How AI and Analytics Are Changing Quarterback Evaluation and NFL Outcomes

Knowledge at Wharton

Ask about Knowledge at Wharton

Episode Description

ABOUT THE EPISODE In this episode recorded on 12/31/2025, Eric Bradlow, Shane Jensen, and Adi Wyner examine how AI-driven metrics, model calibration, and Bayesian approaches inform quarterback evaluation, team upsets, and the evolving limits of sports analytics. Eric Bradlow talks about how modern analytics reshape the way quarterbacks are evaluated beyond traditional stats. Shane Jensen discusses the role of coaching, team context, and patience in quarterback development. Adi Wyner explains how Bayesian thinking and model uncertainty change how analysts interpret surprising results.

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 #FootballAnalytics #AIInSports #NFLAnalytics ----- 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 Moneyball here on the Wharton podcast network. This is Eric Bradlo, professor of
  2. marketing, statistics, and data science here at the Wharton School. I'm here today with my 11 and a half year
  3. collaborators on Wharton Moneyball, but of course a lot longer than that as friends and colleagues. I'm here today
  4. with Professor Shane Jensen, Professor Audi Winer, both professors of statistics and data science. Some
  5. combination of the three of us and Cade Massie are here every week on Wharton Moneyball. And of course, this is the
  6. last show of 2025 as we're sitting here on December the 30th recording the show.
  7. It's been a great year of interesting things in statistics and data science, especially applied to sports. Even more
  8. so happening on telecast now, even broadcasters using them more and more, including more in writing. Uh something
  9. I talked with Aaron Shatz about last week on the show, we're starting to see award winners even more reflective of
  10. advanced statistics. So it's been a and you know we now have thanks to Audi and Wasabi we have a Wharton sports business
  11. journal here at the Wharton School which has supplemented the already great publications that are happening at the
  12. application of statistics data science and sports of which I think it's fair to say the two of you I wouldn't say the
  13. pioneers but the two of you were in the early boat legitimizing statist the application and methodological
  14. development of statistics and data science and sports and so uh it's great to be on the show with you guys for the
  15. last show of uh 2025. >> Yeah, I I'll take the acknowledgement. Shane and I we did remember that years
  16. back when Shane had first started as a professor and we got money from ESPN um to do research into baseball statistics.
  17. That's the paper that we wrote I think is still maybe my top favorite paper of all time.
  18. The only reason the only reason I didn't call you guys the pioneers is of course
  19. you know uh Shane remembers I mean we were both graduate students at Harvard stat uh Fred Meller had written many
  20. many papers of course the original Efron and Morris shrinkage estimation paper uses baseball data um you know our
  21. former colleague and you know friend who passed away recently Dave Schmidline wrote a number of papers including you
  22. know can a hot goalie take you win you the Stanley Cup. So the only reason I was saying is I was pulling you guys in
  23. the first wave or in that >> we weren't the first to do sports at >> No, no, no, no. And of course a mutual
  24. >> your son's very old as well. So that can't can't be >> and also we all have a mutual connection
  25. to this person. Of course, one of my adviserss, one of Shane's adviserss, Hal Stern, obviously did a lot of work in
  26. there. Obviously, his PhD adviser was Tom Cover, someone that Audi knew extremely well. Tom Cover is is uh is
  27. possibly one of the most influential statisticians in my career. Not only did I uh go to Stanford because he was a
  28. professor there, but I did research information theory because that's what he did. And he wrote one of the classics
  29. and I today we kind of we don't we discount it, but he wrote one of the most original and beautiful papers in
  30. baseball analytics um called the offensive erra. It's a classic paper from 1975.
  31. Hm. I I it's not a paper I've seen, but I will look. The stuff of Tom's that I was familiar with is because I, you
  32. know, I was dabbling in lottery stuff. He had done a lot of work there. Uh specifically what h how do you infer
  33. sell entries of a contingency table? You only observe the margins, which is the classic thing, you know, for the
  34. lottery. They tell you how much how frequently each number is used in the lottery, but they don't tell you the
  35. probability of the n tupils. Like they will not give how many times, but how do you infer something from that? So I
  36. became familiar with a lot of his work uh but not the offensive erra one. I I I paper it's it's cover and king. We could
  37. devote a um we could we can devote a session to talking about it some future time. We could also talk about how how
  38. the oldfashioned method that they used wouldn't would be um just too simplistic for today's analyses. But uh listen, he
  39. also, you know, Kofer was right there with Shannon and Thorp inventing gambling um techniques to be used in the
  40. casinos. And I remember going to his house as a grad student. He showed me the old roulette um device that they had
  41. made made to predict what fraction of the wheel they expected the ball to land in with excess probability. And they
  42. would let they would use that to uh place bets on a roulette wheel. Incredible. and they just f felt it was
  43. too risky. I mean, because of course it was monumentally illegal to use that kind of device, [laughter]
  44. but but they they they were doing it. And of course, Thorp um who was at MIT wrote the first book called Beat the
  45. Dealer on on how to beat blackjack. Impenetrable book. Today, you can you can ask Chat TBT to make you an expert
  46. in a few hours. Um but back then, he had that book and it was impossible to decipher. And I remember spending quite
  47. a bit of time learning how to be a a uh a a a competent card counter from that book. Um it kept the the pool of people
  48. able to do it low because the barrier to entry was so high. Probably a good thing for all. Well
  49. guys, we always um today we don't have a guest on the show. Uh it is December the
  50. 30th or we could have gotten a guest. I figured why not just you know you guys are the guests. So, I figured I'd take
  51. this opportunity to, you know, what caught your eye in sports. And of course, we always do it from a
  52. statistics angle because we're a sports and statistics show here. So, Shane, I thought why not? Let's start with you
  53. and we'll just alternate between you and Audi and I might chirp in every now and
  54. then with some stuff, but Shane, let's just start with you. >> Well, I mean, obviously, I guess what's
  55. caught my eye. I'm absolutely over the moon with Drake May and the Patriots this season. And I mean, the my
  56. expectations were low at the start of the season. I think most people's word, but to have somebody who is in the MVP
  57. conversation and to be a team in the conversation for the number one seed in the conference, I just I I mean, it's
  58. one of the many I think very unexpected things uh to happen this season. Not not
  59. the least of which is kind of some of the sort of top contenders not being anything uh this year. So, yeah. So, I
  60. guess that's what's kind of caught my eye. I continue to be amazed by the run they're on. And maybe you get you can
  61. kind of temper my excitement a little bit or something. Tell me something that you know makes me not absolutely gaga
  62. over this Drake May situation for potentially the next 10 years. >> I would think something you might be
  63. equally happy about and I'd love your thoughts on this. Maybe just from a statistical perspective or even if it's
  64. I I don't know that I'm looking for a precise answer, but how you think about it, I think you have to be equally happy
  65. that Mike Frabel's the coach >> because you know, you put in an average coach there and you know, it's hard to
  66. know what the counterfactual is, but it's not obvious that there's a 13 and3 team sitting there. It's not obvious.
  67. You know, Mike Frabel's relatively young for a coach and he's got a long way to go. He could be there for all of Drake.
