The Limits of Prediction in Sports and Elections

Knowledge at Wharton

Ask about Knowledge at Wharton

Episode Description

ABOUT THE EPISODE Polls, prediction markets, and statistical models can all offer useful forecasts. But what happens when their errors move together?

Eric Bradlow, Adi Wyner, and Shane Jensen break down correlated error, the limits of election forecasting, and how early results can help people assess whether a model is missing something important.

They also discuss MLB roster construction, NFL parity, college football playoff scenarios, Conor McDavid’s historic scoring pace, tennis rankings, and the Sixers’ early-season performance.

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.

----- 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. Warden Moneyball Ready! Let's do it! Number of game-winning drives in the fourth quarter. That's deceptive.
  2. You really should compute, did this person exceed expectations, as opposed to, did they win the game?
  3. Swung on! In deep to left! There she goes! I think we are going to see expanded playoffs.
  4. I mean, although I could have predicted how you feel about it, by, does this change the game from when Teddy Roosevelt watched it?
  5. If yes, then I oppose it. Diego Luna got him on ice skates, shot, it's a goal! With respect to basketball, a three-point brings you 50% more points.
  6. It's supposed to be compensated by an appropriately lower probability of success.
  7. But it's just not happening in today's game. Curry, wide open three! Foul! Curry from downtown!
  8. Michael Jordan was the best of all time. And, if your answer is Jordan, you're a little bit older.
  9. If your answer is LeBron, you're both wrong, and a little bit younger. From the campus of the University of Pennsylvania Wharton School, this is Wharton Moneyball.
  10. Welcome. Welcome to Wharton Moneyball, the podcast edition here on the Wharton Podcast Network.
  11. This is Eric Bradlow, professor of marketing, statistics, and data science here at the Wharton School.
  12. Some combination of myself, and you can see today I'm here joined by my two longtime friends and co-hosts,
  13. Professor Adi Weiner and Professor Shane Jensen. Between the three of us and Cade Massey, we're here, I'm going to say every week, on Wharton Moneyball.
  14. And, of course, we're here talking about lots of different topics around the worlds of sports, analytics, and its intersection with business.
  15. And today, guys, though, we have a unique day. I thought it's important that we start a little bit with the following.
  16. We record, as all of our listeners know, on Tuesdays. Today is, of course, Tuesday, November 3rd, which is Election Day.
  17. And it is a big day in the world of statistics, because there are these things that happen months and months leading up to the actual Election Day, which are these polls.
  18. And so that, of course, is we in marketing. I'll put on my marketing professor hat.
  19. There's issues of stated intention data. Someone can state one thing and do something else when they actually vote.
  20. There's, of course, the issue of sampling frame. Who are the people being contacted?
  21. There's the issue of the sampling bias. There's response bias. There's all kinds of issues that are involved with samples.
  22. And then we've also got something else, which has opened up in the last couple of years, if you'd like,
  23. which is now these Kaoshi and Polly market, which are contract. I don't want to call it betting.
  24. People writing contracts on who they think is going to win. And I think what all of us wonder is, you know, what's going to be the most accurate of anything?
  25. Is it kind of the polls? Is it betting markets, which are kind of a wisdom of the crowd feature?
  26. Is it prediction models? So, of course, Nate Silver has his prediction model out.
  27. All kinds of others. Cook Political Report have their mathematical prediction models out using historical data.
  28. So let me first start. I was going to talk about a different topic, which is correlated errors.
  29. I want to get to that. But let me just start with you. The biggest business stories move fast, but the headlines only tell part of the story.
  30. On This Week in Business, Wharton faculty break down the news and trends impacting businesses and consumers.
  31. Listen to This Week in Business from the Wharton Podcast Network. You have three possible things, and maybe your answer is going to be model averaging in an ensemble.
  32. But if you had to, can you tell our listeners how you think about survey-based predictions?
  33. I'll call them model-based predictions or betting market-based predictions. OK, well, betting markets, I'll start with first.
  34. I mean, they're not doing anything other than aggregating everybody's wisdom, which can be very effective because it's able to do things without a model.
  35. So the surveys are designed like what we would call scientifically. Let's be quite honest.
  36. There's a lot of art form in that as well as science. You know, you have to get a representative sample.
  37. That's very hard to do in practice. It's getting harder and harder. How do you get people to participate?
  38. Nobody participates in surveys, right? Come on, guys. You get a random call. You're not going to take it.
  39. That's already introducing a lot of bias. They're clever enough about stratifying.
  40. There are advanced techniques, but they're susceptible to bias. Then, of course, there's the problem of noise.
  41. You just don't have enough observations. And then thirdly, people can change their mind.
  42. That does indeed happen often at the last minute. The betting markets, you call them the prediction markets, contracts.
  43. It's really a bet. Let's be honest. They can be good because they can get a sense of the direction of things.
  44. And I know this is going to be a little traumatic, but 2016, we sat here in the studio with Cade, all of us,
  45. arguing about what's going to happen in the election. And there were betting markets at that time, and there was polls,
  46. and it looked like it was a slam dunk. Hillary Clinton had that wrapped up. And I was arguing with you guys that, damn it,
  47. there's been enough elections in the past year where the polls just didn't make any sense, I mean, didn't jive with the elections.
  48. We just have to add that extra randomness here without any cause. And this is what the betting markets do.
  49. They just throw that in just thinking, you know what, there's something we predict might go badly.
  50. Just to be clear, Adi, what you're talking to our listeners, there's noise one can add, which could not change the mean prediction,
  51. but change the variance. Or there's, as you said, maybe it's not what we call white noise
  52. that's just being added, like independent noise state by state. Maybe there's correlated noise.
  53. Could you talk to our listeners about why it matters, whether it's independent or correlated?
  54. Well, actually, it matters enormously because if you're trying to do an aggregate model to say, for example, the Senate goes Democratic
  55. or the House goes Democratic, that depends on a lot of elections. A model that treats each of those results as independent error
  56. due to modeling and whoever responds in surveys will predict a, make a forecast with uncertainty that's too low.
  57. Correlated error means if one House race goes one way unexpectedly, that's important, unexpectedly, probably many others will go there too.
  58. So it's not the multiplicative unlikelihoods or likelihoods which drives to zero of extreme events in multiple places.
  59. They all just kind of go along. And that's what happened actually in 2016. There was a problem in one place,
  60. and it just percolated throughout the system. So that was, it was the model that was wrong.
  61. The people who said that it was a slam dunk were wrong. But I will point out back when Obama was elected in 2018, 2008,
  62. wait, not 2008, 2008, Nate Silver became well-known because he predicted the results of every state.
  63. But it really wasn't that hard because if one went one way, they all went in that same way.
  64. And it was, it wasn't, it wasn't like, it was just me to finish here. Just the, it looked like it was this impossible task.
  65. You know, there's about 10 or so tough calls and he picked all of them. One over two to the 10 is like one in a thousand.
  66. They aren't really because of correlated errors. It was more like one and a half or one and a quarter.
  67. So that was, I remember that. Who's counting though, Adi? Sorry to interrupt Adi, but I just want to kind of point out that what you need,
