the natural markets, if you will, but then
  58. sports was always very constrained. So that got me excited. And then the nice thing is they had
  59. put a nice Python package together. So I was able to get that going and I needed our CPUs here at man
  60. group to run some of those Bayesian. I think I oversampled. I'm sure if we talk about some of
  61. the details, but I cranked up 40,000, four samples at 10,000 each. So I had kind of these four climbers
  62. going. So that was part of it. And then the other thing that allowed me to do it over COVID, as my COVID
  63. break, I just decided to collect as many game-by-game results for as many sports
  64. as possible, going back as far as I could. And part of it is just my philosophy.
  65. I thought there hadn't been a lot of work about cross-sports analysis, I guess. People had gone very deep into baseball or
  66. NBA. And I thought there might be something interesting if I just took some basic techniques like
  67. Bill James' Bad Green Theorem, whatever it might be, and just apply it across sports, and
  68. maybe I would learn something that way. So that started that. So the minute I said I had that
  69. data, then I read the paper, the Gertie Newman paper, I had the Python package. I had to get a little help from
  70. our tech folks to get it all installed. And I just cranked it over Christmas, over
  71. the Christmas holidays, holidays. As one does, as one does. No one was on the server, so I
  72. felt like it wasn't too bad. It wasn't crowding out real work. And then essentially with the development of some
  73. of the large language models and Claude, they hadn't contemplated some Davidson model and sort of
  74. the ties and the home field advantage and stuff in their basic model. And since I was doing things like English
  75. Premier League, and I knew the home field advantage could be pretty profound in effect, maybe
  76. your estimates of competitive depth, sort of making mistakes, that I thought if I could just
  77. use Claude to say, hey, add the Davidson model, add these other things. And it actually built it pretty quickly for
  78. me. And then I kind of cranked that out. So then I had a full working paper just from the estimation.
  79. So looking at competitive depth in markets, in this case, sports. And then the final point, because I built
  80. a season simulator for every league, using like, let's say the 2024 schedule, I was able
  81. to have Claude come in and help me optimize that and make that a little quicker.
  82. So I could take depth of competition, sort of model the league, and then simulate the
  83. league over the actual season. So I could translate from kind of single game action into what it meant for end
  84. of season, sort of relative rankings. And so the way I was gonna originally set this up for our investor base is,
  85. hey, a lot of our job is to find portfolio managers with skill, right? And one of the ways you do that
  86. is, yeah, you wanna look at their track record, but are there other things that you
  87. can look at, entry exit skill, other things that aren't just the pure sort of realized
  88. track, in this case, realized wins. And so just thinking about that as a framework, and then people get the sports and
  89. the other nice thing in our space, there's probably, in our global reach at the company,
  90. of the 16 sports I looked at, there's probably at least one fan in the audience for each one of those, right?
  91. So it's sort of as a no brainer. You can't just do one league, right? Because people get very frustrated.
  92. So you have to all, the breadth also helped from a marketing perspective. You got pretty broad, like super rugby or
  93. something down in the South Pacific. I was like, that's getting out there. But you looked, I mean, you went, you
  94. know, not only all the North American professional, but also down into collegiate basketball, men's and
  95. women's. You got 14 European soccer leagues. So you did a nice canvassing. Let us start at a high level and
  96. check my understanding. My impression is, again, so kind of riffing off Mobison, I'm jumping over, I'm sure I'm
  97. jumping over lots of academic references since Mobison, but let's just start with Mobison.
  98. The way you write it, I believe, is Mobison fundamentally wanted to say which sports are
  99. more skill-based versus chance-based as a single characteristic. And you come in and say, well, technically
  100. there's two main things going on here. One is the fundamental chance nature of the
  101. sport, the competition, but also how compressed are the teams and their abilities.
  102. And the more compressed, then it's gonna look more like chance, even if it may be
  103. a very skill-based exercise. And then the more spread out they are, the more depth, the less depth there is,
  104. the more diagnostic these games will be. And so you're saying we really ought to
  105. technically separate these things. But then you find that they basically don't separate.
  106. This is the frustration, I suppose. So you do some real nice modeling, but then in the end you're like, well, you
  107. can't actually separate the chance nature of the sport from the heterogeneity of the quality of
  108. the team. So there's a metric that he uses called contribution of luck, which is sort of the
  109. variance of a coin flip. If your league was completely random, what would you expect to see in the cross-sectional
  110. variance of team winning percentages? And so on that metric, NBA looks really, really good.
  111. There's not a lot of luck involved and then you kind of go to the other extreme.
  112. And if you actually run that, and I did this looking for that cause I had this long history of data, it's actually time
  113. varying. He talks a lot about it being sport specific, but if you go back to the
  114. equivalent of the English Premier League in the early 1900s, it's like a random number generator
  115. back then. You think baseball's bad, English Premier League first division pre-World War I is really random.
  116. So some of it felt not so intrinsic to the sport, but maybe about the competitive
  117. environment. And I thought that was a bit missing. And if you actually do some work, and
  118. there's a couple of papers that came after the Mobison book, that you're not quite completely
  119. controlling for sample size and the dispersion effect. Yeah, okay. And so the reason NBA looks really good
  120. in that particular model and what ultimately you see is that it's got good depth, kind
  121. of in the middle of the pack depth, but they also play 82 games, right? And NFL's very wide, but they don't play
  122. a lot of games and Major League Baseball, it's very narrow, but they do play a lot of games.
  123. So you can- Real quickly, Greg, you said middle depth, wide depth, narrow depth.
  124. Say, elaborate what you mean by that. Cause you just said- So just a measure of how dispersed the talent is in
  125. a given market, right? Or a given league. So you can, some of the basic models assume a normal distribution of talent, let's say,
  126. and then what's the standard deviation of that talent would be one way to think about
  127. that. And just to re-summarize what you said, you said baseball's, on that dimension, baseball's kind
  128. of, no, no, you said basketball's kind of in the middle, football's, professional football has greater
  129. spread, and college even greater again. College is greater. And then baseball is the most compressed, just
  130. in terms of distribution of talent within the - Correct, and over time, it's gotten narrower
  131. and narrower. Okay, so we have the basics on the table. Okay. Let me get just one more basic on
  132. the table, cause Greg goes through this exercise to say, basically a win in baseball reveals
  133. the percentage signal in a single win in Major League Baseball is like 1.7%. It's
  134. 98.3% noise and 1.7% signal. And then he does the same thing for all the other sports, but then you stack
  135. it up by, baseball plays 162 games. So by the time you get that sample size, you hit more signal, but you can
  136. do things like, okay, at what point in the season do the team's records have more
  137. signal than noise? Very helpful way to compare these different leagues. And you do that for all four major
  138. North American sports. And I got that idea from Tom Tango, actually, in some of his original way back.
  139. And that was, and I think Michael used a little bit of that as well when building his metrics.
  140. So yes, it was nice to have that come out to be consistent. Well, let's just put that observation on the
  141. table and then we'll open up to questions. Because I want our listeners to understand that
  142. that's in some ways, the most practical implication of all of this is, okay, how do
  143. the leagues compare on this very important thing? And one thing you need to do when
  144. you do that is to consider how long the seasons are. So you reach this 50% point at
  145. different places in the different sports, but then you have to translate that down into percentage
  146. of the season. So MLB, NHL, NBA, and NFL, you need 69 games in baseball to get to the
  147. 50% signal. So we're saying the record in baseball, major league baseball, after 69 games is half signal,
  148. half noise. And that's 43% of their regular season schedule. NHL, remarkably to me, this is remarkable, 36
  149. games on NHL, which is also 44%. So at the same point in the calendar, the NHL, their diagnostics of a game and
  150. their sample side, their schedule match up with baseball in that way. But contrast that with the two others.