  68. Matter of fact, if he's rational, he should be there for all of Drake May's career. How do you think about the role?
  69. >> No. No. I mean, I I think you it speaks to I think the difficult like, you know,
  70. kind of the good fortune I guess the Patriots perhaps have had here or just the difficulty of turning around a
  71. franchise that had like a top three pick in the last like three or four drafts is
  72. that you need to kind of, you know, teams that are that bad usually need both the coach and the quarterback kind
  73. of like refreshed. And getting both those kind of an elite level version of both those at the same time is not an
  74. easy feat. And of course they correlate with you know kind of I I I think they feed back on each other once you uh have
  75. them together. But again, the I guess good luck basically that the the Patriots have had that I think that's
  76. really facilitated their quick turnaround. Whereas I think the Jets might have had I think the Jets had a
  77. good coach and uh you know Salah a couple years ago but not the good quarterback and they you know it's hard
  78. lining those things up. >> So let me ask you a question now. Um how much do you basianly update? In other
  79. words, I think we all know last year I think they were three and 14 or whatever they were last year. They weren't a good
  80. team last year. >> Yeah. Um, let's suppose this year your prior might have been, well, Drake May
  81. is going to improve. You know, you got Mike Brael, maybe there would be a 500 team. Well, let's put this way. I think
  82. a reasonable confidence band might have been between six to 10 wins somewhere in
  83. that band. Um, they've obviously exceeded that, but now that they're in the playoffs,
  84. possibly at least the two seed, maybe the one seed. I think they've locked in at least the two.
  85. >> Yeah. >> Yeah. Well, well, I mean, I Yes. I think Jacksonville can maybe catch them, but I
  86. mean >> Oh, you're right. Maybe. So, okay. But they're one of the top. >> I mean, they're hosting they're
  87. certainly hosting a home their first they're hosting game in the divisional round at the minimum.
  88. >> Um and and possibly uh skipping uh the uh wild card round two. They still could
  89. >> right how much how much do you believe they have a legitimate chance to win the Super Bowl?
  90. >> I mean, it seems very wide open. I I mean I I can talk myself into it. I I mean and Audi probably wants to jump in.
  91. I I they haven't I would love to see them beat more good teams. You know, they haven't did they did not play the
  92. toughest schedule. >> Just so you know, Aaron Shats just on before just to interrupt just for one
  93. second statistic. I think I tweeted out or exed out on W Moneyball last week. Aaron Shatz had an interesting stat
  94. which was this Patriot team up till last week. I assume the same might be true has the third easiest schedule since
  95. 1978. Like it's a fact. Like that's >> Yeah. I mean it's certainly not going to look harder after last week.
  96. >> Yeah. Right. So now maybe it's the but I'm Yeah. Go ahead, Audi. >> I mean so this is what this is what
  97. confuses I mean, all season long people thinking that that we're saying that the
  98. Patriots are the best the I mean, the the the most overrated team g I mean, given their opponents
  99. >> and they just, you know, and there's been and this is a short season. I mean,
  100. no matter how you slice it, we've got 16 games, not a lot, right? And the standard deviation in a football game is
  101. about 14 points. So, what how truly good are they? And and and that would be the
  102. my question. >> I think that is the primary question. I I would just say that that all all the
  103. AFC teams I think it's just particularly wide open the year that all the AFC teams kind of have also questions. I
  104. mean, Denver has had a tougher schedule, but has been barely ekking out their wins and, you know, I mean, if you had
  105. to talk I mean, they would probably be my pick to go to the Super Bowl, but like but you know, I could certainly
  106. argue that the Patriots could beat them in a game, you know, and and and similarly, you know, the I mean, the the
  107. Jaguars, I mean, think about all the top, you know, the teams that are going to be playoff teams. I think you could
  108. argue certainly um the pay I bet you the Bills will be something like the sixth or seventh seed and will probably be
  109. favored in like every >> Let me just let me just bring let me just bring that up by the way. It's
  110. actually something very interesting. So um this is something that's fascinating.
  111. I did I did just use chat GPT for this but it's not that anybody could do this. It's not that impressive a thing. So
  112. right now the betting favorite for the Super Bowl, the betting favorite to win the Super Bowl are the Rams plus
  113. 475. Now they lost yesterday to the Falcons. >> They're going to be like the five or six
  114. seed, right? >> They're likely the six. They could be the five. So I just want to say this
  115. again to everybody and I want to add a couple stats from Aaron Shats and then Audi get your perspective on this. Right
  116. now, they're the betting favorite as potentially the six seed facing the world champion Philadelphia Eagles on
  117. the road. Okay. According to Aaron, according to his DVOA metric, which he invented, you
  118. know, it's something that pretty well known in the analytics literature in football, the Seattle Seahawks
  119. and the um Rams, according to sorry, the Seahawks and the Rams, according to him,
  120. by the way, are two of the strongest teams like in the top 10 ever by his metric. So do you find it
  121. odd and I also this is something I did do by the way I did ask chat GPT took it only about 15 seconds to do I told it to
  122. uh provide me a plot on the x-axis of the betting line implied probability so on the x-axis is you take the betting
  123. line you get an implied probability of winning the super bowl on the y ais is a strengthbased probability of winning the
  124. super bowl now we can define what that is it ended up using some ver some amalgam of ELO and other based ratings
  125. from profocus NFL.com etc. I asked to do a bariat scatter plot for me and I also
  126. asked it maybe to Shane's point tell me the and so then asked to put a 45 degree
  127. line on the plot and then tell me which team is farthest from the 45 degree line. It turns out, by the way, it's got
  128. the Bills being the farthest, meaning it has the Bills to Shane's point stronger than the implied betting odds
  129. by the largest gap of any team. So, Aie, any >> hold on just just sure I'm correct. So,
  130. I understand very clearly what the implied odds of winning the Super Bowl are. That's easy. Um you're using now a
  131. probability of winning the Super Bowl using using essentially a statistical model. Is that
  132. >> so what it did is I'll tell you what I asked it >> and then I'll aren't the probabilities
  133. pretty small. >> They are small. I'll tell you what it did. >> So I just this is literally I always
  134. think it's important when you tell someone you used a large language model. What exactly was your prompt?
  135. >> Here's exactly the only thing I asked it. I'll say verbatim to our listeners here on Morton Moneyball. I mean,
  136. tomorrow will be a different answer anyway, even if it is. >> Yeah, of course. But here's what I said.
  137. I said, provide me a plot of betting line implied probability on the x-axis versus strength probability on the y ais
  138. for each team to win the Super Bowl. And so, and that's what it did. And I know what it did because it now that chat GPT
  139. 5.2 into thinking mode exists. It's telling me what it's doing at each step. So, I'm watching it. I'm not just
  140. getting the output of this. I'm watching it. It's creating a simulation. It simulated 10,000 playoff scenarios going
  141. forward and simulated those out based on strength parameters from an amalgam of NFL.com, Profoot Focus, and ESPN. That's
  142. how it got the strengthbased probabilities. And the other ones it got from I think it was BM MGM and some
  143. other and some other betting line. That's that's what it did. >> No. So did you adjust for the vig
  144. directly or did you not bother with that? >> Good question. It automatically now
  145. adjusts for the vig when it it I didn't tell it. I just told all our listeners exactly what I typed in. It in its
  146. thinking adjusts for the vig when computing those Super Bowl-based probabilities. In other words, it know
  147. it it adds up to more than one because of the vig. It renormalizes it because of the vig. It does that automatically
  148. now when you ask it to give you an implied probability. >> I assume the point is that the bills are
  149. are are the bills the one that are the most diff. >> Yeah, they're most likely to win.
  150. They're the biggest they're much more likely to win according to the statistical model than the implied.
  151. >> That is correct. They're the most under underbet team if you want to call it that. They should be
  152. they should have a better betting odds than given the strength model uh predicts. Uh they have the largest
  153. deviation from the 45 degree line. Now now the betting odds have to do with um how much action there is. So on some
  154. level you can say it's because they're being underb, right? Not enough people are are taking them. Um or it could be
  155. that the statistical models have see things that the public just doesn't agree with.