  68. kind of, I think what happened in 2016 is you don't, you didn't just have correlated errors,
  69. which I think is drove kind of the extremeness, but it's also, you had a close election.
  70. So it's like, you know, I think where the correlated errors really manifest is like, you know,
  71. you might have a few things that are like, you know, a lot of different states that are like near 50-50
  72. and they'll all kind of flip in the same direction, like for, you know, 52-48 or something like that to have a very, you know,
  73. I mean, obviously the Clinton-Trump popular vote was very, very close, but, you know, the congressional map, you know,
  74. and her congressional map was like fairly, the congressional map was, you know, kind of switched the other way because of that kind of correlated error
  75. in close elections. Yeah. Let me talk about that in a couple of ways. So I'll give, you know, the Nate Silvers of the world some credit.
  76. So what recently they've done, and Adi and Shane, I'd be interested in your thoughts on this.
  77. I have three comments. I'll only do one at a time. So his first comment is given the recent polls and the averages,
  78. if one looks at the distribution of errors and the correlated distribution of errors over the last 20 or so elections,
  79. and one draws one from the, let's call it the high point of the distribution, the middle of the distribution and the low end of the distribution,
  80. in other words, low correlation, moderate correlation, high correlation, even under the high correlation scenario, he says,
  81. the Democrats have over a 50% chance to win the election. I thought that was actually quite a clever analysis.
  82. What he's not, he's assuming that the error structure shares some commonality over time.
  83. He's drawing, like Shane said, a magnitude of that, adding that onto the existing polls and then saying,
  84. does it change the outcome? Does it change the outcome of a state? With a really high correlated error model,
  85. it sounds like they're still above 50%, but does it, is it much, how much more are they for a low correlation model?
  86. How much does that correlation matter? You could imagine, like the way he's kind of framed it is,
  87. like a big polling error that would cut across states might be in the 4% to 6% range.
  88. I think we'd agree that would be a significant polling error. Wait a minute, just quickly tell me the numbers.
  89. I'm going to tell you right now. The predictions are the following. There's probably 10 states that are between,
  90. let's call it the 45% to 50% mark right now. In the polls, Democrats are leading probably
  91. in nine out of the 10 of those states. Now they're leading in most of them by more than 4%,
  92. somewhere between 3% to 6%. So the idea is even if they suffer a 2% to 4% moderate
  93. correlated error, they still win more than half of those states, which means they end up with more than 50 votes, 50 Senate seats.
  94. That's kind of the analysis that's been done. And I think that's pretty reasonable.
  95. I mean, it has a loss function in it. It has a decision theoretic argument to it.
  96. It's using an empirical distribution of errors over recent elections. I mean, guys, I don't have any criticism of that.
  97. Shane, I'll start with you. Yeah, no, I mean, I think that's about as well as I could have done it if I'd taken up the enterprise.
  98. I mean, you still kind of wonder if there's something idiosyncratic to Trump-involved elections, basically,
  99. that kind of like isn't being captured with that. Well, he's not running. No, no, no.
  100. That's important. He's in the background, certainly, of this situation. So like, yeah, recent times, let's say.
  101. But let me also say. I guess if I was to modify methodology at all, it would be interesting to see how much it changes
  102. if you oversample from kind of recent elections then when establishing that norm.
  103. Let me also say something else that's interesting. People, because they know we're statisticians.
  104. People say, so, Eric, are you going to watch the election? I said, of course. And then people say, so what do you think?
  105. And I said, here's what I think. And Adi, I think you'll like this. Since I'm Bayesian, what I'm going to do is I'm going
  106. to observe the data from the first couple of states. So what I'm going to do is I'm going to observe Maine,
  107. which closes early, which has a competitive Senate race. I'm going to observe South Carolina, which ends early.
  108. I'm going to observe North Carolina and South Carolina, which end early. Then, and Florida.
  109. And it's not so much about who wins. It's about how close are they to the forecast
  110. and do I see correlated errors? And I make the claim that after observing those four states,
  111. I'm going to have a fairly good sense for me about both the accuracy of the polls
  112. and the prediction for the final evening. Adi, what do you think about that? That's my claim.
  113. Yeah, I would do the same thing. I think, I mean, I'm expecting it to be correlated
  114. but not so correlated like we saw in 2016 or even 2020. I do think the mood of the country is already baked
  115. into the forecast, which is a blue wave. I'm, of course, watching very closely election
  116. in New York for the governor. Of course. My daughter works for the governor. So I want to make sure that-
  117. Governor of New York is looking strong. Yeah, so that's the first election I'm looking at
  118. because that's the most important for me, family wise. But after that, and actually that election,
  119. it could be interesting, not because I don't think she's going to lose, but I do think that if the Republican
  120. is looking better than predicted there, that would be a very interesting piece of data.
  121. It doesn't have to be just in the Senate races. The governor races could be informative about this as well.
  122. But it's also interesting, I give them credit in the following sense. Actually, Nate Silver did something I thought
  123. was really great. He obviously did a massive simulation. He computes a wide confidence interval.
  124. In other words, the 48 to 54 for Democrats is all within the 95% confidence interval.
  125. But then what I liked he did is he talked about the highest probability simulates.
  126. In other words, what's the modal simulate? What are the ones around it? He also then said,
  127. here are what extreme simulates look like if it's highly correlated errors. So the thing I liked about it is,
  128. I mean, Shane, obviously neither you nor I know what's under Nate Silver's hood or anything else like that.
  129. But what we do know is that sounds like a pretty, not only principled way to talk about it,
  130. but how you describe a very complicated, high dimensional distribution to a lot of people.
  131. Like here's the modal outcome. Here's ones near it. Here's one right at the 95th percentile.
  132. Here's one at the five percentile. Here's one at the extremes of the percentile.
  133. I thought that seemed pretty reasonable. Yeah, and I think also taking into account
  134. kind of the weird kind of like that you have these kind of discrete, like these kind of values within state
  135. get discreetly mapped to count. Like if you were simulating a Senate total or something like that,
  136. there's this discretization phase too that you have to, or I'm thinking of like kind of the congressional
  137. counts when we do presidential elections as well. There's like that extra level of complex math,
  138. you know, kind of count math on top of, you know, all these like sort of the simulations
  139. within each region. Yeah, what's interesting, of course, is the House everyone considers
  140. pretty much done at this point. Like the Democrats are going to control the House.
  141. What's interesting is always looking at the gap between the House and the Senate.
  142. And part of it, of course, is there are more House races. And so you could argue the,
  143. you know, let's call it as Adi was describing, the overall Democratic advantage just nationally will play a much more
  144. direct role in the House races, of which there's 400- And also like incumbent effects
  145. and stuff like that, right? I mean, I assume there's much stronger incumbent effect in the Senate
  146. than the House as well. So you kind of, it's- Much, much. It can kind of move with the
  147. very recent feelings of the nation a lot more of the House. Yeah, what was actually also interesting is,
  148. you know, I'm sure people will end up studying this over time. When you look at the time series of predictions,
