Why NFL Week 1 Can Fool You

Knowledge at Wharton

Ask about Knowledge at Wharton

Episode Description

ABOUT THE EPISODE NFL Week 1 can create instant hype, panic, and major changes to power rankings. But which results actually matter?

Tej Seth, data scientist, explains how NFL analysts separate real trends from early-season noise. He explores why offensive scheme can be more predictive than one game's box score, why turnovers should be downweighted, and how analysts adjust projections after surprising performances.

The conversation also looks at the Ravens' offensive changes, the Chiefs' running game, the 49ers-Rams matchup, and the key tests ahead for NFL contenders.

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. Welcome, welcome to Wharton Moneyball. Welcome to a full hour of sports analytics
  2. here on the Wharton Podcast Network. This is Cade Massey hosting on Tuesday afternoon,
  3. mid-September. Finally feels a little bit like fall because I am in eastern Pennsylvania.
  4. I'm on the campus of the University of Pennsylvania with my three longtime collaborators in person,
  5. in studio, back home where the show belongs in the Huntsman building, the former studio of
  6. a franchise platform. We won't name Shane Jensen sitting here to my right, Audie Weiner straight away, Eric Bradlow
  7. to my left. Gentlemen, afternoon to you. Good to see you in person. No dust. Yeah.
  8. Great to see you, especially in person. Yeah. Always fun. We only do this a couple of times
  9. a year when we all manage to be physically on campus at the same time, and it always feels categorically different.
  10. We're going to go our usual hour or so today. We're going to have a guest in the
  11. first half hour. We're going to talk open lines. The term open lines goes back to this
  12. studio when we actually had open lines and phone calls. Marge from Mississippi used to call.
  13. Instead of Marge today, we have Tej Seth. Tej is joining us to talk a little about NFL.
  14. He's joining us via Zoom here in the studio. He's a, he's a large Tej. He's about 1.5 Tej.
  15. And I think it's more like 2.0. It may be two. He's arguing proportions with me.
  16. He's probably right. I'd saying two. You guys have different angles. We did. Frigging armature involved.
  17. I'm not adjusting properly for my oblique angle. Parallax. We, if you, if you missed the pre
  18. -show and you should be glad you did, you missed a whole lot of prime number jokes.
  19. So there's no telling what's going to happen. We're in person and we're having fun.
  20. All right. So Tej is a football analyst. He is, I think it's safe to say kind of one of the new crop football
  21. analysts. Tej got an early start. He was still, I think at university of Michigan when Eric Eager spotted him or he
  22. spotted Eric and their eyes kind of glimmered and they got a little, little smile and
  23. they've been working together in some form since he went to work for PFF when Eric
  24. was there. Then he's been doing some work for Sumer sports more recently. Provides content for them.
  25. He is an active participant on all things. Twitter analytics is Tej FB analytics is Tej.
  26. Yes. Do I have that right? At Tej FB analytics, great follow in the world of football.
  27. And I think it's Tej's first time. He's part of the Zach mafia, Zach Bradlow, Zach Drapkin, that whole generation of bigger, better,
  28. faster, more. Um, they're doing things that we, this is my first, they were, they were my first
  29. crew. That's right. And we're the first year. They're amazing. Datable winners. And I advised them back then.
  30. They do. They call you Obi-Wan. I, you know, on that comparison, all right, Tej, uh, let's welcome, man.
  31. Good, good to see you. Thanks for being here. Thanks for having me on. I've been listening to this show for the
  32. past couple of years. So I was really excited to come on here for my first time and appreciated Eric
  33. Eager, uh, you know, helping, uh, helping us set this up. Cause I think it'll be really fun.
  34. It was already really fun in the, in the pre-show to talk. I have numbers and other stuff.
  35. So we're glad to have you. Um, we want to hear we, we, we warmed up on NFL last week, but now
  36. we've got actual games to talk about. And of course there's always the risk of overreacting to one Sunday afternoon, or in this
  37. case, Thursday, Wednesday, Thursday, Sunday. We have four days of games, so no, we have more data, right?