  156. >> I'm kind of fascinated on the other side here that the Rams um are the you said where the betting
  157. market comes. >> They are there. This is from from the from ESPN and FanDuel. The Rams are the
  158. number they're both at they're plus call it 475. >> Because you know the number uh the two
  159. other teams in their division are playing off for the number one seed. Well, that's that's why I brought up
  160. this upcoming weekend. So, I think that's why I'm speaking to you guys. >> I I mean, I I I don't know if again,
  161. this is maybe uh >> that is interesting. I don't know if there's some kind of built-in
  162. >> um you know, Stafford effect or something like that, though. He certainly looked uh less than impressive
  163. uh last night. >> Well, let me ask you guys a question. How much value do you guys place?
  164. Suppose I say the following. Um, I'm going to keep the teams the same because obviously if you take Stafford off the
  165. team and you put some average quarterback on it, obviously the Rams are very different. How much value do
  166. you put on, you know, let's look at it. Rams, Shawn McVey has won the Super Bowl. Denver Broncos, Shawn Payton has
  167. won the Super Bowl. Mike Frael, I know, hasn't won the Super Bowl as a coach. He's won it many times as a player.
  168. >> Tommy Mc has also lost the Super Bowl. >> He's also lost the Super Bowl. Yeah, but
  169. he's been he's been to the Super Bowl and he certainly has his teams have performed well in the Super Bowl. Do you
  170. guys put any value on the coach, let's call it let's even winning the Super Bowl cuz maybe that's a that's too
  171. specific a metric. How about playoff success? Like suppose we did the followup the po I'll give you another
  172. coach that's won the Super Bowl. How much would you change the Rams odds or the Broncos odds if I told you Mike
  173. Tomlin was the coach of their team? Well, I I mean, >> he's won the Super Bowl. I
  174. >> I mean, obviously, I think playoff success does move the needle a bit, especially coaches that have
  175. demonstrated playoff success. At the same time, you know, I think the McVey, it's based much more on what we know
  176. McVeyy's body of work in general than his particular playoff success. For example, Nick Serriani, I bet you has a
  177. better playoff record than Shawn McVey or than probably anybody we're talking about right now. consider him based on
  178. that like probably the best playoff record of any playoff coach out there right now. Would you consider him the
  179. best head coach kind of among the playoff teams? Obviously not or at least I wouldn't either. So
  180. >> no I would not. >> So I I kind of think it's you know there's a lot of small sample randomness
  181. to kind of play off record. Shawn McVey has been around for long enough where I think yeah I think his body of work both
  182. regular season and playoffs clearly suggest he's elite coach. Let me just stick with the Patriots for a second.
  183. Then Audi, I want to get to you about what caught your eye in sports. There's an going to be an interesting debate
  184. about the MVP of the league this year. I think it's down to one of two players. I
  185. think everyone agrees it's either going to be Drake May or Matthew Stafford. I think that's likely to be true. Okay.
  186. What's interesting, back to the analytics versus not passing yards, Matthew Stafford.
  187. Touchdown, Matthew Stafford. touchdown interception ratio Matthew Stafford. But if you look at any let's call it more
  188. advanced metric >> any rate stat >> Drake may >> so I think it's going to be fascinating
  189. and by the way not that there's let's call let's we all look at there's a recency bias I don't think particularly
  190. Matthew Stafford paid particularly well in the game last night against Atlanta and I would say I don't know um I'll
  191. just put what Shane put in the rundown May was the first player in NFL history I understand it's you know they p cherry
  192. picked these He threw for five plus touchdowns, 250 plus yards, and completed 90% of his passes in the game.
  193. Not bad. >> Yeah, he was out of the game by the middle of the third quarter. >> Right. So, I all I'm saying is I think
  194. this will be another piece of evidence because based if if I just took the names off and I just provide you the
  195. advanced statistics, I think Drake May would win the MVP this year. I do. I personally, it's not about deserve it. I
  196. actually do not think he will win because I think Stafford leads in the traditional metrics, especially
  197. touchdowns by so many. I think it'll just overwhelm it. But that's just my opinion.
  198. >> Yeah, you're probably right. Um I I I think it's kind of similar to, you know,
  199. kind maybe like the Mike Trout uh um Miguel Cabrera kind of, you know, MVP race from a few years ago where Miguel
  200. Cabera, you know, Mike Trout kind of had the sort of rate stat like was was was dominant but Miguel Cabrera was, you
  201. know, had had these kind of gotty kind of totals. >> Absolutely. >> So Audi, what caught your eye? Why don't
  202. we move on to you? I mean, not that by the way, this is the nice thing. We've talked about we talked about the
  203. Patriots in some sense, but we got to talk a little bit about coaching. We got to talk about, I'll call it advanced
  204. stats for the MVP versus kind of more traditional stats. We got to talk about the Rams and maybe they're the um you
  205. know, they're the third, you just pointed out the third best team in their division and they're the favorite to win
  206. the Super Bowl according to the betting odds. They're the third. Forget again, you might say, well, strength is they're
  207. the third best in their division. They play other teams in their division the same as Seattle and the 49ers do.
  208. They're in the same division. >> All right. So, a lot of things have caught my eye. I I I'll start at the
  209. bottom. Um the Jets, that's my my defunct my def de facto favorite team or historically the favorite team. Um how
  210. awful they've been and is just sort of disappointing and also considering I I been spending time thinking about Sam
  211. Darnold. If you're thinking if Aaron Shat says the Seahawks are this, you know, one of a great team, how could how
  212. does Sam Darnold go being so bad with the Jets to being I mean, I don't I don't think he's particularly
  213. extraordinary with the with with Seattle. I mean, he's had a very good season, but and so it makes me wonder
  214. like how how good are we really are at assessing the quality of these players? And I think that reflects back on our
  215. discussion about Drake May. I mean, what do we really know? It does. It this season does seem to be like ex with the
  216. exception of some very awful teams including the Jets, the Giants, the Raiders. Um, and if you look at it,
  217. there's there's at least three or four horrible teams. There's a a whole bunch of teams top 10 that are not necessarily
  218. that distinguishable from each other. Um, and then I mean I watched the the Eagles game um on uh Sunday and they're
  219. not a good team. I mean they're they're not a good team. Um >> they have a defense. They have a good
  220. defense. Can >> they have a good defense and by and defense just just to talk about as
  221. general defense is the least predictable aspect of a football team. And I mean that not only from next season
  222. >> to next season that doesn't predict well. But even next game it doesn't predict that well.
  223. >> Um so what do we really know? And so I I'll turn it to you. Um uh is the Jets
  224. are the Jets problems so managerial coachwise ownership? What do you see the problem there? I mean I mentioned them
  225. but Jets pay people just I mean they put their hand on their back and they go h it's really a weird feeling. I mean what
  226. is it? >> I so let me first say it's it's I don't I don't buy particularly any of that
  227. franchise stuff like franchises matter, ownership matters. um them having first class facilities matter, but obviously
  228. the Jets have a great market, infinite wealth, um all of that. I think you have to I go back to what Shane said earlier.
  229. The Jets sooner than later have to decide whether Aaron Glenn is the coach that can build and turn around that
  230. team. Because if he's not, what's going to happen is you're going to draft probably, you know, uh, another
  231. quarterback, right? >> Yep. Yep. You are. And now the question is, if he can't develop that
  232. quarterback, maybe it's the quarterback's no good or maybe Aaron Glenn's no good as a coach or who has
  233. offensive coordinator is no good. Shane, I know you wanted to jump in, please. Yeah, I mean I would just sort of I
  234. think what the Jets have exhibited over the last like like part of their continual mediocrity and I think it's
  235. kind of partly the age we live in I suppose is there there's no patience, you know, like like Aaron Glenn's got
  236. like a year or two before, you know, he's going to get canned and he doesn't look I mean I mean, you know, I'm not
  237. really I don't think he necessarily is a keeper, but you know, part of the reason, you know, the Darnolds, you
  238. know, and and the Gino Smiths and everything of the world is that you've got like a year or two to really hit big
  239. or you've got essentially a year to hit big in New York in the New York market or or you know all of a sudden the
  240. pressure is on you starting to be considered a bus. I kind of like look at some of some of the like like Pton
  241. Manning's first year. Look at look at like or or Rogers sitting for a few years. Like there's like a lot of I I I
  242. think we >> Tom Brady's first year >> that's not I I I mean sure if you want that for the standard, but like I I
  243. think people like need to be more patient with quarterbacks and recognize that they'll often struggle because
  244. they're usually on a bad team um coming out of the draft and and you kind of have to build a whole organization
  245. around them. And I think, you know, the Jets keep kind of hoping for a quick fix
  246. or like, you know, that a number one will come in without support and like, you know, just be awesome out of the
  247. gate. And sometimes that does happen. You know, there there have been, you know, teams that have kind of hit very
  248. quickly like that. But I I think >> I think there doesn't seem to be a lot of long-term vision there.