  151. So for NBA, 14 games into the season, 14 games into the season, you've got as much signal as noise in the record.
  152. That's only 17% of the season. So this is everyone's sense that basketball season's
  153. too long. And then for the NFL, it's 12 games, which isn't that much, but it's a big
  154. fraction of their season. 71% of the major American sports is the deepest into the season that the record
  155. actually captures 50% of the signal. Okay, with those basic facts, Greg, do I have that roughly?
  156. Yeah, there's a basic test theory about that and how you take a correlation, sort of
  157. the correlation between outcomes and talent and then what that would look like over the course
  158. of a season, Spearman-Brown correlation. There's some math behind it, but yeah, basically
  159. that's the idea that you've got this, ultimately you care about the final season correlation between
  160. the talent of your team and what actually is the record. And that can be manipulated either by really
  161. high accurate sort of game outcomes, or you play with a lot of noise in the games, but you play a lot of games
  162. basically. And then the other factor is how many teams you have. So you find that the chance of the
  163. best perform, the best actual underlying team reaching the number one spot is lowest in collegiate
  164. basketball because there's like 230 teams or whatever. So all these things can combine to, you
  165. could, if you wanted to, construct like optimal tournament, speaking more broadly about regular season playoffs
  166. or whatever, if your goal was in fact to crown the best team. Anyway, okay, we've got some questions from Shane
  167. and I think Shane was first, Shane. Well, I guess just on a methodological issue,
  168. you have this depth of competition parameter that's pretty important obviously to this entire modeling enterprise.
  169. But with all these kind of models, as you sort of said, you're assuming kind of symmetric talent distributions throughout.
  170. And I guess my question is, if you thought about it, if you instead actually modeled
  171. like non-symmetric talent distributions, and I think we probably all think talent is probably not
  172. necessarily symmetric. Would you, would that kind of, is that depth of competition parameter kind of capturing the
  173. extent to which these talent distributions are skewed and there's only like a certain percentage of
  174. top teams that have all the talents? Like, you know, I kind of feel like you're maybe getting that depth of competition parameter
  175. is essentially capturing what is actually inherent skewness that's not being captured by your model in
  176. the underlying talent distributions. So I got concerned about that. There's a couple of other papers where you
  177. can assume equal distribution of talent and you can kind of derive a lot of interesting
  178. closed form stuff. But rather than kind of go down that route, I went, I gathered additional data, which
  179. is not in the paper, which I think I'm eventually put in, where I went to almost all of the leagues and I got
  180. money line odds from the betting markets to use kind of a model free, sort of what is the, how does the market price
  181. dispersion? And from that, you can sort of do it in log odds space and they can talk about all the kind of technologies around
  182. it, but essentially changing, taking money line into a probability that team A beats team B.
  183. And then you can look at the dispersion of that. And I calculated that and it lines up
  184. almost perfectly with the bottom-up analysis that I did through the assumption of the, kind
  185. of the Bradley Terry assumption of normal distribution. So I would say it's not terrible, it's
  186. not perfect, but I think for the data that I had, none of the conclusions change.
  187. And in baseball in particular, I think a given game you'd expect it to be very tight, depends on who's pitching, those kinds of
  188. things. But that data set was a really interesting data set that took a while to download,
  189. but like, and I can only do 2021 to 2025, but every single game across all those leagues and probably monitoring 40 to 50
  190. bookmakers, you can kind of aggregate it up. So it's pretty powerful, cool data.
  191. There's a company that does this and I was able to download all that. Greg, we skipped over the fact that you
  192. said bottom-up, model-based evaluation, because you're, in the course of this analysis, estimating strength
  193. parameters for every team in your data set across all of these leagues, a single one
  194. for each season, but that's what you're talking about. And now you're saying, I can do it
  195. that way, or I can go out to the market and get information. I can infer those underlying strength parameters from
  196. market information top-down. And you're saying they coincided very well. They're pretty close.
  197. Well, I guess I have too many, so I'm gonna have to start with, I guess I'll start at the bottom.
  198. I mean, I actually wrote a paper on measuring luck and skill sort of in general.
  199. And the key, basically two components has to do, essentially, how much learning does it take
  200. to go from the bottom to the top? And you essentially divide up that path in levels.
  201. So from the bottom to the top, there are levels. And a level is where the team, at
  202. one level higher than the one below it, consistently beats them. And so that's a very loosey-goosey idea
  203. because it's very hard to make perfect sense of that stuff. So that leads to the question, and this
  204. is how I always like to think about skill versus our compression of talent is really
  205. what we're talking about here, because it's all skill-based. There's very, very little actual dice-oriented luck
  206. in this kind of sports. Although there is, how we describe that luck is complicated, right?
  207. I mean, if you think about it, what is luck, right? We just saw a penalty shootout that determined
  208. the Champions League. Is that luck? Is that skill? I don't wanna get into that. That's sort of a sideshow.
  209. But the questions that I always ask about sports is, which seems to me to summarize
  210. it, is if you look at the best team versus the worst team and they played each other, what's the probability of the best
  211. team winning? And then look at the best team versus, say, the average. Let's say, look at the best team versus
  212. the first quartile. And sort of look at the dispersion and the probabilities. And we all know that for baseball, for
  213. example, that's much tighter. You don't see even the very best team playing the very worst team still at most
  214. might have an 80% chance of probably even less of victory. But we know in basketball, best against worst,
  215. it's probably nearly 99% chance of victory. And then in football, it's more compressed slightly.
  216. And that's kind of a way to understand how much skill and luck there is in the game because of that.
  217. And that directly translates out into questions like what's the probability that the best team wins
  218. the title at the end? But that, of course, then sneaks back to Shane's question, which is normality is not gonna
  219. really work. If you have an outlier, they're gonna run up the score. And then I guess those are all statements.
  220. I'll just sort of finish with a question, which is at any point in your analysis,
  221. do you look at the actual outcome of the game other than who won and who lost? And doesn't that reveal information?