  149. the, I'll call it, the mathematical markets, the polling markets, and the betting markets
  150. had some discrepancies for a while. What's now interesting is I'm watching them today.
  151. I literally have three screens, three things up on my screen. I'm looking at all of them simultaneously,
  152. is that they've essentially, not shockingly, they've essentially converged, which means there's almost no discrepancy now
  153. between the most recent polling averages, betting markets, and let's call them mathematical prediction markets,
  154. which probably make sense. That's just my opinion. Can you bet a parley on, say, the Senate and the House?
  155. That's a great- The answer is you can. You absolutely can. As a matter of fact, Donnie,
  156. you brought this up. So if you go on to, well, write a contract, I know that's what you technically mean.
  157. If you go on- Yeah. If you go on to Cauchy and Parley markets, they give you the entire two-by-two matrix,
  158. of course. Dems win this. Senates win that. You can bet on any of them. They also give you,
  159. just like in football, you can bet on an exact number. Dems win 53, 47, 52, 48. You can bet on any, sorry,
  160. write a contract on any permutation you want. You can bet on the House on ranges.
  161. Democrats will end up with 230 to 240, just like in sports betting. But you can also take the contracts
  162. to infer a probability. Absolutely. Because they imply one. So I'm curious to know, for example,
  163. a simple one would be, what's the difference in the conditional probability of the Senate going Republican,
  164. given the House does, compared to the unconditional probability? So for our listeners,
  165. what that means is a bet price tells you the implied probability of, say, the House winning
  166. and going Democrats. And then if you look at the parlay, that gives you the joint probability.
  167. You divide by the conditional, you get the conditional. You divide by the first unconditional probability,
  168. and you get the conditional by the definition of conditional probability. And that should be informative
  169. about how correlated the markets, those races are. I would argue they should be quite correlated.
  170. So whatever that probability is, it's a bit higher. Did you do it yourself or did you ask your AI buddy to do it for you?
  171. I haven't actually computed that yet. But what I will say is, what I think you'll find is that,
  172. essentially, the probability of the Democrats winning both is more than the product
  173. of the two marginal probabilities. Right. So that's the inflated conditional. Correct.
  174. It's the inflated conditional probability. And that's what I've been tracking kind of in my own mind all along.
  175. I've been looking at those two. So it's just kind of interesting. Well, guys, the good news is we'll find out tonight.
  176. But again, for those people, I'll be looking at the early states to see kind of what the information is.
  177. Shane, do you want to add anything else on the elections or are you good? I'm good.
  178. Okay. Well, guys, so that's been the first half of the first half. Now let's switch over to the regular content,
  179. as we say, which is sports. Shane, why don't we start with you? Kind of been a lot of sports going on.
  180. Obviously, NFL, MLB is not going on, but it is. There's the CBA. There's a lot going on in the NHL,
  181. the NBA, tennis. Why don't you start us off? What's caught your eye in sports? I've been watching a lot of football
  182. over the last week. And what caught my eye, first and foremost, was your Tampa Bay Buccaneers
  183. being the first team to win a game since the seventies with no completed passes.
  184. That's very impressive. Minus two net passing yards in the game. Congrats on that.
  185. Well, I watched the entire game. It was a monsoon. Literally, heavy pouring rain.
  186. And I hate to say it, but you know, Adi, you don't have to be an expert coach in football to know
  187. if you're gaining 350 yards on the ground and it's raining heavily and you didn't fumble the ball
  188. the entire game, maybe you should just keep running the football. You got it. Momentum, right?
  189. By the way, let's be clear. They only threw three passes. It's not like Baker Mayfield
  190. was 0 for 50. He was 0 for 3. Yeah. And two of those were dropped, by the way. But that's a separate issue.
  191. Um, yeah, no, that's a big deal. Look, it's a very interesting situation. You know,
  192. obviously, people that listen to our show know I'm a big Buccaneers fan. They're 3-5,
  193. but the team winning their division is 3-4. They went to the Panthers and just beat them
  194. 43-13 in Carolina and they just beat the Falcons, who are the other top team in their division,
  195. 16-0. So, you know, this may be the first time. But Shane, it's not impossible for them
  196. to believe 7-10 could easily win that division. the amount of between-team variation,
  197. between-week variation in all the NFC South teams is off the charts, I feel like,
  198. this year. The Falcons are, I think, the most extreme. They beat the 49ers. But then,
  199. you know, I have... Michael Pennix looked awful. He didn't even look like a professional quarterback.
  200. He threw four picks. And, if they had let him continue to throw the ball, he might have thrown
  201. seven or eight in that game. Actually, he looked... If I'm going long on the Atlanta Falcons,
  202. I'm going to be very concerned after that football game. Because, like, rarely do you see
  203. an excellent pro quarterback. Everyone can have a bad game, I get it. But, you know,
  204. I've watched 50 years of football. I've never seen, you know... I mean, you said yourself
  205. it was, like, a very extreme weather event. How many times have we observed other...
  206. I mean, they didn't let Baker try and do that. I mean, like... You know, the counter argument
  207. would be, it was coaching malpractice to even have a quarterback try throwing it
  208. in those conditions. But, I mean, I'm not... Pennix was drafted when? Can you guys fill me in?
  209. When was he drafted? This is his second year? Second year? I think so. Third year,
  210. right? He was the same one as... I mean, because the question that I always wonder about...
  211. Caleb Williams. The question I always wonder about quarterbacks is, how long does it take
  212. before you know you got a dud? That's a good question. That's a great question. For example,
  213. you look at... You know, everybody who's been listening to me over the years knew that
  214. Trevor Lawrence was the highest rated quarterback or player that I'd ever seen in my draft evaluation.
  215. High school draft... Not high school, but high school analysis that I had done a few years back.
  216. And obviously, he was picked number one, but it took a while before he started to do better
  217. and better and better. And now, he really looks like he's potentially, certainly...
  218. One of the best quarterbacks in the NFL. Yes, exactly. And it took a long time...
  219. You guys... Someone like, you know, Sam Radford. What the hell was that going on?
  220. I think what normal people... What people say... I don't know. This is a good analytics question,
  221. but if people want to post to us at WMoneyball, they can post to us and do some analysis on this.
  222. What most people say is... And the nice thing is, it is consistent with the rookie contracts.
  223. By the end of your third year, most people feel they've got a good sense about whether you're a
  224. quote-unquote franchise quarterback or not. You're right. Shane was right. This is Michael Penish's
  225. third year. He has played 20 NFL games, so he's not a franchise He's not played 40 NFL games,
  226. but he's played 20 NFL games. And, you know, I think this is the kind of situation...
  227. And he's 26. This is the kind of situation where I think the Falcons will make a decision
  228. by the end of this season whether Michael Penish is their future or are they going to try
  229. to draft another quarterback in the draft. Because remember, they had Kirk Cousins
  230. as his backup or starter, and they traded away Kirk Cousins. And he obviously starts
  231. for the Las Vegas Raiders now. So they obviously felt even after year two they had somebody.
  232. We'll see. Speaking of quarterbacks, is Aaron Rodgers still a good one? Or is he just...
  233. Or are the Steelers just looking good around him? He's looking good. I mean, you know,
  234. I don't think he's... He's certainly past his prime, but, you know, I think he's looking
  235. better this year than he did last year with the Steelers or in any of his Jets time.