  249. >> Well, I'm going to say I'm gonna let me say one thing about Sam Darnold, then I
  250. want to ask a question. So, let me start with Sam Darnold. So, I just saw something a stat the other day. There
  251. are only six quarterbacks in the history of the NFL to win 13 games two seasons in a row.
  252. Okay, Jim Darnold is now one of them. Now, great. Maybe it's teams around him. Maybe they're be he's being coached
  253. well. Now, of course, Brady's on that list five times, but let's I'm just I'm counting Brady just once. I'm not
  254. counting Brady five times. But Sam Darnold is one of six quarterbacks on two different teams now. By the way, he may
  255. be the only to do it on two different teams to win 13 games in a row twice. I mean,
  256. it's incredible. I mean, at some point, you have to give something to Sam Darnold. Great. Maybe he's a system
  257. quarterback. Great. He's a system quarterback that maybe win 14 games and be the number one seed in the NFC.
  258. Just make sure you didn't say he's hasn't won 13 in a row two seasons in a row twice.
  259. >> He's won 13 in a row. >> But I'm saying we're two different teams. Yeah, he's done it twice last
  260. year and this year he's still got a game to go. But I'm saying he was the Vikings
  261. quarterback last year and now he's >> Yeah. >> And that was my point. I mean, he's not
  262. just good. He's very good yet we all thought of him as terrible with the Jets. >> That's my point. Really?
  263. >> Maybe I'll give you another name you speak to. Tell us again the whiner rating system of Trevor Lawrence. Maybe
  264. it takes a little bit of time and how's Trevor Lawrence playing? How would you feel? By the way, if you're the Jaguars
  265. right now, I know there's there's probably half a dozen teams there if I know the Patriots would be one of them.
  266. If you're the Jaguars, you wouldn't trade Trevor Lawrence for anybody right now.
  267. >> No, like not remotely. But my question is this. Let's maybe But there's a handful of teams that
  268. Yeah. No, no, no. I'm saying there's a handful of teams that would make that same statement. The Chargers might make
  269. that statement. The Patriots would definitely make that statement, but the Jaguars are now one of those teams that
  270. would make that >> like top 10 quarterback. Finally. >> Oh, yeah. Let me let me ask you guys a
  271. statistical question. Yeah. >> Know, one of the things that we brought often in baseball, we have these ways of
  272. measuring the quality of a player outside of their quality of opponent and quality of the team around them. for
  273. baseball, it's easy because the their their players around them don't matter that much. Opponents matter um and we
  274. and we and we can adjust for that and we can generally measure the quality of a player in the abs in the in the outside
  275. of their context. >> Have we gotten any qu ability to do that for quarterbacks? Is there any way to
  276. really question measures? >> Well, the problem with EPA, and I'm I'll bring this up. problem with EPA is
  277. you're if you do you really care about EPA or do you really care about >> I mean are you ask
  278. yes I mean as as a peripheral as a ratebased peripheral it's kind of like do I really as a pitcher do I really
  279. care about strikeout rate I mean yeah >> but that is kind of like pretty what I'm
  280. trying to do >> let me try to answer your question because I looked at this so if you also
  281. look in the rundown down. Although I had a mistake. I I I don't know. I was dreaming the Bucks were 6-1 at one
  282. point. They were 6-2. Whatever this Let me type in again what I asked chat GPT. How has Baker Mayfield's quality of play
  283. declined from when the Buccaneers were I should have said 6-2. 6-2 until now? Please be specific on what metrics you
  284. are scoring him on. Is there any way to know whether it is in quotes his fault or whether it is due to poor coaching or
  285. a poor offensive line? So, I only copied the first half of what it showed, but Audi, it gave me a bunch of metrics
  286. that it thinks are predictive of quality. His yards per attempt has gone down. His passer rating's gone down. His
  287. completion percentage has gone down. His interception rate has quadrupled. His touchdown rate has been flat. his sack
  288. rate has gone up. And so it also had a bunch of other more advanced stats that I I probably should have put there. But
  289. it >> it probably doesn't have it probably doesn't even have access to PFF, but his
  290. PFF kind of grades per game have been going down. >> It had it had PFF grades, but it did. By
  291. the way, this is beauty. By the way, look, just because I'm Morton's vice dean of AI doesn't mean I'm all in. But
  292. let me just say I'm sort of all in. >> No. Yeah. Yeah. You understand? For those of you that have access to it,
  293. >> use chat GPT 5.2 thinking mode. And and let me just say why. It's I'm not saying
  294. its answers are perfect, but I'm saying you get to watch in real time its thought process. It's literally telling
  295. you step by step. So even said Shane, it it's almost said in like you and I were
  296. speaking. It said, "I'd love to have PFF data, but that's behind a firewall." >> It said that. Okay. Yeah. So I I did
  297. some play. Let's get back to the So the metrics. The reason I brought it up is those are metrics.
  298. >> Yeah. No, the thing is so but let's go back to my question. Those metrics would
  299. say that Sam Darnold was bad when he was with the Jets and now he's good with the
  300. two teams that he's been very good with the Vikings and Seattle. What I'm asking
  301. for is it because Sam Darnold is now better or is it because we just weren't evaluating him properly? Well, I mean,
  302. clearly for now, >> I think you can conclusively say and and he also was not good with Carolina, if I
  303. remember correctly. I mean, he's been around. >> He was not >> um >> the Jets did not put him in his in the
  304. best position to win. That that's a pretty safe statement. I mean, the ceiling on Sam Darnold has been
  305. demonstrated to be much higher than the Jets could even, you know, muster. So, that I I mean, there's obviously an
  306. organizational effect here, I think. >> Well, let me ask you a question. We've seen it happen not just to Sam Darnold,
  307. Gino Smith, you know, I mean, >> let me ask you guys both a question. When evaluating Sam Darnold or any such
  308. quarterback, will you allow me to compute a set of metrics, but I'm in a condition on various things? For
  309. example, I might condition on a bin of win loss records, or I might condition on the team making the playoffs. And so
  310. how does his metric how do his metrics look compared to other quarterbacks that have played on similarly successful
  311. teams? And I won't even use postseason I won't use postseason outcomes. I'll use
  312. let's take all teams that have won 10 plus games. Look at the metrics of their quarterback. look at in a multivariate
  313. way where Sam Darnold lies on those multivariat distributions and try to say is Sam Darnold good compared to
  314. quarterbacks that have been on successful regular season teams. Is that a fair to condition on that and do it
  315. that way? >> I mean yes and one one piece of information we that have already conditioned is you know that previous
  316. MVP discussion we had even though he is on a 13- win team did Sam Darnold come up at all?
  317. So we kind of implicitly know his metrics are perhaps, you know, not what we at least see as driving the success
  318. of his team. >> You have more do you have another way of so I mean my my here's what I would love
  319. if I if I had if I were the, you know, the the uh the oracle or the you know the deciser who can decide what I can
  320. observe. I would love to observe to know how much pressure there is on a quarterback
  321. >> on every play and and I want it. Well, can you um >> You can you can get it publicly on on
  322. every play. How much? >> Yeah, you can get some good proxy pocket last. I Yeah, I should have
  323. printed this out when I asked the question about Baker Mayfield. The advanced stat it had was pocket pressure
  324. rate, how he did against pocket pressure. Those were stats that it was found. No. So those are aggregate
  325. information. What I would like to do is on every play, every drop back, every I want to know how much pressure and I
  326. want to know how open the receivers were and and and so what I'm trying to do is
  327. get something that's like like a what we would call in baseball peripheral. Like
  328. how good at you are doing your your basic job, which is open. Exactly. PF, right?