  222. So if you, for example, if I destroy my opponent, doesn't that give me inside information?
  223. Oh, absolutely. I did not do it. This was just the pure outcome. Clearly, you are throwing out tremendous amount of
  224. information. It's funny, in our high school sports leagues in Massachusetts, they don't even look at wins
  225. and losses to calculate playoff ranking. It's all about goal differential or point differential
  226. and strength of schedule, right? So clearly in those very unbalanced situations, that's
  227. a much more accurate predictor of who's gonna do well in the playoffs. So this was clearly, and you could definitely
  228. augment the data. My simple augmentation was just really the home away thing. That was not in there.
  229. That was a meaningful variable that was missing. But yes, you could probably move things into
  230. goal differential or point differential, which would be much more powerful. Your scaling question's very interesting.
  231. There is embedded in the paper, borrowing from the Judy Newman paper about what the variance
  232. is of two teams playing each other. It's about one. You just assume it's one. So that coefficient of depth has a scale
  233. assumption into it, which then allows you to kind of think about, so it's sort of
  234. relative ranking, but not in some sort of weird absolute sense. Though it does seem that teams typically or
  235. so a lot of leagues have a competition level of about one as it just turns out. But I think that was more of an
  236. outcome that was a bit random. But there's a scaling assumption for sure. And again, I will show you the normality
  237. assumption I think is not terrible. I mean, if you look at it, and there are certain situations and the Dodgers, et
  238. cetera, maybe in baseball, but it's a good place to start. I don't think it would materially change the
  239. outcome if you did some things in the SKU, but that's definitely something you could test.
  240. You could definitely assume, instead of assuming a prior of normal distribution and talent, you could
  241. assume SKU normal, whatever kind of distribution you wanna look at. Guys, let's talk a little bit more about
  242. one higher level methodological question, which is this inability to distinguish the chance nature of the
  243. sport from the depth of the competitors. This is vexing in a way. You'd like to be able to do that.
  244. And I came away wondering if you needed a little bit more of an economist approach
  245. to can you take advantage of some exogenous variation in the world to get some insight
  246. into that. So for example, you've already mentioned the fact that Premier League changed over the last 100
  247. years. And you have this fascinating graph in there where Major League Baseball and Premier League looked
  248. about the same in the 30s or 40s, I forget exactly when. And then they depart radically with the Premier,
  249. not the, I'm saying Premier League. It was first division up until 1992. People get, particularly here in London want me
  250. to say that. Particularly, yeah, exactly. So that was actually a little bump. When they went to Premier League, it accelerates.
  251. This shoots up where you're seeing more diagnosticity in the outcome. And it seems that pretty obvious it has
  252. to be that that's because the team quality is more differentiated than it used to be
  253. because the sport isn't changing as fast. This can't be as changing as fast, it seems to me.
  254. So can you use that in some way? Similarly, and I think less problematically, you have
  255. four or five European soccer leagues and they show up similarly in your data, but not
  256. identically in your data. And they are playing the same sport. There's this, you can't believe that there's differences
  257. in the chance nature of the Bundesliga versus La Liga. I mean, maybe there is, but it seems
  258. like most of that variation that exists has to be coming from differences in heterogeneity of
  259. the quality of the teams. Correct, and that was the basis why I wanted to do the top five European football
  260. leagues. And it was also the rationale for also going back in time. So looking back into the contribution of luck
  261. metric, if you show that it's time varying for a given sport, very time varying, that's
  262. one control, right? Given that the rules haven't changed that much. And then if you take completely contemporaneous, essentially
  263. the same rule set with fluidity across the leagues, that's another natural experiment.
  264. And I saw contribution to luck was varying there as well. And so that, and if you unpack it
  265. and run the analysis, you start to see these differences, even though they're almost identical structurally,
  266. there are these differences in the depth. And part of that is the spending, the various caps.
  267. There's a lot of league dynamics that I don't understand completely, but there's a lot of,
  268. and we're talking about salary caps in Major League Baseball. I think in Premier League, they have caps
  269. on the amount of losses you're allowed to have. Because people don't make money in Premier League
  270. on a given year. It's all made up in the capital value of the team, right, ultimately.
  271. And so a lot of interesting league dynamics come through there. And you can kind of map back to
  272. that. And there's also one of the leagues plays slightly fewer number of games as well.
  273. There's a little bit of that noise in the background. But the nice thing is there's a really
  274. nice data set. The University of Texas professor catalogs all of these games, all the way down to multiple
  275. divisions, going all the way back through time. And so it was really easy for me
  276. to download and run the analysis to see that when that kind of perfectly controlled.
  277. I thought it was a good control experiment. For sure, for sure. Well, do you come away with a better
  278. sense of the chance contribution in the sport by that? Because throughout the paper, you're basically saying, we're
  279. not able to say, we're not able to say. Well, so the one thing I, so what I want to, there's a little bit of
  280. a distinction between, and I think that, again, going back to the original inspiration, the Judy
  281. Newman paper, how do they define luck, right? So people use it kind of generically.
  282. Is it sample size? Is it this concept of competitive disburden? So all the teams are the same.
  283. So it is a coin flip, but they actually introduced another parameter in the model, which
  284. is this concept of how they define luck, which is essentially the probability of an upset
  285. win. So it's like, what's the probability that an infinitely good team loses or an infinitely bad
  286. team wins, right? And so essentially what they're doing in the model is they shrink the Bradley Terry model,
  287. which will give you your probability that the better team wins back to 50-50 probability,
  288. right? And so in the model, there's another parameter they call alpha, you know, with big letters.
  289. So you have depth of competition beta and this alpha, which is the shrinkage parameter back
  290. to pure coin flip. So if a league was purely a coin flip, then this alpha number, and it would
  291. be a number between zero and one, right? You're going to wait between the Bradley Terry
  292. outcome and the pure coin flip. So econometrically, I thought it was a pretty cool way to diagnose it.
  293. Now, the problem when you go through all of this, and that was the point of their paper too, when you're getting these very
  294. low competitive dispersion leagues, right, there's a lot of collinearity. And so I did a couple of charts
  295. where you don't really improve the model fit by adding this pure luck parameter.
  296. And so that's why, so I think why I'm leaning back to this competitive dispersion explains
  297. a heck of a lot in these sports, and it's a little bit less around what actually goes on in the sports.
  298. Now, I could kind of go a little bit on this concept of sports information ratio
  299. that I stole from the investment literature, which does tie in a little bit and is
  300. a bit more consistent with what was his point. But if you talk about the Bill James
  301. Pythagorean theorem, that parameter, it's sort of- So hold on, Greg. Yeah, I mean, I'm going all over the
  302. place, but we can- No, that's fine. Let's do that because this is- You're actually getting somewhere with that point because one
  303. of the problems that, when you've mentioned it several times, and I'm gonna make it really
  304. specific, if every team is exactly the same quality, then whatever determines the outcome of the
  305. game, it isn't skill, it's something. You can call it chance at that point, even if it's not an actual dice, we
  306. call that chance. And then as the skill gets a little bit larger, and that chance component abruptly has
  307. to stay the same, then that's the only way you can sort of measure that chance component.