  38. Is that, is that what that means? Um, Dej, what, what are you thinking about here?
  39. This is the Tuesday after the first weekend of NFL. What are you thinking about after those 16
  40. games that we saw? Yeah, I'm with you guys where I think that people strongly overreact to week one every
  41. year. I usually like to wait for a couple more weeks of data to come in to start to finalize, uh, some, some early thoughts
  42. on these teams, but I still thought that there was some interesting things that we saw
  43. in NFL week one that might be consistent trends throughout the year. And one of those things that I was
  44. really focused on is knowing how important offensive play callers are to their offenses.
  45. I was really curious to see how some of these new offensive play callers were going
  46. to do, especially in situations where these teams, uh, could be contending for the super bowl.
  47. And the one that jumped out to me was the Ravens new offensive coordinator, Declan Doyle,
  48. who came over from the Bears to call plays for them this year, uh, under coach Jesse Minter was really impressive to me.
  49. The Ravens went under center a lot more than they had in the past. Derrick Henry seemed to be more comfortable running
  50. downhill from, from under center. And then Lamar Jackson also seemed to operate the offense really well.
  51. And so I just kind of thought that I'm going to be more bullish on the Ravens offense the rest of the year because
  52. of what I saw from them week one with the the new offense that, that they put out.
  53. Do you worry about their trolling the Home Depot parking lot for wide receivers right now?
  54. Is that a concern at all? I think it has to be. I really like safe hours. With the Patriots in that case.
  55. Yeah, there are, there are some teams for sure where, uh, where the wide receiver group,
  56. especially with, with some of these like really good quarterbacks, uh, with, with the Chiefs, Bills
  57. and Ravens, where it's like their, their pass catcher groups are just like not where you'd,
  58. you'd want them to be, uh, you know, with, with those, those types of quarterbacks.
  59. Um, but I, I think with the Ravens, like in particular, like Zay Flowers is, is maybe the most like underrated, uh, top receiver
  60. in the league. Like he's, he's someone who consistently ranks very highly in a lot of these production level
  61. stats that, that we like to look at, like yards per outrun. Um, he just happens to have these like
  62. fumbles and primetime games that make people over index on, on, uh, you know, his ability
  63. there. Um, but the rest of the Ravens pass catcher group is not, not inspiring for sure.
  64. So Tez, how do you separate out? Like maybe it was because the Ravens were playing the Colts.
  65. And so maybe the reason why they over indexed on Lamar being under center or Derrick
  66. Henry running downhill, although I don't know why you wouldn't every single play want Derrick Henry
  67. with four yards of momentum heading towards the line, as opposed to handing him from the
  68. shotgun, but this is a separate issue. How do you separate those two things out?
  69. Like maybe next week I'll make it up. They're playing a team with an extraordinarily stout
  70. offensive line, defensive line, and then they're going to play entirely differently.
  71. And that's what a good offensive coordinator would do. It sounds like you're, he's, he's accusing you
  72. of potentially overreacting yourself. No, I'm just wondering how does he separate?
  73. Well, you would do that by the way. You would definitely do that. I've got all kinds of thoughts, but I'm
  74. just asking whether how he separates out the play calling against a given defense with.
  75. One game. Yeah. One game has lots of unique characteristics. Yeah. No, it's a, it's a very fair point
  76. to bring up. Yeah, maybe I, maybe I am overreacting to week one. I, I, when I, when I think about
  77. week one performance in general, you know, some of the like overlying stats, like a team's
  78. EPA per play or their success rate is definitely noisier than I think the schematic principles
  79. that they actually showed in week one, because you just have a lot more control over
  80. the scheme that you're putting out there. And so it's, it's a combination to me of this off season, the Ravens had many
  81. press conferences where they talked about going under center more. So my prior on where, on how much
  82. under center they were going to use this year you know, the percentage of it rose.