  329. >> No, PFFs does that. Yes, they do that with with ratings by by raiders. They like look at it and they Yeah, they they
  330. do that. You can get a B you can't the peripheral you can get is a B kind of a a binary count variable but playbyplay
  331. did they do a good job or not as graded. Yeah. >> But you would agree that we're not far
  332. away from a video enabled large language model trained by humans to be able to ingest
  333. film data. Determine some definition of pocket pressure. determine some definition of how open the receivers
  334. are. Determine some definition of expected yards gained given this situation. I mean that data will become
  335. like even if it's not PFF doing it. Um somebody could that AI that is what AI is
  336. extraordinary at. It's trained to do that some blackbox version of that soon. >> Yeah. Doing that's what they're doing in
  337. soccer. are doing it in and continuous. >> Nobody will be telling it to anybody.
  338. >> Yep. >> Well, guys, >> all right. But before I let you go, guy, before you go on to your your I just
  339. want to say just >> I know this is football football, but we are at the baseball solstice today at
  340. this very moment. We're exactly halfway between the last pitch of the World Series and the first pitch of spring
  341. training. It just needed to be noted. >> Like it's been a particularly cold stove. No,
  342. >> it has been the news. >> Well, guys, since we are I just one other quick thing. This is not this is
  343. not genius math. I just wanted to see if it could do what we teach in our basic stat classes. We obviously have might as
  344. well we should spend a few minutes on it. We have four NCA games coming up. Uh NCA football games coming up. Um Oregon
  345. a 2 and a half point favorite over Texas Tech even though Texas Tech was the four
  346. seed. That's fine. Indiana by seven over Alabama. We could question, you know, if
  347. we look historically, there's no reason. I mean, this is another thing like how much weight do you put on Alabama's
  348. historical success? Should they be less than a seven-point underdog? We have Georgia 6 and a half point favorite on
  349. Miss. We have Ohio State 9 and a half point favorite on Miami. I just asked Chad GPT a very simple question, which
  350. is what's the probability that at least one of the underdogs advances? And of course it can do it can do what I call
  351. you know high school level math stat. The probability of least at least one is one minus the probability of none. So I
  352. I put this in the rundown. It actually computed that. It assumed independence. It took the betting odds. It implied it
  353. implied win probability. It can I mean it's not shocking that it can do that calculation. But I'm just saying you
  354. know it's not like I gave it any real I didn't tell it where to get the data from. I didn't tell it to do, you know,
  355. it could have done it in a more brute force way. All I'm commenting on is that apparently there's an 81% chance, which
  356. again, if you asked most pe This is why we need a show like ours in my view. If you asked most people on the street,
  357. what's the probability that Oregon, Indiana, Georgia, and OSU are all going to win? They're going to say a
  358. lot more than 19%. A lot more a lot more than 90. >> You think so? I mean, aren't aren't
  359. don't most people think of the games is pretty tight? >> No. >> No. >> See, they're one tight game, which is
  360. Oregon, Texas, Tech, and and maybe, by the way, a lot of people have Oregon as their prediction to win the whole thing.
  361. Um, and then I, you know, the betting odds don't have Indiana, Georgia, or Ohio State as particularly close games.
  362. I don't think anybody would be surprised. As a matter of fact, if you remove the Oregon game out of there and
  363. I said, "What's the probability one of Alabama, Old Miss, or Miami advances or at least one?" I think that probability
  364. would probably be no more than one minus.8, so I don't know, uh, maybe 10%. Yeah, I just felt like most people
  365. thought there of there not being an overwhelmingly dominant team in uh, college football this year. I think I
  366. think I remember Kate saying like OSU and Indiana are significantly above everybody else this year. We could
  367. debate whether Georgia, you know, Georgia will be there or not at the end, but either way, I just thought it was
  368. interesting. I I just thought that it's not that the calculation was very sophisticated or getting the data was
  369. that hard or sophisticated. I just thought it surprising that there's an 81% probability assuming independence,
  370. which by the way, I'm not an independence guy. I look there's no reason the game should be dependent on
  371. each other. No reason whatsoever except >> there's none. Except I can tell a story
  372. where if the first game happens and let's say you know well it only takes one of these events to happen for my
  373. event to happen that I asked for. But let's say Miami plays a really close game against OSU. Does that give any
  374. increased belief for whatever that's worth to any of the other underdog teams? Any rational
  375. >> answer that can I go first? Yeah, >> I would say yes. It means to suggest you
  376. have model miscalibration. Um, and remember the same model is applied to always. Will it
  377. >> exogenous latent variable like it's you know what >> crazy monsoon rains throughout of the
  378. United States that don't that basically you know knowing one game 63 tells you a
  379. lot about the Nets game being 6-3 or something like that there I mean that that's an example of something that
  380. could cause >> No but I think more dependent model if Miami I mean >> Ohio's pretty heavily st favored over
  381. Miami. No, that's that's the first game by the way. That's actually there's >> if Miami wins.
  382. >> Yeah. >> Um there there can be one simple explanation for that is they just they
  383. just you know underdog wins, right? Big deal. And that's probably the the more likely explanation. But this an
  384. alternative explanation is your model was miscalibrated to start with and my posteriors on on each of those
  385. possibilities will in will change after Miami victory. And that so if Miami wins, I would suggest that there's
  386. probably a a greater probability that I've done something wrong and the the model is wrong across all of them. And
  387. so I would therefore imagine that the numbers are that it's more likely. >> I hadn't thought about this connection,
  388. but let me tell you about a paper I'm literally just about to submit with a doctoral student maybe today, tomorrow,
  389. as soon as we finish up a couple things. So, I'm sure you guys know this in the this a good statistics point for our
  390. listeners out here on Morton Moneyball. And again, this is Eric Brado. I'm here with Shane Jensen and Audi Winer, some
  391. combination of the three of us and Kate Massie here every week on the Wharton podcast network. So, you guys know,
  392. let's assume you're being basian for a second and you assume that your data generative process is Gaussian. So, a
  393. normally distributed, right? And let's say you put a prior distribution on the mean of that distribution which is also
  394. Gaussian. So you have what's called the standard normal normal model. The nice thing is they're what's called conjugate
  395. which all our listeners probably know what I mean by that. The posterior is normal. But one thing for sure is true
  396. is that the posterior variance has to be lower than the prior variance. That's just a mathematical result of the normal
  397. normal model. Now Audi so it turns out that the minute you move away from these families that of course isn't true.