  308. But now the question you might ask is, and this is where Bill James comes into, well, what is it, what is it the
  309. thing that's causing it, right? So the sport that I know best is baseball, and I can probably tell you where
  310. that is, it's when you hit the ball at a certain angle and a velocity, and sometimes it's a home run, sometimes it's not,
  311. right? And you don't know why, and it could be because it's- Or you throw the ball and sometimes it curves and sometimes it
  312. doesn't. Yeah, little things like that. I mean, it's not chance as in a dice roll, but it's something completely out of
  313. your control. And that's the thing that we can measure. And what Pythagoras is really not measuring that,
  314. it's measuring the dispersion of runs by game. And if you spread them out, you know,
  315. you don't wanna win by a lot, you only wanna win by a little. That's measuring something different, but almost every sport
  316. now has something called X something, XG, X Woba in baseball. I don't know if basketball yet has an
  317. X, which stands for expected. And the difference between those two is in some level what luck is trying to capture.
  318. And maybe we can actually measure that luck by looking at the difference between X, whatever
  319. X number and goals, X Woba and Woba. And I'll just talk about this because like Ben Rice right now in baseball, if the
  320. Yankees, I'm gonna have to bring up a Yankee, it's my job, has hit a cover off the ball.
  321. And for the last two seasons, he's had the biggest differential between his actual Woba and
  322. his X Woba. And everybody was screaming, there's no way this guy is, you know, top quartile.
  323. This is a superstar. And we're just waiting for him to just regress back down to what we would call
  324. ordinary luck, as opposed to bad luck. And that's kind of a way we can measure it.
  325. Yeah, and it goes all the way back to FIPS, fielding independent, you know, ERAs and
  326. those types of, which is a good easy example with clients. I could just, you know, it's really about
  327. strikeouts and walks and home runs. Once the ball's put into play, it's a bit of a random outcome, right?
  328. And what do people control? So completely by, I think, yeah, yeah, go ahead. I was like, the one thing, just because
  329. you brought basketball, I think the reason we don't have an X or expected thing in
  330. basketball is because the actual outcome happens often enough that you can just look at, you
  331. know, like X is important in soccer, X is important, expected goals is important in soccer
  332. and hockey, where, you know, you've got these underlying, you know, kind of like peripherals that
  333. are predictive of it, but the actual outcome doesn't realize as often as you would like
  334. to. And so you want a more kind of smooth measure of things going on in the game, basically.
  335. Whereas basketball, you don't need that as much. It's more valuable, but people still use it
  336. in basketball. So, you know, whether or not, in fact, who was the Cavs coach against the Knicks
  337. after they were down, I think they were down, they'd lost three games, so they were
  338. close to eliminated. The Cavs coach is like, we're actually winning by expected metrics.
  339. And he means like, you know, three-point shooting. Some studies have said, you know, conditional on
  340. the shot being taken and the degree of coverage is like, it's chance what happens once
  341. it leaves a guy's hand. And so they do use expected some in that way. Greg had this observation, it may not be
  342. in the paper, but he's mentioned it to us on the side, that I think you guys will find interesting.
  343. I'd love to hear him say a little bit more about it. And it's how James's Pythagoras formula, parameter shows
  344. up across sports. So before we do that, Greg, can you just remind our listeners what Bill James's Pythagorean
  345. theorem does? And so we can talk about this. It's a way to translate runs scored and
  346. runs allowed or goals scored, goals allowed for a team over the course of the season
  347. and how it explains their ultimate winning percentage. And theoretically, if there's a big deviation, maybe
  348. that's due to random noise or luck, right? Either direction. Maybe the team got lucky in one inning
  349. games or didn't get lucky in one inning games, but it's sort of back to Adi's point about, it's more of the expectation over
  350. the course of the season run differential matters. And that's again, back to my Massachusetts high
  351. school point, goal differential is how you predict future performance. And so there's a way, basically the Pythagorean
  352. theorem translates scales that runs scored, runs allowed into a number that maps to the winning
  353. percentage of a particular league. And so it does vary by league. In short, expected number of wins, given the
  354. number of runs you scored and gave up or given the number of goals you scored and gave up.
  355. It's expected based on that. And there's a parameter in there that's vital. It calls it the Pythagoreas because it's the
  356. square root of the sum of the squares of those things, right? But for baseball, it happens to come out
  357. as a two. So that's why, yeah. Okay. So the exponent in that little formula, so this is classic Bill James.
  358. This may be the quintessential Bill James, which is like not sophisticated math, but it's deeply
  359. insightful and parsimonious. And now others have used that and people talk about it.
  360. And so what do you, so the exponent in his formula for basketball happens to be two.
  361. And then you go out and find what trick. Yeah, so baseball two, basketball it's 14 and
  362. whatever the number, you can do this a bunch of different ways, but what I was feeling in this, and it kind of goes
  363. back to maybe our sample size argument about basketball, right? Or maybe baseball.
  364. Imagine baseball, they played every game was a thousand innings, right? So there's a little bit of the game
  365. unit itself has embedded, I think, different observations where the NBA has got a hundred possessions.
  366. Yeah. You're only getting four at bats per, so there's a little bit of noise that comes
  367. across in the baseball. And so what basically, there's a lot of fundamental reasons why this parameter, and you can
  368. make assumptions about is it a Weibull distribution or whatever, the run scoring mechanism.
  369. At the end of the day, think of this Pythagorean coefficient as a scaling number that
  370. translates run scored versus runs allowed to winning, right? And ultimately, if you plot it, and if
  371. you do it for all the 14 sports, you almost get a perfect way. First observed, this is completely empirical.
  372. Then I was able to show it sort of mathematically. It's perfectly linear in what I call the
  373. sports information ratio, which is the mean of goals per game divided by the standard deviation
  374. of goals per game over the course of the season. Okay, and so I call it the high,
  375. so basketball, I'm gonna quote this off the top of my head, basketball is like an
  376. eight, and major league baseball, English Premier League, is like a one or a little over
  377. one. And this relates to the information about a team's quality that you infer from their-
  378. It's basically that game has a lot more, theoretically, that game in the NBA has a
  379. lot more information. So they play 82 games, and each game has a lot of information.
  380. So it kind of builds on that. But what I haven't done yet is build a unifying theory that if you do like
  381. Bradley Terry, I'm close, but this is just, again, me being an amateur, of how the
  382. competitive depth and the sort of sample size of the game or information ratio of the
  383. game come together to describe things. But anyway, if you do the math, you can kind of, with making some very basic
  384. assumptions, there'll be some scaling coefficient, but it should be linear in this information ratio concept.