  236. You know, and they've got, like, a very good team. So it's the type of thing where he can kind of be,
  237. you know, I don't know what you'd say, like, you know, like 60th percentile of quarterbacks
  238. or something like that now, or 70, you know. But that obviously... How about you would agree
  239. with the following, Shane? They can win that division with Aaron Rodgers. Yeah. Yeah, no,
  240. especially with Baltimore, the issues that Baltimore has. Lamar Jackson's probably
  241. out for the season. I mean, no one trusts Cleveland. The Bengals have no defense,
  242. and they got lucky to beat the Titans. They should have lost to the Titans. So, I mean,
  243. why can't... Pittsburgh is kind of looking like that's the strongest team in that division.
  244. It also gets back to what Shane talked about earlier, which is, I don't think Mike McCarthy
  245. is asking Aaron Rodgers to do, to play like the old Aaron Rodgers. He's asking Aaron Rodgers
  246. to do what he does well. He's putting him in position. He's not asking him to, you know,
  247. sit in the pocket. Sit, sit, sit, sit. Wait for someone to come free. It's a lot...
  248. Or shorter passes. Let the receivers do the work. Don't let him get beat up because he can't.
  249. He's not as mobile as he used to. I think he's coaching... I think he's being coached well.
  250. And the play calling is consistent with his current skill set. I do... I mean, I think probably
  251. Steelers fans are a little skeptical only in the sense that, like, you know, they've watched their team
  252. eke into the playoffs for so many years and then not go anywhere in the playoffs.
  253. That's kind of like McCarthy's post-pre-playoff MO as well. But I could talk myself
  254. into a Pittsburgh run this year. But, like, they don't necessarily look like a team
  255. who's only going to eke their way into the playoffs and then just get destroyed again.
  256. Like, I think... I mean, partly it's that I don't see that many really dominant teams in the AFC.
  257. So maybe Pittsburgh is looking a little better just by that. I don't know. Do you see any
  258. dominant teams in the NFL at all, Shane? The Bears look pretty dominant. I gotta say.
  259. I mean, maybe recency bias on that. The Bears... On the AFC side, the Bills look maybe
  260. the strongest but can get got. So, no. I think the years where it was like... I haven't seen
  261. a team, you know, in... out there right now that's as strong as, you know, KC was
  262. at their peak or anything like that. By the way, do we think it's interesting? You know,
  263. every year, because, you know, I was a fan of the Dolphins in the early 70s. At least my brother was.
  264. I didn't really... You know, obviously there's no more undefeated teams. And, you know,
  265. the Larry Zoncas and the guys, they always have a celebratory champagne and cigars
  266. when there's no more undefeated teams. I'm just so triggered by that because of, you know,
  267. 07 or whatever, the perfect season. Yeah. It didn't end perfectly, that one? No, no.
  268. We don't have to relive it. Well, you brought it up. Do we think that it's... Look,
  269. if games were independent, then it's probably not surprising by week seven. But I'll look this up,
  270. you know, offline. I'll post it again at WMoneyball. What do we think the distribution of week is
  271. for which there's no more undefeated teams? And do we think week seven is in the left tail
  272. of the distribution? Wait, who's undefeated now? There isn't. The Vikings lost. I was talking about
  273. kind of waiting time by week of the... Waiting time distribution until there is not
  274. an undefeated team. And whether, like, week seven is kind of a little bit early for that
  275. or not. Well, listen, if it's a coin toss, you'd expect one to go at least 5-0. Right,
  276. because that's one out of 32. And then you have 32 teams. And then the expected number
  277. is there you go. But it's not a coin toss. There are teams that are better. And so,
  278. I think week seven is maybe a little average early. Maybe. I think it's below. It seems below the median
  279. only because I can kind of think back over the last few seasons and think about teams
  280. that were 7-0 and 8-0 type of thing. I mean, even the Pittsburgh Steelers a few years ago
  281. were winning something like 9-0. And again, it wasn't... Right, so this year the longest was 6-0.
  282. It's 6-0. Yeah. So anything caught your eye in sports? Anything you're paying attention to?
  283. Yeah, well, there's of course, you know, there's a beautiful quote by Bart Giamatti
  284. that talks about how baseball was designed to break your heart because it's through the summer
  285. and just as the weather's getting bad and it's starting to cool down and when we need it most,
  286. it leaves us, leaving us to face the winter without that daily. I mean, the thing about you football fanatics
  287. out there is it's not a daily. It's once a week. I mean, your team plays. I mean,
  288. you can obviously watch some other games, but baseball is part of my routine for so many months.
  289. I know. And it takes place in the background in the way a football game doesn't take place,
  290. whether that's the radio or just on where I'm doing something. but I'm still, so now we're into the offseason.
  291. I just want to reflect a little bit on the end. Obviously, the Blue Brewers, they were,
  292. they just, they just juggernauted it. 103 wins. A couple of things I don't think we've talked about.
  293. The entire, the entire playoffs went to chalk, I think for the first time and I don't know when.
  294. Really, really bad year for the Coinslip model. Not any time where it was this big
  295. in terms of this many, but every single series went to the favorite. That's just crazy.
  296. How many, considering how many series there were. And usually. And how narrow the favorites,
  297. you know, the favorites really narrow. Like wild card series, you can imagine some big mismatches,
  298. but by the like end. By the end, it's usually pretty tight. And I mean, even, I mean,
  299. the betting markets probably might, they might have had the Dodgers as the favorite in the betting,
  300. but certainly by record, the Brewers were the better team. And so that's just one,
  301. one nice observation. The Brewers just, you know, upended so many, so many of baseball's new methods
  302. that they need to be studied for a lot of things. This is a team without superstars.
  303. They had some great pitching. Yes. They had Yale, which is past his prime as a superstar.
  304. Yep. Dodgers were well more stacked in that department and as were other teams, but they just work together
  305. as a team. They did things. We had Mike Petriello on. They did, and we talked about
  306. the different ways that they approach hitting in particular. I don't think, I don't want to think
  307. when it comes to the pitching staff, I don't think they were particularly remarkable,
  308. very good, but not particularly remarkable. It's just their approach to scoring runs
  309. was so markedly different than my usual Yankees. I will say a couple of things about them that I think are telling.
  310. One is that the speeches made by Hal Steinbrenner and Aaron Boone after they lost again
  311. was almost identical to the same speech they've made for the last five, six years.
  312. The players have to get it done. It's not the fault of the manager and the general manager.
  313. My response to it, you keep putting together the same team and you're expecting a different result.
  314. They're a team that's built with everybody, with all cylinders firing, with the judge
  315. and back in the lineup. They're there to make the playoffs. They do not appear to be a team
  316. to win the playoffs, at least all the way through. And a lot of that has to do with
  317. they just are so one-dimensional and therefore are easily exploited by top pitching staffs.
  318. We know they've won one title. We know they've won one title in the last 26 seasons.
  319. Have they made the playoffs? Almost every year. Yeah, so that's remarkably... One year they did not, I think.
  320. I think only one year they haven't. That's remarkably tough to do. Oh, it's been more than one, but...
  321. You know, but it's to what I'm starting to formulate as a hypothesis is that what the Yankees
  322. have capitalized and done very well is get into the playoffs. And you can't just do that.
  323. Now, I think for, on average, that's probably just build your best team and that's your best approach.