  83. And then we saw that actually play out in week one and they used it a lot and it was successful.
  84. And so now I'm feeling decently confident that we'll continue to see that throughout the year.
  85. The, the, the drawback of that is you lose some of the Lamar design rush stuff from not being under, like not being in
  86. shotgun as much as they have been these past couple of years. So maybe there becomes a point in season
  87. where Lamar, Lamar's design rushing threat is just not as strong as it had been in
  88. the past couple of years. And they actually do switch back to their normal shotgun rate of the past couple of
  89. years. Like I could definitely see that happening. Shane was going to jump in here.
  90. I just want to kind of come back to the Zay Flowers comparison you made earlier,
  91. where you sort of, you talked about how, you know, we, he's got very good average
  92. production, but people tend to over weight or over-index prime time turnovers. So is that like an analytical statement where
  93. we are like, like our turnover? Cause I could counter argue that turnovers actually
  94. are maybe underweighted in a lot of analyses. And too much of a focus is on average production.
  95. Is it, was that kind of like our turnovers not to be emphasized relative to say average production because they're less predictable of the
  96. future or they're just less important to the outcome of a game? What's that kind of based on?
  97. I think it's the former, but I'll let Tej answer. They're considered not predictable, even though they are
  98. very important. Yeah, I totally agree. I think they're extremely descriptive. They can often like swing outcomes of games,
  99. but when it comes to Flowers in particular, like any sort of receiver in that mold
  100. where like on a per route basis, they are able to generate yards and be a big function or have a big function of
  101. the offense, but like they've had turnovers go against them because of like either bad fumble
  102. luck or other factors from a predictive standpoint. Like I would still believe in that player
  103. more so than maybe some other people just because of how random the fumble recoveries can
  104. actually be. So Tej, what would be wrong with someone doing the following to not overreact in week
  105. one? I assume one could construct a dataset, understand what is stable across weeks in the NFL,
  106. then score teams on their week one performance, and then adjust their performance on that measure
  107. based on week one performance. Wouldn't that be a reasonable thing to do that would prevent overreaction?
  108. Like, for example, like I'm upset because the Buccaneers gave away in my view, 21 points,
  109. okay, well, but maybe Baker Mayfield's not going to fumble the ball two times next week
  110. and throw one interception because that's not particularly stable, so couldn't one just create stable statistics,
  111. score teams early on with small data, and then do some sort of inference on that?
  112. I definitely think so. I think that when you are looking at something like a metric like expected points added,
  113. for example, and you see how it can capture from a descriptive standpoint, the turnovers that
  114. the Bucs had, and some of the late down performance that the Bengals had in that
  115. game, and if you were to just use the raw EPA values going into next week, you would probably end up overrating the Bengals
  116. and underrating the Bucs. But when you downweight turnovers and you downweight late down performance and things that are less
  117. stable week to week and less predictive of how a unit will do over the course of a season, I think that you can
  118. end up getting a better value on how well the team should actually project going forward
  119. based off of what we saw from them in week one. I think in the framework you guys are
  120. discussing, I think it's worth noting some games, I think might contain absolutely no information actually
  121. moving forward, you know, something like guys, like I think some game outcomes. Some game outcomes.
  122. Well, yeah, maybe there's kind of 120 plays. Yes. But even on kind of the play level,
  123. like, you know, again, did we learn much about the New York Jets when they beat the Tennessee Titans?
  124. You'd like to think that that's a much improved Jet squad, but the Titans looked like
  125. they could be a historically bad team this year. And so do you get information about the
  126. Jets, like if the Titans are that extremely bad, do you gain information about what do
  127. you think? I mean, this is my Jets you're talking about. They won a game. Let's celebrate.
  128. Right. No, I mean, I don't want to take away from the outcome achievement. Is there a hope here?
  129. No, I mean, it can't be, right? No, I mean, so like the thing that you learned about the Jets from you mentioned
  130. like, oh, the Titans could be historically bad. I think you learned by the Jets winning
  131. that game pretty easily is like, okay, the Jets probably won't be historically bad this season.
  132. You know, they could be, but like the chance of them being that has decreased a decent amount based off of that week one
  133. game. Well, Eric gave us the FPI power ranking. Not of all the teams. Only I didn't, I couldn't.