  385. Are you talking about the sharp, almost like for a sport? Well, so I stole this.
  386. We spend a lot of time in sharp ratio. Philosophically, I know I'm gonna be doing a
  387. little bit more rambling. I always felt like sports, they never spent, it's always about point estimates.
  388. There's very little, usually around uncertainty, around parameters. And I live and die by the mean
  389. and the standard deviation. So a few years ago, at one of our conferences, I came up with a thought
  390. experiment. Which team would win or lose more games? A team that gives up exactly two runs
  391. every game or a team that gives up zero 50% of the time or four 50% of the time.
  392. Which of those two teams would be better? And then you can do the same thing on the offense.
  393. You know, he scores exactly two. And it was this whole Jensen's inequality, blah,
  394. blah, blah, convex, concave. And basically you want a very unstable defense and a very consistent offense was sort of
  395. the method. But it only matters in low sharp sports or low information ratio sports like soccer and
  396. baseball. You get the NBA, it doesn't matter at all. So that was, I added another variable to
  397. the Bill James equation and it loaded, but only in ways that would be consistent with
  398. the IR. And that made a nice story for my investors that you wanna buy high sharp ratio
  399. strategies. Risk management is much more important than low sharps. So anyway, that was the, that's exactly the
  400. point. And back to the paper, you're observing that this, the exponent in James's Pythagorean is linear
  401. in the sports information quotient. There's a scalar to it, but yeah, exactly. It aligns in the data.
  402. It's kind of a cool little paper. That's neat. And that's only because you're able to look
  403. at however many different sports you looked at. That's a, that's a- I just throw
  404. out something that our listeners should sort of know. If you just look at the point differential
  405. as a, for every team, that almost is almost perfectly linear in their winning percentage.
  406. Not over the history of the game, because if you go to the periods of time where no one was scoring and you look
  407. at the massive, you need, then you need the curvature. But if you just look within a season,
  408. for example, at the, at the winning, the delta, your point differential for every sport, goal
  409. differential. Actually I use points for soccer. It's linear in your winning percentage.
  410. It's very, very linear. I mean, and you can see where this comes from. You're looking at this ratio to a power
  411. is the first Taylor series. So first term is linear. And then- Linear, exactly. And then what I could find, that beta,
  412. the slope coefficient that takes that goal differential to map it to winning percentage.
  413. That's the number I'm talking about. We'll scale with this information ratio. Completely agree with that.
  414. Okay, team. We should let Greg go. He's coming to us many hours ahead of us and we've talked with him longer than
  415. we thought we're going to talk with him. But Greg, thank you for sending us- Please add anything.
  416. Yeah, of course. Thanks. Where, if our listeners are interested in seeing your paper, where could they find it?
  417. So we've got, on this particular paper, it's not quite out yet, but the linear information
  418. ratio is out on our, if you just Google it, Pythagorean theorem man group. There's also a Python package behind it.
  419. If you download it, you can run it. Advertises a bit on our Arctic DB, which is one of our database tools for Python.
  420. Well, you know, speaking of that, I noticed in your footnotes that you grabbed your data.
  421. A big chunk of your data came from Neil Payne, friend of the show, Neil Payne, and sort of a co-founder of FiveThirtyEight.
  422. And it goes all the way, the fact that that's there, goes back to FiveThirtyEight's founding
  423. philosophy. They decided right up front, we're going to use Elo for all of our stuff because
  424. it generalizes across sports. We know it's not perfect, but it generalizes across sports.
  425. But because they had that philosophy and Neil's maintained it, you were able to grab data
  426. and run this. If you are a friend of Neil, have him update for 2025. I had to do that myself.
  427. Well, so he comes to the summer to campus to talk to my students every summer. He said that one of the tricks he
  428. talks to the students about is he's very good about keeping an ongoing log of every
  429. game that happens. So that for all the sports that he follows and he writes about, he's always updating
  430. his database just on a daily basis. So then when he's ready to run something, he just has it there.
  431. He doesn't have to do this mad dash search. If you guys want either the GM data which I sent over and that other thing,
  432. and also any of this game level data going back to the 1870s, I can cost those 16 leagues.
  433. I'm happy to share it if you need it. We would love it. And I'm going to probably have one of
  434. my staff get in touch with you to get that. Greg, we have lots of- Because it would be a fantastic tool.
  435. Yeah, we have students doing projects all the time and they would find that useful, I'm
  436. sure. And perhaps in some world, they might be useful to you. So Greg Bond, thank you for being with
  437. us. Greg, again, Chief Investment Officer at Mangroup. Mangroup is a very large alternative asset management
  438. group. He is based out of Boston. He runs Americas for Mangroup. He's also the PM, the Portfolio Manager for
  439. their flagship fund. Greg, great talking to you and good to see you. Good to meet you all.
  440. Thanks. Interesting time of year, guys. Championship time of year. I think this is as many championships as
  441. we're ever going to see in the calendar. We are starting the NHL Standing Cup Finals
  442. tonight. We've got the Carolina Hurricanes playing the Vegas Golden Knights. I'm seeing, what, 58, 60% chance on
  443. the Hurricanes. Shane, you've got something up there about their - Yeah, I think my- Yeah, I
  444. mean, the Hurricanes definitely are pretty heavily favored in this, but the Avs are really heavily
  445. favored against the Knights, too, so what does that mean? You know, they've got that- I haven't
  446. watched- For some reason, I was all about the Eastern Conference this year. I haven't seen as much from the West.
  447. When I saw the West, they- Who said this is- I'm going to show all my ignorance here, Shane.
  448. The captain of the Knights was injured for the first couple rounds of the playoffs.
  449. He came back mid-playoffs, and he's this big block of a guy. He does not look like a typical hockey
  450. player. He reminds me of, and this is another name I'm not going to have, Kentucky, years
  451. ago, had this Fred Flintstone Square-looking quarterback, playing quarterback. He was so much fun to watch because
  452. he was so differently shaped than the typical quarterback. Yeah, I think you're talking- I mean,
  453. the first one you're talking about is Mark Stone, right? Oh, Mark Stone, exactly.
  454. I think you have to like- Pretty block of stone. He's like, you know, I feel like he's
  455. like a rock out there. You, I think he's fun to watch and fun to pull for, just because he's so
  456. unusual out there. So, and who's their little guy that's so good? Oh my God, the former Leaf, who just
  457. looks- Oh, Mitch Marner. Mitch Marner. Yeah, Marner, just different than everybody else.
  458. So there are some fun things about the Knights and the way they took care of the abs.
  459. It's not obvious to me, but the Hurricanes did get that 10-day rest. They smoked the freaking Canadiens far worse than
  460. the outcomes looked. Anyway, that's our- No, I mean, Hurricanes have been kind of, you know, they're I
  461. think 12-1 now in the playoffs so far. So they're kind of, you know, I mean, if they were to, I mean, deal with
  462. the Knights in a similar fashion, this would be kind of a relatively historic run.