  324. But when you have... So you have an approach like the Yankees, and I wonder whether it extends
  325. back in the minors, that they actually groom players to have a certain profile and that they're too one-dimensional.
  326. We talk about correlation with elections. How about correlation with types of hitters?
  327. If they're all the same, then what do you... And the people that they go after all seem to be in the same style.
  328. I do think that the Yankees... I don't think it adds up. The Yankees aren't alone in this,
  329. but I do think teams like the Yankees that seem to have... kind of only have the sort of like
  330. three true outcome kind of guys... Yeah, and the other six players can be pitched around.
  331. ...are more vulnerable in the playoffs to high-end pitching, which you see basically constantly now.
  332. Constantly in the playoffs. And so I agree with that. But you gave me good news.
  333. You gave me good news, Shane, while we were prepping for the show with something I didn't know.
  334. Jazz Chisholm isn't coming back. They didn't send him a qualifying offer. They could still sign him
  335. as a free agent. You never know where they might sign him. The thing about... And at some level,
  336. Chisholm is emblematic. A lot of talent in that guy, but not a lot of focus or even a lot of hustle.
  337. And I have to say... How much of that is kind of... Again, I think the question that
  338. you have to ask is how much of that is inherent to... Is a player selection problem
  339. for the Yankees? That they just shouldn't bring in people like Jazz? Or how much of that is
  340. Jazz in that particular environment? Like a development board? They don't run enough
  341. building drills? I mean, I'm being very simplistic. Now you're talking about what we call the intent.
  342. It's the organization. Yeah. Shane, this is the great thing that we analytics can ever answer
  343. is these so-called intangibles. The interactions between the players and each other
  344. and the coaches and the manager. We don't know how to measure that and we haven't yet.
  345. Maybe... I don't think... If you think about it, I don't think Aaron Boone is a pretty...
  346. He's not a kick-in-the-pants kind of manager. He's never... Jazz Chisholm would have had
  347. a very different experience if Girardi was still the manager. Adi, so you work obviously with a lot of students.
  348. You work with a lot of sports teams. You've obviously placed probably more people
  349. at sports teams than anybody currently on the planet. You would have to think that any good analytics organization...
  350. And the Yankees have analytics people. The Brewers have analytics people. They all have analytics people.
  351. You would think that one of the most important models that they would build... And I did this for five years
  352. when I was with the Eagles. The most important model we built was predicting the playoff performance
  353. of various players. Besides just overall performance, how are they going to do in the playoffs?
  354. So I would assume the Yankees can do this. I would assume... I understand there's some sparse data,
  355. but there is... It's a tough prediction problem. It is a tough problem, Shane. But I've got to believe
  356. there is some signal there that they have to notice what's going on. And I've got to believe
  357. that the prediction models would say what you said, Shane. Every season, they're likely
  358. to have three, maybe four guys that are going to be predicted to perform well in the playoffs.
  359. And the other five or six just aren't against good pitching. They're... Adi, what do you think?
  360. There's got to be something. I mean, that's a particular question I'd love to see answered
  361. in some degree. I have a former student who's actually getting his PhD at Boston University now
  362. in statistics, who's actually studying some of these things in baseball. And maybe we'll get
  363. some answers to that. But the big question that I always try to get answered is, is there...
  364. If you try to forecast a interaction, a competition between a pitcher and a hitter,
  365. one has an ERA, you can think of as a WOBA against, and one has a WOBA for the batter.
  366. What's the best prediction of the outcome? And it's usually the average. But I always argued maybe
  367. that there's some sort of non-linearity there that maybe favors the pitcher when the pitchers are great,
  368. favors the batter when the batters are great, obviously creates the right average.
  369. It has to average out, right? And I've never gotten a good answer to that question.
  370. And that might be related to your playoff question. But when you get back to the playoffs,
  371. and what I'm really paying attention to, I just can't wait until the Yankees say goodbye
  372. to Grisham, who I always thought was a statistically bad choice to bring in. He had a great year last year,
  373. hit 34 home runs, but he was defensively not what he was when he was playing, I guess, for the Padres.
  374. And this past year, he was not only the worst defender by stat casts out of average measure
  375. among any center fielder, he had a terrible year at the plate, reverting right back
  376. to his historical mean. And how do you bring someone who's likely to regress so strongly to the mean
  377. just based on the data? And then, of course, talking about hustle. I never felt he was
  378. a particularly hustling guy, but at the Boston Yankees playoff game that I attended,
  379. where you really can see fielding, which you can't see, you really can't see it when you're watching it on TV.
  380. It's at the stadium that you see it because you can see how they react right off the plate.
  381. When you're watching it on TV, they're showing you the hitter. Grisham loafed at a shallow single,
  382. allowing it to turn into a double. One of my students who knows how much I have a distaste for Trent Grisham,
  383. looked it up and said that was the slowest running speed of a center fielder on a fielded ball all season.
  384. Well, he's saving himself to run to first base, right? Right, of course, he's saving himself.
  385. So talk about no hustle. Before I can say that's been the first half of Port and Money Bowl,
  386. let me just say, thanks to ChatGPT, the answer to my previous question is the median is 9.5 weeks
  387. for the undefeated team in the NFL, and the percentile of week seven is sitting there at about 34 percentile.
  388. So it's not really rare, but it's certainly in the lower part of the distribution.
  389. Maybe we could use that kind of as a proxy for my kind of anecdotal observation that I don't see a lot
  390. of very dominant teams out there this year. That would be an A. I don't mind using that
  391. as a proxy for that at all. So that's been the first half of Wharton Money Bowl here on the Wharton Podcast Network.
  392. Stay with us and join us right after the break. Welcome back. Welcome back to Wharton Money Bowl
  393. here on the Wharton Podcast Network. This is Eric Bradlow. I'm here today with my two longtime collaborators
  394. and friends, Adi Weiner and Shane Genshin. Some combination of the three of us and Cade Massey are here
  395. every week on Wharton Money Bowl. So guys, this is an interesting week in tennis
  396. for one reason. So you probably looked at the rundown, but hopefully you have not.
  397. So Alexander Zverev, as you remember, coming into this season was the quote unquote
  398. best player without a major. And he won two majors this year. Of course, Sinner was out
  399. and Alcaraz was out. He's now the number one player in the world by the ranking system.
  400. And does anybody want to guess if you've looked at the rundown, do not guess. He's played 12 matches this year.
  401. Okay. Against top 10 opponents. Anyone want to guess what his record is? He's played 12 matches
  402. against top 10. 12 matches against top 10 opponents. 10 and 2. Yeah, 10 and 2. Yeah, 9 and 3.
  403. Right. 10 and 2. Yeah, it's 3 and 9. But the tricky is this. It's 3 and 9, Adi. 3 and 9.
  404. Against top 10? Who are the top 10 people? Well, I could start naming them. Sinner, Alcaraz, Djokovic.
  405. He just lost to Diminor. So he's just like had like a lot of seeding luck. Yeah, so that's,
  406. thank you, Shane. So that's my point. Adi, he's 3 and 9. And this is the number one player
  407. in the world. So a couple things. He's not the number one player in the world. He's ranked number one.
  408. Yeah, I know. So, and he won two majors. Look, part of it is, as Shane said, is seeding luck.
  409. The second thing is what it represents is, this was true, I remember when my kids