  398. Matter of fact, if you just make a simple assumption of a mixture distribution as your prior, then that
  399. phenomenon goes away. The reason I bring this up is I'll use your words. I'll just repeat back to our listeners what
  400. you just said. Let's suppose Miami beats Ohio State. So, one possibility is now I'm less certain about my model. I have
  401. to raise the posterior variance compared to the prior variance. Now, I've got an
  402. observation that suggests my model might need a wider variance. Now that means even if I don't change the mean
  403. strengths of the teams of the other six teams playing the probability of the lower team winning would go up just by
  404. the widening of the distribution. And no, no, I know I'm saying of course because you're a statistician, but I'm
  405. just pointing out to people that >> wait and I'm not even sure that your first assertion is true that the
  406. posterior variance has to be less than the >> It is 100% true. >> Even if even if there's a mean bias
  407. because there's that, you know, the discrepancy part of it where there's there's the extra variance blow up if
  408. the prior mean and the data mean are super misaligned. >> So there's two ways by the way in a
  409. mixture you can actually get increased variance. One is if you get a larger separation of the means which we call a
  410. change in polarity and the second way you can do it is of course if the uh within group variances change but under
  411. the standard normal normal model if you look at the equation for the posformation information adds so you can
  412. say 1 / sigma^ 2 + 1 / to^ 2 inverse that's the posterior variance it always has to be smaller than 1 sigma squar So
  413. the normal normal model always implies always implies a lowering of variance no matter what information signals received
  414. and I'm pointing out that that doesn't recognize the fact that there could be a model misspecification or other things.
  415. So I'll be happy to send you the paper Shane. It's also we're going to post it on SSRN but I find it and the example we
  416. use is I don't know Shane what's your you live down in center city. You live in Philadelphia. What's What's your
  417. favorite restaurant in Philadelphia? >> I like Barcelona down in East Pass. >> Barcelona. Great. And so you have a
  418. strong belief that it's a good restaurant with a pretty narrow posterior. Okay. Because you've been
  419. there a bunch probably and you like it a lot. And now all of a sudden imagine you
  420. go to Barcelona and you have a bad experience. You could easily imagine your posterior variance being a little
  421. bit wider than it was before. But the normal normal model will never allow for that. it will only allow for a smaller
  422. posterior variance no matter what that information signal is. And that's just an it's a mathematical restriction of
  423. the normal normal model. And the reason I know this, by the way, I found this out. I I'll use Audi as an example.
  424. Audi's a brilliant man. I give him an I'll use the words of educational testing because I ETSs for a number of
  425. years. I'm giving Audi an SAT. Audi's getting questions right right. I think Audi's really smart with
  426. a low posterior variance. I give him an easy question. He gets it wrong. Well, a
  427. normal normal model will lower the variance of Audi's ability, but of course any rational person would
  428. probably raise the variance of now because now he's gotten now I'm less sure that he's as smart as I thought he
  429. was. And so I published a paper of this. This is my first ever academic publication. I literally in homage to
  430. Fiser and it is related to that. I called it negative information that you can actually receive signals that raise
  431. posterior variance but not under a simple normal normal model. >> That would also but you would also think
  432. I'm stupider too. >> Yeah, it would take the mean. >> Yes. Yes. Yeah. I'm talking purely about
  433. variance stories. Of course I can compute the posterior mean. I mean of course that's going to go downward. That
  434. wouldn't surprise anybody. But the and the normal normal model will allow for that. It just won't allow for the
  435. variance to go up. It will not allow for the variance to go up. Either way, that's a total aside. It was just based
  436. on your point about if Miami were to somehow beat OSU, we have to put more variance now in our model beliefs, which
  437. would by de facto raise the probabilities of the underdog teams. >> You got it. That's a good that's a
  438. technical explanation for what I was predicting. Um, >> I want to say what caught my eye, and
  439. I'd love to get Shane's thought on this. So, now this is the third week since I pointed this out.
  440. that the Colorado Avalanche still only have two regulation losses. They've played 38
  441. games. They've won 29, lost two, and lost seven in overtime. Now, by the way, you might say it's an
  442. equally statistical anomaly that they've lost seven in overtime given how good they are. They're on pace for 140
  443. points, which would break the record of 135. But, you know, I say this every year about halfway and the teams peter
  444. out and they might end up in the high 120s. I mean, there's no way you would predict that they're going to win 140
  445. points. Shane, just give I mean to me, they've played almost half the season and they've only lost two regular season
  446. games. Isn't that nuts? >> Yeah. I mean, it's it's it's it's it's a good record. Like you said, I I I feel
  447. like we've done this to ourselves several seasons now. um in the last like decade or so where uh I mean first of
  448. all even if they kind of maintain that pace you know you know like they may not even want to be a historical pace going
  449. into the postseason because the last couple teams that were on that historical pace flamed out in the first
  450. round of the postseason >> like the Lightning I remember at least >> Lightning and the Bruins the Bruins who
  451. have the third point record that team that year they flamed out quick >> yeah they won they lost in the first
  452. round same as the and they have the record at 135 points Yes. Yeah. Yeah. I mean, again, these
  453. are all sort of the 82 game kind of records or whatever. You know, this is uh you know, the Montreal Canadians had
  454. like uh I think won like >> 131 or 132 or something for years for like 40 years theirs was considered the
  455. real record. But uh but yeah, I mean I I do think they are on a historical pace.
  456. The specifically the two losses does yeah stand out. I mean, if they had if they finished the season with like, you
  457. know, say less than, you know, eight losses or something like that, that'll probably be like the loss record
  458. certainly. Um, and I know, I mean, they they've just been absolutely excellent uh um looking unstoppable. It just it
  459. does seem I I think it is a curiosity that the last few years we've seen these kind of seemingly unstoppable teams roll
  460. through the regular season and then get immediately stopped. I don't think it >> is there any reason you see that um any
  461. reason I maybe we've talked about this I just don't remember why you know I think
  462. it was the the Canadians record was something like 1977 I may have or 81 or somewhere whatever '7s
  463. >> in the 70s okay and then that record held for 40some years and then all of a sudden now the Bruins broke it and now
  464. we're seeing pot and the lightning had a great season and then the Bru any explanation for that I mean just there
  465. are a bunch of really crappy teams teams. We're seeing the variation in team strengths go. Is that what's
  466. happening? >> Yeah, it does seem like that. But I don't actually know. I mean, we should
  467. need we we would need to get a real hockey expert on that kind of is looking at that distribution, I guess, season to
  468. season a little bit more. It could just be kind of these sort of one, you know, I mean, c certainly comparatively
  469. to like the when the Canadians were doing, I feel like the lead is less dynastic. I mean the Canadians back I
  470. mean this is pre free agency and all this type of stuff. I I feel like you you'd kind of think that most of the
  471. sort of forces that are around now influencing hockey would go towards more parody not less. Uh but again the tail
  472. behavior of that we can we've sort of seen in baseball that like you know there's been obviously a lot of parody
  473. kind of in general I think over the last couple decades in baseball in terms of the teams that contend versus not and
  474. some of the forces in play but you know there still is like occasionally you'll have a season where there's like three
  475. 100 win teams and there'll be like none the next season. I don't I don't know the kind of tail behavior even if you
  476. don't necessarily have a even if you have a relatively stationary system. I I I don't know what I don't have a good
  477. intuition about that. >> Yeah. I think as I remember maybe just go back to baseball for a second. Am I
  478. I'm right. Right. There was no 100 win team in baseball last year. Right. >> There was none.
  479. >> And we were shocked. Shocking. Right. >> Dodgers started off with like 13 wins in
  480. a row or something and they were their preseason prediction was nearly 100 to start with and after winning so many we
  481. would have thought for sure they would have won. We we thought it we made those.