  463. I think the modern kind of, I put it like the Oilers in like the late 80s had one where they went 60-2,
  464. I think, in the playoffs. So, you know, if you're harping back to kind of Gretzky era Oilers playoff dominance, you're
  465. doing pretty well. Right. And they, so they could drop one and tie or they can sweep and get the
  466. modern record. This is pretty impressive. What about the NBA? The Spurs took down the Thunder in seven.
  467. That was a great series. Back and forth, back and forth. Beginning of the season, we wouldn't have expected
  468. the Spurs and now the Spurs are pretty big favorites against the Knicks. Knicks have looked great, but they haven't played
  469. anybody. You know, these two teams, this was something like only the eighth or ninth team time
  470. that the two best regular season teams played a game seven in the NBA, something like
  471. that. So that was, these are head and shoulders the best records coming into the season.
  472. It'll be interesting to see whether the Knicks. It reminds me a little bit about the,
  473. you know, back in the day when like the Lakers and Kings were battling back and forth and the rest and the West and
  474. like, you know, whoever was in the, sitting there in the East was a little bit of an afterthought because did it, you know,
  475. did it really matter? I mean, I'm not. Am I right? The Knicks have a much worse winning record
  476. during the regular season. Regular season, yes. But they've had a very impressive playoffs.
  477. Yeah, extraordinary playoff. Yes. The question's whether that's just a weak field
  478. or they've actually caught some fire. And of course the big question's what are they gonna do with Wendy?
  479. Yeah. Which would be. Yeah, yeah, I mean, to a certain extent, them getting stomped by the Spurs wouldn't necessarily
  480. be evidence that they're not actually a very strong team. It seems the Spurs seem.
  481. That's right. Relatively palpable right now. Well, I mean, people, I don't know whether
  482. there's any value to this, but the Knicks did beat the Spurs two out of three, I believe, during the regular season.
  483. Does that mean anything? I don't know. Do we have a regular season for any purpose?
  484. Right, it's so sketchy. It's really, I wanna go back and look at the minutes that Wendy played, that kind
  485. of thing. The Spurs are bigger favorites in that championship series than the Hurricanes are.
  486. I think the Spurs are like 64%. Yeah. Fun fact. The Odds are about 200. They're about minus 200.
  487. So, about that. Guys, one of our listeners sometimes sends information about the ticket prices, and apparently you can
  488. get this off at ESPN. I haven't looked. I think this is like your typical get -in-the-door ticket prices.
  489. So, San Antonio is like the 19th or 20th biggest market, and the NBA, New York's
  490. obviously the biggest market. What do you think the difference is in the price? I don't know whether it's average or marginal,
  491. but ESPN will give you a ticket price to get in the door. What do you think the difference is?
  492. Like, what do you think the ratio is? But for the next Spurs game, New York versus, net games, New York versus San Antonio,
  493. ratio. Are we all supposed to decide now and then hold our numbers so we don't cheat?
  494. I'm sorry, I'm gonna just up and say five. Five to one. I'm gonna just say what I think the
  495. numbers are. I would guess you can get into the, in San Antonio for around $400, and I
  496. don't think you can get into the Garden for under $2,000. So, that's five. You're a little low on both.
  497. Your ratios are very close. The last numbers I saw were about $4 ,000 for the Garden, and like $900, $890,
  498. $900 for San Antonio. So, not quite five, but in that direction. So, y'all did good.
  499. Odd, your numbers were a little low, but y'all did good. Okay, that kicks off. Yeah, I'm surprised you can't, I mean, is
  500. that, the game hasn't started yet, so there might be an aftermarket reduction. This is one observation from a couple days
  501. ago, but there's gonna, you know, it's the same, this is the thing, it's the same
  502. matchup, just very different cost of taking it in in person. All right, I wanna do a little bit
  503. of college, because this is an extraordinary week in college sports with a handful of things
  504. going on. So, let me give you some rundowns. I'll try to find something interesting to say
  505. in a couple of these. I've got one especially interesting thing for you guys, you baseball people.
  506. But I wanna start with the men's golf tournament, because I talk about this tournament every
  507. year, because it's got such a cool format. They take 30 teams, they play regional qualifiers,
  508. I think five of them, they send six teams out of each to the nationals. That means 30 teams and some individual medalists
  509. arrive. They play four rounds of stroke play to cut it down to 18 teams, eight teams,
  510. eight teams, and then they go to match. Of course, match is where the drama is, match is where the fun is.
  511. So, it's like Ryder Cup style team match play, five guys a side, and that's how they're gonna decide it in the end, and
  512. that's good drama. But to make sure they get the best teams possible there, they play four freaking rounds
  513. of stroke to get there. And so, they ended that yesterday, they started match play today.
  514. How did they do to get the top eight? One of the fun wrinkles here, guys, is Mark Brody, who we had on like two
  515. weeks ago, three weeks ago, Brody has been involved with the ranking system. So, you can look at what the seeds
  516. are of these teams, but you can also look at what their rankings are. That rankings, according to Brody's of the final
  517. eight, one, two, three, five. So, four of the top five, and then 10, 11, 13, and 30.
  518. UCLA, the supposed 30th best, I have a lot of confidence in Mark's ratings, 30th best
  519. team in the country comes in, they land a spot in the top eight. So, here's what they do.
  520. Today, they play the quarters and the semis. So, four match play team matches this morning,
  521. and then two more matches this afternoon to decide the semifinalists, and then the finalists go
  522. in and play tomorrow. So, my Longhorns, number two team in the country, get a good draw.
  523. They got the 30th team, they got UCLA. They should whip those boys, and they got
  524. smoked. They got taken out, first team out this morning. Auburn's number one player, this Jason Coyven, Coyven?
  525. I'm not sure I say his name right. He's like, number one amateur, has turned down
  526. going pro. He lost today. They're playing the semis this afternoon. It's just good fun.
  527. Match play, team match play is great fun. But when I was poking around on this, I found some interesting records.
  528. Let me give you a couple of trivia questions. Who do you think has the most men's
  529. golf titles in NCAA history? They started playing them in 1897, apparently. Can you give us, is it an older,
  530. or are we looking at the first half of golf's history, or the second half? No, no hints, no hints.
  531. You have to guess it out yourself. All right. Adi, predict the home team. Yale dominated for like the first couple of
  532. decades. Even Harvard was smoked by Yale for a long time. Yale is, if you go back to the
  533. 18 whatevers, you're gonna hear a lot of Yale. Yeah, a lot of football, a lot of
  534. baseball even. You know, the Yankees used to play in the Yale baseball team. They haven't won a golf title since like
  535. 1943, but they still have more than anybody else. Princeton is like tied for third, I think.