  134. It's okay. Most of them anyway. And, and, and it tracks change, helpfully tracks
  135. changes. Now these are changes in ranks, which aren't very well behaved because ranks at the tails
  136. are much wider than ranks in the middle, but nevertheless, you see some interesting things that
  137. relate to this conversation because what Eric described is simply the way you update your power
  138. rankings after one week of performance. And I, when I saw this chart, I was like, there's a lot of movement here
  139. for one game. Now I realize, you know, you come into the season, you've got priors, you know, the
  140. least you'll never know less about a team than when watching before watching week one.
  141. And so of course you expect some movement. Can we, can we debate that just for one second?
  142. So, um, I disagree with that premise and you're Bayesian. You added information.
  143. You're going to put the, can you contribute information? I'm going to tell you. Right.
  144. So, right. Yeah. So again, let's, let's take an example of the chargers. Okay. And I think going into the game, we
  145. had a strong belief that the chargers were a good team, but there was a small probability maybe that they weren't particularly good.
  146. Okay. Now we observe the data from the game. Now I would make the following argument.
  147. There's still a probability. They're a good team and still a probability. There's a bad team, but probably more probability
  148. is on the bad team. That mixture can have higher variance than the prior than the original distribution did.
  149. I would call that as with statisticians, a form of negative information. And so we have a wider understanding.
  150. There's more variance after the game. So from an information perspective, it can be
  151. lower. And matter of fact, I have a paper on that exact topic. And a cautionary tale.
  152. I think the chargers are a great example of a team where you should never have too strong of a positive prior going into
  153. the season. I think they, for a while, for about 25 years in a row, we have had a, you know, their, their actual in season
  154. performance has probably underestimate under either way, I just disagree with the premise that more information
  155. leads to a narrower posterior, but we have to recognize an unusual event occurred by the,
  156. according to the prior, the prior said, they're unlikely to be bad. They were terrible.
  157. So I was a good statistician. So what the math shows is prior again, I will emphasize the degree of discrepancy between
  158. the observed value, the performance and the prior exactly determines whether this phenomenon can occur.
  159. And if it does occur to the degree to which it occurs, of course. Okay. This is, this is helpful to me.
  160. I mean, I, I, I'm not the Bayesian statistician y'all are, but I, I do do some Bayesians at least in for our
  161. listeners that want 10 seconds. This is only one thing you have to assume. Notice.
  162. I was very specific. I said, the prior was a mixture. If you assume a unimodal distribution, like a
  163. normal thing, like team's performance is unimodal. You can never get this effect.
  164. And what you said is mathematically true. And that's why I do not believe in unimodal distributions.
  165. You have to fit things with mixture priors. Otherwise you get this wrong set that information
  166. always goes down, which it does not have to. I did note that you said in your first statement, you said mixed and something like,
  167. okay, this is the secret to it. So just say, give us one more line, give it, give the lay audience one more
  168. line on what you mean by mixed model or mixed, mixed distribution. So I'll use an example from my home
  169. field of marketing. I always, this is how I describe it to someone when I'm just at a party
  170. and they say, what are you working on? That's interesting. Here's what I say to them.
  171. No, they do. I, so what I say to them, I said, I'll pick Shane. Shane, what's your favorite restaurant here in Philadelphia?
  172. Oh, that's a tough one. Um, Tres Jalapenos. Okay. I don't know that, but I'm going to
  173. try to go there. You like that restaurant a lot. So if on the X axis, I had how much you liked the rest or the
  174. quality of the restaurant, you agree that your distribution is probably far to the right.
  175. Cause you said it's your favorite, but are you willing to admit that there's a small
  176. probability that maybe it's not of high quality? It could be 1%, some small probability.
  177. Okay. Oh yeah. Shane, wait, Shane then has an experience at the restaurant. He watches the outcome of the game.
  178. He now has a negative experience. That's what a mixture is. I know, but help me understand what you
  179. did with the prior there. Cause normally you would expect, look, you've got, as you said, a unimodal distribution, but you've
  180. got support over the full range. You've got mass over the whole support. I know, but you must have, you must
  181. have a mixture. I, you must have an underlying state. So it's one of that. Yeah.