  410. grew up playing squash. You play more tournaments, you get more points. It's not the best 10.
  411. It's the total number of points. You play more, you get more. And so if you stay healthy,
  412. maybe that should be part of the ranking system. And you play the right tournaments.
  413. And as Shane said, it's also, you want to talk about not independent. See, let me tell you why
  414. I think Djokovic right now, in my mind, can never win a major again. Forget that he's 39,
  415. soon to be 40, and his skill set is not quite what it was. He's now ranked number 10
  416. in the world, okay? If you're number 10, in the third round now, you're playing a seeded player.
  417. In the fourth round, you're going to play top 16. In the fifth round, you're going to play top 8.
  418. So his getting good draws is gone. It's gone now. He's going to play four top 16 players
  419. unless they get eliminated to win any major. And I don't think a 40-year-old can do that
  420. for five matches. And that's my point, is that Zverev now, he has the possibility
  421. of getting really favorable draws, more so than someone that's ranked 15 or 20 in the world.
  422. How old was Federer when he had that? Didn't he have like a run right to the finals
  423. of a tournament when he was in his 40s of a major? Yeah. So Federer made it at age 39.
  424. Okay. It was 39. Matter of fact, you may remember this, Shane. Vaguely, I guess I do.
  425. Federer at age 37 had not won a major in four years, and he was stuck on 17 majors.
  426. He won three majors after the age of 37. But he had gone four years with zero. That's fairly remarkable.
  427. It's remarkable. He goes down. One of the ones he lost, which crushed me because I'm a big Federer fan,
  428. was, you may remember, I forget the year, maybe 2018-19 against Djokovic, where he was up,
  429. this was Wimbledon, he was up 8-7 in the fifth serving. The greatest server of all time.
  430. The eight-time Wimbledon champion was up 40-15-8-7. Had two max points. Yeah. This is the one
  431. I was thinking of, by the way. Lost that game, and then it went to 12-all, and at that point,
  432. they had a tiebreaker because of the, you remember the John Isner match that went 70-68,
  433. so they got rid of that. At 12-all, they now play a tiebreaker at 6-all, but they had a tiebreaker
  434. at 12-all, which he lost to Djokovic and lost 13-12 in the fifth. But he was serving 40-15,
  435. and you think about it, how do you won that match? First, he might have 21 majors
  436. and Djokovic might have one less. He would have nine Wimbledon titles, which I'm not saying
  437. that's much better than eight, but nine's better than eight. It is better than eight.
  438. I always felt that match, he probably has to regret more than any match he played in his career.
  439. He was serving 40-15 for the match, which is just incredible. I thought you guys would find
  440. Zverev being three and nine against top 10 players kind of interesting. Remind me, Eric,
  441. a little history for us grandpas or soon-to-be, or I am, you know, future. Connors was how old
  442. his last majors, Jimmy Connors? I'm going to guess. I'm going to guess. I could look it up.
  443. I'm going to say no more than 30. Now, what you're remembering is at age 39, he made the semifinals
  444. of the US Open, but he played a 24-year-old. Why can't I think of his name? The guy that's won three or four majors
  445. with the red hair. Oh, I just can't remember his name. I apologize. No, no, no, no, no, no, no.
  446. Jim Currier. Thank you, Jim Currier. Who became the number one player. Currier, who became number one
  447. in the world and destroyed Connors. Because I just remember that in our time, just winning over at age 30 beyond
  448. was considered ridiculously unlikely. And that Connors was only in the semifinals, right?
  449. Yeah, so just to give you an idea, here's some stats. I mean, Bjorn Borg won 11 majors,
  450. but he retired at age 26. I know he won his last major at 26. John McEnroe won seven,
  451. won his last major at age 27. So Pete Sampras had actually won his last major at age 28,
  452. but then he played one, he had one last run where he beat Agassi in the finals of the US Open
  453. at age 30 and then retired immediately. So, you know, people always wonder how many would Pete Sampras had won
  454. if he had kept playing beyond the age of 30. He had 14 by the age of 30. What I also noticed is
  455. as great as Alcaraz is, I just saw a list of most tournaments won before the age of 24
  456. because Alcaraz is 23. I think he's number 12 on the list. No, no, no, I'm saying
  457. Borg, Federer, Nadal. It was definitely much more of a young person's game. Yes, it was a young person's game
  458. and also they just played much more. They also played many more tournaments. One of the reasons for that
  459. is the prize money was a fractional compared to what it is today. The prize money was a fraction.
  460. Shane, since you're our resident hockey guy, I see you put some stuff in the rundown.
  461. I see this strange thing. I don't know. Maybe Connor David's off to a decent start this year.
  462. Could you tell us a little bit about that? Yeah, I know. Honestly, it hurts me kind of talk about it
  463. because I'm no fan of the Oilers. But yeah, I mean, it's just worth noting. I think in his first 15 games,
  464. Connor McDavid has 39 points, which is, I mean, if you multiply it up, it puts him like over 200 points
  465. for the season, which has only ever been done by Gretzky in the 80s. It's only ever been done like four times.
  466. What is the highest number ever? Has anyone ever had 200? Yeah, as I just said, it's only been done four times
  467. by Wayne Gretzky in the 80s, basically. I can't remember what the highest one is.
  468. I think it's like 220 or something like that. But Mario Lemieux came very close,
  469. I believe. But yeah, basically... 215 by Gretzky in 85, 86. Gretzky's got the top...
  470. Sorry, no, no. Gretzky's got the top four. Back on, Gretzky's the only player to break 200.
  471. Lemieux had 199. And by the way, Gretzky has number six and seven also, if that's worth anything to you.
  472. Oh yeah, no, no. I mean, I lived through it. I lived through every single one of those points.
  473. So yeah, no, I mean, Connor's not going to continue. It would be the story of the millennia, I think,
  474. if he somehow was able to maintain that rate. But still, he's doing things that have not been seen
  475. since Gretzky, basically, in scoring right now. Can you give us a couple... I have a question for you.
  476. But first, one is just a simple fact. I could look it up. Roughly, how old is Connor McDavid?
  477. Are we talking about someone in the prime of his career, growing into the prime,
  478. after the prime? Yeah, I mean, I'm trying to think how old he is now. We could quickly look it up.
  479. But I mean, he's already been in the league for 10 seasons. So I mean, I would say he's...
  480. Early 30s? Late prime. Late prime, okay. So look, I fell in love with statistics.
  481. I told people this story. I heard one of your colleagues, my colleagues, Ed George,
  482. give a talk when I was a 17-year-old about shrinkage estimation. Of course, it was on baseball,
  483. and he was doing, you know, Efron Morris shrinkage estimation back then. And someone's batting
  484. 20 for 50, that's 400. Of course, no one's batted 400 since Ted Williams. Where are you going to
  485. shrink them back to? Talking about Stein estimation, et cetera. All right, so, Shane,
  486. I'm going to put... It's not putting you on the spot. I'm asking you a question.
  487. Conor McDavid right now has played 15 games and has 39 points. So he's averaging 2.6 per game,
  488. which if you multiply that by... I think they now play 82, he'd end up with about 210 points.
  489. Now, you're going to shrink that back. Yeah, I am. You had to give me a point estimate
  490. and a confidence interval right now for Conor McDavid. Where are you going to center your interval?
  491. Is he going to score 160, 150? Where are you putting it? I mean... I'll give you a hint.
  492. He's got 67 games left. Yeah, yeah, yeah. And I mean, he's gotten to... I mean, he's had seasons
  493. where he's gotten to 150. That would represent that, you know, but that's kind of...
  494. Like, if you use his past years as somewhat of a kind of mean to shrink to, you're pulling them more to,
  495. like, you know, 130, 140, I think, there. What would be my prediction? I mean, he is kind of off
  496. to maybe the best season of his career. He just turned 30, by the way. Or he is 30.
  497. So I would say, yeah, I mean, sure. I would center it around 150, 160, somewhere in there,