  536. But here's a couple of runs that were just amazing I came across. Houston, University of Houston, won 16 in a
  537. 30 year stretch. They had this coach, must've been a legendary coach, like from the 50s into the 80s,
  538. 16 out of 30 national titles in golf, and they were runner up four other times. So 20 out of 30, they were one
  539. of the top two. That's the- That's amazing, because it's like - It's ridiculous.
  540. You wouldn't necessarily think again about a coach having, I mean- Recruiting maybe?
  541. Must've been a built-in recruit? Right, I mean- You know, that's a great question.
  542. A pipeline with that kind of success over three decades is pretty, given all- Unbelievable,
  543. unbelievable. I wanna raise this point here, because we're talking about a long time ago.
  544. And you immediately went to the pipeline recruiting. I would guess that a sport at that
  545. period in its history probably is, a coach is really gonna take a mediocre player and
  546. can potentially make them quite good. Golf? Yeah, I mean, if you think about it,
  547. I mean, how much, I mean, if you think about it, the game could not have been as advanced in the 50s, 40s, 50s,
  548. as it is even in the 90s. No, that's true, but we are talking 50 through 80s. 50s through 80s.
  549. Yeah, 50s through 80s is- By the time you get to the 80s. Peak Palmer Nicholas. So it's not an advance.
  550. But I lived in Texas through some of that era, and I knew a couple of U of H golfers, but I had no
  551. idea they had that kind of record. Anyway, the other one that jumps out, Oklahoma
  552. State, get this. Oklahoma State is, I think they're right up with Princeton. They're like 12 or so champions.
  553. They have won a men's title in golf every decade for the last seven decades. That's a different kind of streak, right?
  554. So now we've made three very different patterns, but they have been good and competitive and
  555. at the top of the game for a long time, but kind of regularly. And they're in the final four.
  556. They won their matchup this morning. I believe they won their matchup this, no, maybe they were going into, maybe they had
  557. them late. They are- They are winning a Superbowl in every decade or whatever, you know, what
  558. they've been. Who is that? Are you gonna tell me they're doing one game of Superbowl in almost every, like 80s,
  559. 90s, 2000s, 2010s? I'm saying, I missed who you're saying. Who's, I should know who that is, who
  560. you're saying. The New York Giants. The Giants? Oh God, I wouldn't have come up with
  561. that. Oh yeah. But then they're like, they're terrible. The other nine years- Right, yeah, no,
  562. again, who knows what the, I don't know what Oklahoma's doing necessarily in between their like
  563. championships. They could be tanking and then coming back. I don't even know. That's, you know, over like one per decade
  564. is kind of, or at least one per decade. Well, they had a couple of twos. They won three.
  565. I think in the 80s, they won three national titles. Okay, let me give you two other sports.
  566. So baseball, y'all's love. They just finished regionals and the Supers are seeded.
  567. The Supers are sweet 16. Eight teams are hosting super regionals. The cool thing about the baseball, both softball
  568. and baseball play the same structure. 64 team tournament with 16 four team regionals,
  569. but they go through this weird cadence. I've told y'all this before. They do double elimination and then they go
  570. to, at the Supers, they play two out of three. And then in the College World Series, they
  571. first play double elimination and then they play two out of three championships.
  572. So it's this weird iteration between double elimination, two, three, double elimination, two out of three.
  573. And when they do the two out of three, like I was watching the, you know, the college softball or whatever yesterday, and they
  574. do the latter two of the three as a double header. Is that correct? So is that- No, those were elimination
  575. games. So we'll talk about that momentarily, but when they get to the final, that was just
  576. to make the final. So that was the last of the double elimination semis. And then the finals they'll play, I think
  577. for the last 20 years, they've been playing two out of three. Men's baseball, number one and number two got
  578. knocked out in regionals. So UCLA, Georgia Tech got knocked out in regionals. Okay, here's what I want to get to,
  579. softball. So not because Texas is in the final, but we have been paying more attention because
  580. Texas won last year, they beat Texas Tech. They're again playing Texas Tech. And so we've been watching this thing since
  581. regionals and it's just amazing to see a tournament get whittled down and whittled down.
  582. And you'd never think your team's gonna end up in the finals when there's that many
  583. teams, good teams. And the horns are back. They came out of the loser's bracket, won
  584. two yesterday against Tennessee to get into the finals. And then Texas Tech did the same thing,
  585. came out of the loser's bracket, won two out of three against Alabama. Alabama, number one seed, SEC regular season champ.
  586. Tech is like sitting there at number 11 seed. They went two in a row. Here's the thing though, guys, this is gonna
  587. get a lot of attention. So Tech could easily beat Texas. And I don't know what the odds are,
  588. probably close to even. But they've got, remember Najeri Kennedy, this pitcher that came from Stanford, Tech got her, they
  589. paid a million dollars. She was this first big transfer in softball. This was two years ago.
  590. They got her and they, as the first season she plays for Tech, they go to the World Series for the first time ever
  591. in Tech's history. And they go to the title game first time. And to remind you what happens in softball,
  592. a good pitcher can pitch every game. Like pitchers sometimes pitch every inning of the
  593. post-season. It's not completely unheard of that that to happen. And pitchers can be dominant in softball.
  594. And so it's a huge advantage. Okay, but here's the thing that's going on right now.
  595. This is completely crazy. And it's gonna be a thing. It's gonna be really interesting.
  596. There are rules. Have you watched softball before and noticed that they can pull a pitcher and they put
  597. the pitcher back in? This is, sometimes it's this jarring thing that happens. A fielder was injured for like an inning
  598. and then they came back in the game. So there's this designated player rule where you
  599. can have a 10th person on the roster that's gonna bat for somebody in the field.
  600. And it just gives them a little latitude in the way they move people around. And then around pitchers, they can pull them
  601. and then put them back in. And so it's not crazy. It's not unheard of that a pitcher might
  602. be pulled and then come back in later in a tight situation. They might start their best pitcher.
  603. They wanna give her a break or whatever. They might bring her back in if the other person gets in trouble.
  604. But it's not super common. It happens, but it's not super common. Okay, the Tex coach, Glasgow, the Tex coach.
  605. So last year he had Kennedy, best pitcher in the game. They got a ton of NIL people and
  606. transfer people and like eight of the nine starters are transfers in a Tech. And they're the best offensive team in like
  607. years. And they've got Kennedy and they've got another great pitcher. Caitlin Terry, I think is her name.
  608. So, and Kennedy's lefty, Terry's righty. Terry's also a good fielder. She can play left field.
  609. Kennedy can play first base or they can sit him on the bench as a special designated player.
  610. And so what Glasgow started doing and he talked about it with the players all season
  611. long and didn't do it. And then in the regionals, no, the super regionals, Tech is 11 seed, which means they're
  612. traveling for the super regional. They go to Florida, they go to game three and all of a sudden Glasgow busts
  613. out this strategy of, I'm gonna start moving, I'm gonna start shifting. I'm gonna platoon my pitchers against your lineup.