  482. >> We've also had a couple teams that have kind of hit historical lows over the last couple years in baseball. Is that
  483. the new rule or is that just kind of a a two-off kind of scenario or something? >> Yeah, I lost that bet. Remember this
  484. year I I predicted I forget who it was. Was it the >> Colorado? You thought they would be
  485. >> Yeah, I thought and I don't know how many did they end up at 50. >> They they did not set any records. No,
  486. they they >> was the record, but I I know that I had predicted I think I predicted like 35
  487. and you guys said they're going to start winning. >> Yeah, I just followed Gton. You know, I
  488. regression to the mean in the end of the day. Hard to top. Well, guys, in the last few minutes, let me throw out some
  489. topics just to get your reaction to. I put this in the rundown before last night's game where he got injured and
  490. now he's out for four weeks. I'm talking about Nicole Joic, uh, the three-time MVP of the NBA. Let me just tell you
  491. guys his stats. 29.9 points a game, 11.1 assists, and 12.4 rebounds. He leads the league in
  492. assists and rebounds and he's averaging 30 points a game and he's a center. So at some point
  493. we're going to have to start Am I wrong that we're going to have to start listing this man? I if he hadn't injured
  494. his knee he's literally injured it last or yeah injured his knee. He's out for four weeks now. He might have won the
  495. MVP again which puts him at four MVPs but he's got three. Not bad. You're an NBA guy. Is this one of the
  496. greatest players in NBA history? >> I mean, starting to look that way. I I I think it's kind of um
  497. you know, I mean, I I think with these er era kind of comparisons, I actually am not the student of his that NBA kind
  498. of history. Um so I I I guess you know, if we can kind of try and talk at least anecdotally about the kind of dominance
  499. he's displaying, what the what what kind of a historical analog would be. Would it be kind of Shaquille O'Neal? Is that
  500. kind of, you know, I'm trying to think of sort of center points and reboundated in in this gamechanging way. I know
  501. Shaquille probably doesn't have the point totals, but would that be kind of the closest analog that we've been
  502. looking at? >> I think so, except he's a much better passer. I'm pretty sure I don't know
  503. that this is true. I'm my I'm going to say a statement and our our listeners at W Moneyball can say Eric doesn't
  504. remember when this happened. I'm pretty sure Wilt Chamberlain was the last center to lead the league in assists.
  505. >> Wow. Yeah. So, that's pretty amazing. >> I mean, to me, the 11.1 assists are more
  506. impressive for a center than the 30 points and 12 rebounds. I mean, that's impressive, too. But you I'm sure you
  507. could look it up while I'm speaking here. I'm I'm I know I know for a fact Chamber led the league in assists one
  508. year, but it might be the last time a center led the league in assists, >> right? And and would you sort of say
  509. like you know physically like is he even the kind of most unique sort of center in the league right now because you've
  510. got somebody like Wemmen Yama. So like you know >> oh Joe gets can't jump. He can't run
  511. fast. He he's got no lift. Um he's not particularly muscled. Um might have you know no I know.
  512. >> Is he he is he the most uninterestingly dominant player ever? [laughter] >> Yes. Yeah. I would say that's I I'd say
  513. that's I would say that's fair to say. You know, it's also, you know, since I'm a big Larry Bird fan, I was always a fan
  514. of his that and partially because of that they always said, you know, Larry Bird couldn't jump, he couldn't run
  515. fast, you know, um, etc. But his knowledge of the game, you know, and his ability to lead teams, I think when I
  516. see Jokic play, I'm just, you know, guy almost never makes the wrong play. Either way, I just think we have to
  517. start talking to him about the best ever. Um, I you mentioned golf. We have to, you know, in homage to him today.
  518. Um, today's Tiger Wood's 50th birthday. >> Wow. Congrats. >> Young man. Young man. You know, you have
  519. to remember for me, Tiger Woods, we were both at Stanford at the same time. Um, so
  520. >> his undergraduate and your graduate school. >> Yeah. I I was finishing my graduate
  521. degree. But here's my question. >> Yeah. something magically happens in the world of golf which could help Tiger
  522. Woods when you turn 50 is there's this thing called the senior tour big advantage you get to use a cart and
  523. we know part of the challenge is he says I can swing a club I just can't walk well you can use a cart third they only
  524. second they only play three rounds instead of four so this is one of those interesting
  525. questions like I mean he can still play the Masters by the way and he can still play all the majors I think I'm going to
  526. make a forecast here. I think he moves to the senior tour just because his body he just I mean he has played one
  527. tournament in two and a half years. He could play as many senior tournaments as he wants and then plays the four majors
  528. like if he comes back why Shane any thoughts? >> Yeah. No, I mean the only thing I would
  529. sort of say is you know we don't know kind of like motivation for actually coming back you know I mean he he could
  530. just ride off into retirement. He certainly got the accolades and everything like that. I think it will be
  531. kind of whether his accumulated injury history if he can kind of have that more great you can kind of put it together
  532. physically for the for the senior tour. I mean it would be wide open. I mean he'd instantly be favor, right? I think
  533. coming into that uh situation. So um so yeah. No, I mean I would I kind of I would love to see it. I just you know um
  534. don't know necessarily it'll be kind of I think a motivation factor. So, I have one more topic I want to ask
  535. you guys about. So, I I wasted time on I guess it was Sunday morning and I watched this awful made for TV tennis
  536. event where Nick Curios do you know who Nick Curios is, guys? >> Yes, he's a pretty bad tennis player.
  537. >> Well, he's he used to be okay. He was a top 10 player at one point. He made it
  538. to the Wimbleton final. He lost in the Wimbleton final. He's currently number 672 in the world
  539. >> because he hasn't played much. Um, he played the number one woman in the world, Arena Sabalanka, in a best of
  540. three match. Okay. And but the reason it was awful was because you only allowed one serve, both players.
  541. >> His court was actually wider than her court. So literally, Shane, they made a
  542. court that looked like this on one side of the net and then like this on the other side of the net.
  543. >> For our listeners who weren't seeing, it was substantially bigger. his his he was
  544. he was very limited. >> He was Yeah, it was limited. He won 6363. >> And by the way, he was also toying.
  545. >> He was No. No. So, look. So, I asked Here's what I asked GPT. Can you rate the strength of men's and
  546. women's tennis players in one ranking? Now, here's what it did. It did it. But I'll give it credit. It says tennis
  547. abstracts ELO is calculated separately within each tour's match pool because there's no cross tour match data. The
  548. absolute ENO numbers aren't calibrated for men's versus women comparison. So this combined list is best read as top
  549. by ELO within their respective tours. So I give it credit for recognizing that. Interestingly,
  550. Center and Alcarz are at the >> I'm glad TPT is trying now pushing back on some of your more uh more elaborative
  551. plans here. >> I agree with that. Um it has S and Alcarez on top and then it has the next
  552. four players are all women. Sabalanka third, Switite fourth, Rabbakana fifth and Koko Goff sixth. So my only point is
  553. is that um it is true it was this funny match but look even in their primes remember this isn't Billy Jean King
  554. against a 55year-old Bobby Riggs by the way you know which was the original Battle of the Sex's back in 73 I mean
  555. Curios is a 30-year-old former top 10 player and they're playing best of three um you know even in her prime Naverova
  556. they always wanted her to play the and her comment was she couldn't even beat the top hundth man in the world and said
  557. it wouldn't even be close. She her comment was she would have lost six love six love to number 100 man in the world.
  558. All I'm saying is this to me was not an event to show the relative. That's what we I mean that's kind of what we're
  559. getting at is if it if it was really a fair sort of you know fair playing surface literally um
  560. >> yeah play the match >> the usual what we call a tennis match but ranking with the top women like what
  561. would be the kind of fair ranking of like say number one in the uh women's side versus like the men's side and you
  562. know Naverova's estimate is like for herself when it was herself is like at the 100 level. No, no. She said she
  563. would have gotten beaten easily by theund. She might put herself at 500. >> So either way, it was just I I just
  564. found it interesting. I I thought put this way. Um it was an interesting match to watch,
  565. but it didn't tell me anything about like I do want I wouldn't mind answering the question. I mean, look,
  566. >> why don't every tournament we just kind of have like a round like some play like
  567. like you know the people that like lose out in the early rounds? Couldn't we do a little bit of extra like kind of like
  568. games to like like across the across the g uh genders to sort of uh try and sus this out slowly?