  498. which would be his highest season so far. As far as a 95% interval, well, I mean, I feel like, you know,
  499. there's at least a 5% chance of, like, an injury soon or something like that. So the lower tail of that interval
  500. has got to kind of go down to basically something like, you know, 50 or something like that.
  501. He's currently at 39. The upper interval, I think, would not probably get to... I would not include 200 in there.
  502. You know, maybe to 180 or something like that. What is his average rate? What would be his mean historical rate
  503. for his last five years or something? For his last five years, we're talking about, like,
  504. maybe 130. So not even... So around one and a half. 140. More like 140. 140, yeah.
  505. So maybe that's about 1.7. Yeah. 1.7. Multiply by 65 games. It's another 112. So he's around 150.
  506. Yeah. Oh, here it is. His entire NHL career is 1.56 points per game. So you multiply that by, you know,
  507. 67 games left. That gives him another 110, 150. Yeah. His mean would be about...
  508. His own mean... Now, of course, you guys know, in a small sample, do you just shrink
  509. towards his personal mean? Do you keep the fact that, you know, his first couple seasons,
  510. he wasn't necessarily the Conor McDavid he is now. So, like, how do you guys... I'm just asking you, Shane...
  511. I'll tell you how I think about it. Okay. Go ahead, Adi, first. I don't... I mean, the bottom line is that
  512. he's... I think he's probably about three SDs. Standard errors. Bigger than his historical mean.
  513. Right? Possibly. In that neighborhood. So I would not... I would think jumping from a 1.6 average...
  514. I mean, he'd have to be extremely lucky going forward for the rest of the season.
  515. So I would not... I would not move his true value up very much. Maybe I might move it from...
  516. Maybe you said it was 1.7. Maybe I move it to 1.8. No, 1.56 is what it's at. 1.56.
  517. So maybe I would move it to 1.7 at most. And then he'd have to be... I mean, he'd have to...
  518. He's certainly going to hold... He isn't going to hold that. But how many do you have to get
  519. to be 200? 2.5 for the season? Oh. Yeah, he needs 131 more points. No, he's going to have to
  520. outperform his... He basically needs two a game. Two. Yeah, he's going to have to
  521. outperform his historical mean by, like, five standard deviations. 200 is not in my end point.
  522. Yeah, I mean, nobody in modern times has come really even anywhere kind of close to the...
  523. I mean, I'm trying to think of... Who the highest, like, career... Like, season points.
  524. That wasn't Gretzky or Lemieux, basically, season points. No, I'm sorry. He needs more than two.
  525. He would need... Sorry. He would need... Like, he needs 161 more points over 60-something games.
  526. He needs, like, 2.4 a game, something much, much higher. Guys, have anything else
  527. caught your eye in sports? I mean, look. I was watching the Sixers, and it's very interesting.
  528. So, they're 7-1, which is great. How's the run? They lost to the... It's great. They lost to the Pistons,
  529. which was not a great loss. They're actually... I was at home. They're playing the Pistons
  530. again tomorrow night. It's early. They seem to be playing well, passing the ball well.
  531. They seem to be spreading the ball around. All of that's going well. Have you updated at all
  532. your belief about the Sixers, or not really? They're going to be a very good regular season team.
  533. They beat in the Celtics. They beat in the Knicks. They beat in some good teams.
  534. I don't know, guys. Any update on the Sixers that you guys were thinking about? I don't think I'd update too much,
  535. because I already had them as a top two or three team in the East. I guess I kind of baked
  536. into my personal forecast about what they would do. I mean, they've looked cooler
  537. and better than I thought. I'm not saying that this is right in line of what I expected,
  538. but I don't think I would necessarily... I think it's going to come down to kind of almost sort of
  539. maybe a three- or four-way coin flip in the end at the East anyway, and I wouldn't necessarily
  540. update too much based on their hot start. Yeah, I think the only question I have is, you know,
  541. we've talked about this there's only one basketball hypothesis. Right now, it's clear
  542. that person one on the Sixers right now is Maxie. Person two is Jalen Brown. And as long as LeBron and Embiid
  543. on the offensive end of the court are three and four, and they're happy with that,
  544. I think the Sixers will be fine. But the minute that they have to go to Embiid, who's not his...
  545. Let's be clear. He can't jump, really. He's not the same athlete he was. He's still 7'2".
  546. He can still score, all of that. But as long as they stay with that pecking order,
  547. some combination of Brown and Maxie and LeBron and Embiid are three and four, I think they're in good shape.
  548. Yeah, I just, I guess, how much, you know... And as long as everybody can be healthy come playoff times,
  549. I think they'll be able to kind of manage that. They're already doing load management.
  550. They're just going to be managing that kind of load throughout the season and staying, keeping...
  551. I mean, they don't have to kind of do as well as they've been doing so far. But I think that will be the challenge
  552. through the season, certainly. Yeah, if I had to predict, I would say there's going to be,
  553. I'm making up a number, no more than 30 games, maybe the most 40 where all five starters play together.
  554. I think they're just, you know, already no back-to-backs for LeBron, no back-to-backs for Embiid.
  555. Like, you're basically getting, with Edgecombe, you're basically getting three of the five starters
  556. in half the games and the other games you'll kind of get all five, which is perfectly fine.
  557. Well, guys, this has been a great Morton Moneyball. I'm glad we got an opportunity
  558. to talk about the election because, you know, I think it's really interesting. I will say one other thing
  559. before we end. This is an interesting week for me in college football, not because I paid
  560. that much attention to college football, but you're almost getting to the point where a few teams
  561. could get eliminated. And here's what I mean. These are two-loss teams that are ranked
  562. in the 10 to 15 range where if they lose, they're now three-loss teams. And it's starting to get difficult
  563. to take a three-loss team. Like, for example, here's some big games this week. These aren't three-loss teams.
  564. BYU No. 7 at Utah No. 3. Big game. Texas No. 10 at Missouri No. 19. If Texas loses that game,
  565. they would have three losses. Alabama No. 9 at LSU No. 15. I'm pretty sure both of those teams have two losses.
  566. If they were to lose, they would go to three-loss teams. Kentucky at Tennessee. Of course, the marquee game
  567. of the week is No. 1 Miami at Notre Dame. That's a huge game. And then, of course,
  568. there's Georgia at Ole Miss. Georgia No. 2, Ole Miss No. 12. Again, these 12, 16, 15, 19, 10, 9,
  569. they're there right now because they're two-loss teams. And we talked about this
  570. a couple weeks ago. If they were to lose, they're three-loss teams. And then you have to start
  571. making an argument this three-loss SEC or Big Ten team deserves it more than this one loss, potentially.
  572. I just think that's going to be a difficult argument to make. But that's just my opinion.
  573. That's why I think the college football season gets interesting to me starting now.
  574. No, and I mean, that's a pretty tremendous slate for this weekend. I feel like NFL,
  575. the NFL slate this weekend, it doesn't look as exciting to me, so it might become
  576. more of a college football weekend coming up. Definitely so. All right, guys. Well, that's been one hour here
  577. of Wharton Moneyball on the Wharton Podcast Network. I'd like to thank my co-hosts,
  578. Adi Weiner and Shane Jensen, some combination of us, and Cade Massey are here every week.
  579. I'd like to thank Jacob Grodnik and the Grod Media team for producing our show. Thanks to Deep Patel,
  580. the big boss, between now and next week. Enjoy your sports, enjoy your election statistics,
  581. and we'll see you next week here on Wharton Moneyball. This has been a Wharton Podcast Network production.
  582. If you enjoyed this episode, be sure to subscribe, rate, share, and leave a review.
  583. Until next time, keep innovating, stay inspired, and join us again.