  614. Or if a person gets in a little bit of a trouble, I'll pull her and put the other one in, let her regroup
  615. and then switch them back. And so he starts switching multiple times within a game.
  616. No one's ever seen it before. And he's got these two all American pitchers, one lefty, one righty.
  617. And he wins that third game against Florida. They get to the World Series. The other night against UCLA in an elimination
  618. game, he had something like four pitching changes and three innings or three, five and four,
  619. five and three. It's like he was back and forth. He was literally- But it's not like
  620. he was going batter to batter and every time there was a lefty, he'd bring in like- Not quite that much, but almost
  621. like that. Like if the lineup had a bunch of lefties at the end, he's gonna make sure
  622. his righty's in there against them. He even had one that I was watching where there was the big hitter for UCLA
  623. was gonna come up like second in the order of that inning. And she had already homered off a Kennedy
  624. twice. And he's like, by God, I'm not gonna pitch Kennedy against her now. Let's get Terry in.
  625. He pitched Terry against them. She made it through a few of them. And then he put Kennedy back in.
  626. And it's just, I have to admire the innovation. And I mean, it's not quite rules manipulation,
  627. but it's completely- Do we know how big the handedness advantage is in softball?
  628. I knew for a fact that when I came in here and told you all these stories, you had asked me for some empirical
  629. observations that I don't have. And I do not have that one. It is the relevant question.
  630. It is the relevant question. And I actually asked one about the kind of the pitchers as well, because I was
  631. watching it too. Like they kept talking about like a rising versus sinking. It actually is rising in this case, right?
  632. It's not like, you know, rising fastball in baseball is a misnomer. A rising fastball just is sinking less quickly.
  633. But in softball it actually, the distances and angles are such that it actually is rising,
  634. correct? I don't know for a fact that's true. I don't know for the, people think it
  635. is. It looked like it. It looked like it on TV, but so does a rising fastball in baseball.
  636. Yeah, right. I'm not gonna say for sure. I know for rising fastballs don't rise.
  637. I mean, I played enough softball. I mean, they can release it at a very low point in their body, right?
  638. And it can end up higher, but I guess it can be arcing to get there. Yeah, I'm just saying based on the distance
  639. and speed, like, is it like an upward trajectory when it hits the plate or not? It looks like it, certainly.
  640. But of course in baseball, we all know that that actually isn't the case, you know.
  641. And softballs like baseball in that they want the release to be the same no matter
  642. the pitch. And so they're not gonna deliver a rise ball at least they're gonna try not to
  643. deliver it differently than the others. But I thought y'all were gonna kick out this pitcher strategy.
  644. I totally admire the innovation, even though it might be the downfall of Texas in the
  645. finals for all I know. It may not be. I mean, if it makes it like, in terms of pace of play, it makes me
  646. glad baseball does not do that to a certain extent. Oh my God. Like, I mean, it's crazy enough in baseball
  647. with the more restrictive kind of substitution rules. Right. It's annoying as hell the way it is
  648. right now. I can't imagine it getting worse. Well, I have been saying this for the
  649. last couple of years, but I really enjoy, and I'm sure I'm biased because Texas has
  650. had a good run in softball, but softball's fun. It's faster paced. Anybody, you know, somebody, some, you know, they
  651. have some big sluggers, but then anybody can hit a dinger like this little second baseman
  652. hit a game winning walk off the other day. It's just, that doesn't happen a lot, but
  653. the risk of it is there at a steady pace. Well, I mean, here's a question for Alan
  654. Nathan, our physicist expert. In softball, it seems, and I've watched some of the home runs, they often just stick
  655. their bat out and it goes over the fence in complete swings. Oh yeah. I mean, and listen, they're using an aluminum
  656. bat, which is with incredible springiness and their pitchers are pitching quite fast and the ball
  657. is, I mean, I think it's an interesting question. I mean, how much, let me make it
  658. fundamental. How much of the power in softball is generated from the bat, the batter, or the
  659. pitcher's speed? I'd love to hear that sort of broken down. I can tell you that in baseball-
  660. I can tell you the speeds. The speeds, I don't know the batter part of that, but they report exit velocities during
  661. the games and they're in the high 70s for these home runs. And the pitchers are pitching, their change-ups
  662. might be in the 50s and their top pitches are 60s and the top pitches are typically in the 70s.
  663. I think the Tennessee pitcher famously touched high 70s last year. But so it's 70 in, 70 out, you
  664. know, roughly. 73 in, 73 out. At 40 feet, oh my God. 60. Is it, it's 60 feet, is it not?
  665. No, that's baseball. No, no, it's 42, I think. And with their stride and at that distance,
  666. it is a softball, so it slows down probably substantially from hand to plate because it's
  667. such a big area, surface area. These are lots of questions, but I can tell you that in baseball, the velocity does
  668. matter, but it's about one-tenth relative, matters about one-tenth as much as the velocity
  669. of the bat. Yeah, right, right, right, right. These are aluminum bats, and so therefore a
  670. lot. That's right, that's right. Much less. I mean, if you watch college, you know,
  671. college, they use a different ball in college baseball than they do in major leagues.
  672. It has less bounciness. Oh, exactly. Well, that's because the aluminum bats, I mean,
  673. they come off the bat pretty damn fast. It kills some people. Okay, why don't we leave it there?
  674. Appreciate y'all being up for a little college sports, minor college sports at that, but
  675. it's a fun time of year, and they have some creative tournament designs, which I think
  676. adds a lot of flavor. Why don't we leave it there? We missed Eric Bradlow this week.
  677. Enjoyed the conversation with Greg Bond, Shane Jensen, Adi Wanner. On behalf of the whole team here, many
  678. thanks for Deep Patel and Jacob Grodnick doing things behind the scenes and Marissa Reyna making
  679. the bookings happen. Thanks to you guys for listening. Come back next week. Come back and join us next time.
  680. Between now and then, enjoy your sports.

Brief Summary

Greg Bond discusses sports analytics, skill versus chance, and investment strategies in sports with Wharton Money Ball hosts.

Key points

  • Championship Time of Year. It's a thrilling moment in sports, rich with championship opportunities.
  • Greg Bond Joins the Show. Greg Bond, CIO at Mangroup, shares insights on sports analytics and investing.
  • Skill vs. Chance in Sports. Greg discusses his research on identifying skill versus chance across various sports.
  • Understanding Skill vs. Luck. Exploring the balance of skill and luck in sports outcomes.
  • Analyzing Game Outcomes. The importance of looking beyond just wins and losses in sports analysis.
  • Pythagorean Theorem in Sports. How Bill James's theorem translates runs scored into winning percentages.

Tension

45/ 100Some tension

Related episodes