  569. >> I love even today I love women's tennis. I might love women's tennis more than I
  570. love men's tennis. I love women's I love I have so much uh respect for the women's game. I love everything about
  571. the women's game. But you know just as an academic I must admit you know I I have to admit I'm thinking to myself is
  572. there some way because look we did this at ETS2 like if there's no overlap between two like can I like let's take
  573. an a simp that's not a simple example but it's a simple enough example suppose I want to know who's smarter someone
  574. that gets a 780 on the physics AP exam and someone that gets a 780 on the French seven AP exam. And of course, if
  575. there's overlap, then you could potentially try to statistically answer that question. But I'm trying to the the
  576. bigger picture I'm talking about isn't men's versus women's tennis, cuz I I love them both. Um what do you do when
  577. you don't have overlap between distributions? Do you guys remember the bridging eras in sports paper? That's
  578. the way they did it is that you overlap and you know Mickey Man Ruth played with
  579. Dagi or Garrick played with Dagio and Deagio played with Mantle and then Mantle played with this so you have
  580. overlapping designs but here we essentially have a zero overlap design a can be done well and some overlap you
  581. got the mixed doubles right >> you know that's right >> that's not I mean a is there anything
  582. that can be done >> no in fairness I mean until you have unless you have un there is some
  583. overlap. It's just not at the professional level. You have to go down to the amateur level and from there you
  584. can make some you can do something. But listen, it's it's it's monumental. It's just absolutely I mean it was not only
  585. Martino I think it was also did wasn't it um Serena Williams didn't she say she'd be beaten six love six love
  586. against >> Serena Williams said the same thing >> and uh and and basically she says it's a
  587. different game. I mean they they they're much faster. They hit much harder. Um, and they're not they're not I mean the
  588. it's just to speak back of that about that overlapping errors paper. You can't even do that without without modeling.
  589. You still need >> I mean you the way you can achieve overlap is through model assumptions
  590. through model assumptions. >> I mean it's not like timing. I mean if with swimming and track it's just a time
  591. you can you can compare that. Um that's about the only sport that has an absolute metric. And even then people
  592. talk about Jesse Owens he didn't have he didn't have the track. He didn't have the shoes. He didn't have, you know, and
  593. then you you might want to wonder what would how fast would Jesse Owens have been if he had been trained and and
  594. geared up in today's world. And people try to make forecasts of that. >> I think what we can all agree on here
  595. is, you know, since time happens, weather happens, different equipment happens, there's no
  596. really such thing as an exact applesto apples replication. Like you you you always need to like there's like and we
  597. can decide those variables are >> under the assumption that time is a straight arrow and not a flat circle.
  598. >> Yeah. >> If time is a flat circle we will come back to a replicatable. I mean we we do
  599. like I mean a great question for us as statisticians to ask is across all sports sporting events but baseball
  600. football not only team but also individual which what is the oldest what is the performance of yester year that
  601. would still be considered dominant by today's standards considering there's so much improvement
  602. in athleticism gear etc and what obviously comes to my mind immediately is secretariat who still has all the
  603. records and that's about 50 years ago know um >> 52 years still holds the record all
  604. three tracks >> and [clears throat] it's not even close I don't think um and so beyond that I
  605. mean you can't you can you can go back to you know Hack Wilson 180 plus RBI's 190 that's a different time you wouldn't
  606. even think about it Chad Williams hitting 406 again it's a different time you can't it's hard to compare them no
  607. tennis player um no none of the athletes from the basketball or baseball or football era would you imagine um If you
  608. transform them to today, unless you this sport has changed too fundamentally in each of those.
  609. >> Maybe the the one that came to my mind I I like I like Hack Wilson. I like all of
  610. those things >> to me. I I forget if he won 10 11 Michael Phelps at the Olympics
  611. that won 11. >> Yeah, but but listen, Michael Phelps records are all gone. His last record
  612. >> I just No, no. I'm not saying his records are still there. I'm just saying the his dominance dominance
  613. >> every event he competed in >> eight gold medals in the Olympics. Yeah. Yeah. >> Oh no. I mean that even the fact that he
  614. beat out that Spitz guy who was dominant and like the fact that there seems to be
  615. like once a generation somebody who comes and completely sweeps across >> it. It says to me that maybe this maybe
  616. it's less impre I mean Phelps was the one that did it. So I'm not taking anything away from him. But right it's a
  617. once in a generation >> sport where you can sweep across 12 medals because they're all basic.
  618. >> I mean what I mean here question be what would Chamberlain have looked like if he
  619. had grown up in today? >> He probably would have been I mean he wouldn't have been trained the way he
  620. was back then >> or Bill had a completely different competition, a different style. would
  621. have been I mean you so you do have to recognize that that people wouldn't would would go into the you know one of
  622. the things I remember my favorite was when I when when my kids were on college tours I'm not not sure you did this Eric
  623. you probably you're all your boys went are went to Penn but my kids were looking at different schools and I
  624. remember once um I was at at Yale for a reunion and and I had a senior or a junior in high school and I went to the
  625. admissions committee and they said none of you would have gotten in today based on the standards and most of us are
  626. looking each other is and we wouldn't have looked on paper the way we looked back then today because we would have
  627. been in the same environment as the kids today we would have done all the things
  628. that they would have done to make themselves look differently and similarly with athletes I mean if you
  629. put Will Chamberlain in today's world he would look look like he looked back then
  630. if you put Babe Ruth in the world today he wouldn't be eating hot dogs like that
  631. it wouldn't happen right that world is gone he would have been forced to be as as as uh as athletically and
  632. nutritionally minded as any athlete is >> why we like the secretariat example because we're not Oh, Secretariat would
  633. would have slacked off or something like that. [laughter] >> Oh, Secretariat had heart though, right?
  634. >> Take that out of the equation. >> Well, Secretary had the biggest heart. We have to remember
  635. >> biggest heart we've always heard. Well, guys, >> if we keep talking, we're going to be
  636. talking right into the new year. >> All right. Thank you. This has been a great edition here of Wharton Moneyball
  637. and the Wharton podcast network. Uh on behalf of all of us, myself, Audi Winer, Shane Jensen, uh we'd like to thank you
  638. for joining us today. uh and abstensia Cade Massie, we like to thank you for being a big part of our show for all of
  639. 2025. And um we always like to think I hope all of my co-hosts agree with me, the best is yet to come, 11 and a half
  640. years, but uh we've got a lot more in us. And so please join us for our next show. Thanks to our producers, uh
  641. Marissa Ren and De Patel. Thanks to our associate producer and sound engineer, Dion Simpkins between now and next week.
  642. There's a lot of it. Enjoy your sports. Enjoy your statistics. and we'll see you
  643. next week here on the Wharton Podcast Network.

Brief Summary

The episode discusses sports statistics, standout players, coaching impacts, and the evolving role of analytics in sports.

Key points

  • End of Year Reflections. The hosts reflect on a year filled with advancements in statistics and sports.
  • Drake May's Impact on the Patriots. Shane shares his excitement about Drake May's performance and the Patriots' surprising success.
  • Betting Odds and Team Strength. Discussion on the discrepancies between betting odds and statistical models for NFL teams.
  • MVP Debate: May vs. Stafford. The MVP race is heating up between Drake May and Matthew Stafford, with stats on both sides.
  • Sam Darnold's Surprising Success. Sam Darnold is now one of six quarterbacks to win 13 games two seasons in a row.
  • 81% Chance for Underdogs. A surprising statistic reveals the probability of underdogs advancing in college football.

Episode highlights

Tension

40/ 100Low tension

Related episodes