Brief Summary

Eric Bradlow, Adi Wyner, and Shane Jensen examine correlated election forecast errors, then apply statistical thinking to MLB, NFL, tennis, hockey, basketball, and college football.

Key points

  • Why Election Forecasts Can Miss Together. The hosts explain that election errors may be correlated across races, making dramatic outcomes more plausible than models assuming independent errors suggest.
  • Reading the Early Results. Eric plans to watch the first results from several states and compare them with forecasts—not just to see who wins, but to look for signs of correlated polling error.
  • Prediction Markets Tell a Bigger Story. The hosts discuss how contracts on combinations of House and Senate outcomes can reveal joint and conditional probabilities, offering a way to assess how closely races move together.
  • A Game With No Completed Passes. Shane points out that Tampa Bay became the first team since the 1970s to win without a completed pass, finishing with minus two net passing yards in a monsoon.
  • How long until you know a quarterback is a dud?. The hosts debate how much time teams should give a young quarterback, using Trevor Lawrence and Michael Penix as examples. One rule of thumb: by the end of a third year, teams often feel they know whether they have a fra
  • Can the Brewers’ team approach beat the stars?. The Brewers’ 103-win season and playoff run prompt a look at how a team without a roster full of superstars succeeded. The hosts highlight their collective approach and the unusual fact that every playoff series went to

Episode highlights

Tension

18/ 100Calm

Related episodes