  182. You could think that there's some matter of fact, the way you could describe it in
  183. math, but it's also in English is I think the way science works is you have competing hypotheses.
  184. One hypothesis, Shane's restaurant is good. Another hypothesis is it's not as good.
  185. And you've got distribution, but you give quality distributions for both of those states.
  186. That is correct. And now you have an underlying, before observing the going to the restaurant or observing the
  187. chargers, I've got some probability we're in each of those two states. Then I observed data and now a couple
  188. of things could happen. One is what's called utility balance, which means the closer you get to 50, 50, maybe
  189. he was 95, five to start. Now he has a bad experience. Now he's 50, 50. That will increase the variance.
  190. The other possibility, there's three ways. Second is the humps move farther apart.
  191. The third of course is each hump gets wider. That's another possibility. Those are the only three ways you can
  192. get increased variance. And there's a nice behavioral scientific story for each of them.
  193. And it relates to football too. Fine, fine, fine. And we've done a lot of this. I think it's, I think it's been educational,
  194. been educational for me. I'm going to do one more line on it though, because, and then we're going to
  195. ask Tej, Tej, we're going to ask you, is this relevant to football modeling? Have you played with these kinds of things?
  196. Are you buying what Eric is selling? So I want to ask, but one more, just for clarification, this is quite the simplification
  197. of the world in most situations, because you're saying there are two discrete possible stages of
  198. states of the world, and that alone is a, as a massive, strong assumption that drives
  199. a lot. In this case, it's not necessarily too underline discrete. I mean, that's kind of how you can
  200. model it, but like, you're really thinking about just two different, a distribution that consists of
  201. two different modes, right. With variation around them, like two components, like
  202. two components or more. It could be an eight. It could be eight. There is, you are kind of be more
  203. than one. Yeah, yeah, yeah. You are thinking about some underlying latent discrete
  204. state, whether that maps to like a real life thing or whether it's just a construct.
  205. It could just be a construct that helps get to something, or it could be a theory of the world, one of those things.
  206. I agree with that. Pull us back into football, please. No, that was an awesome conversation.
  207. It felt like I was back in college and I got a little bit hungry talking about the restaurants there too, but I think
  208. it's a really good point to bring up. There are certain times in football where you
  209. get these situations where more variance gets added to your long-term outcome of the team
  210. for the rest of the season. The Chargers was like a great example, I thought, to bring it up there, because I
  211. feel less certain about the Chargers, or about my projection of the Chargers for the rest
  212. of the season. And then the Giants in the other direction is also in that category, where they go
  213. out on Sunday Night Football, and they look really good against the Cowboys. And coming into the season, I had thought
  214. that the Giants were a team where they had some pieces in place, but you kind of felt that they were going to be
  215. a below average NFL team, because the secondary wasn't all the way there, and the offense
  216. had the lack of weapons outside of Malik Nabors, and Nabors was coming back from his
  217. injury. But then I watched that game on Sunday Night, and I'm like, they do kind of
  218. have this right tail outcome all of a sudden that has appeared, where they could become
  219. a pretty good football team and relevant in the NFC. But now I'm less certain about how I
  220. feel about the Giants for the rest of the year because of that game as well. Maybe this is a great illustration.
  221. If we could formulate this underlying latent state, a couple of key injuries like they experienced
  222. last year could move them from, you're sort of like, we've now got them in this elevated part of the projected outcome distribution, and
  223. now all of a sudden a couple of key injuries, and they snap into a whole nother part of that kind of probability space.
  224. I want to suggest, I'm curious whether suggesting that model to the layperson would actually improve
  225. their judgment. I'm going to hypothesize, as a behavioral person here, I'm going to hypothesize that that would
  226. be a helpful heuristic. People aren't great Bayesian updaters. They hold generally too much confidence about their
  227. belief in the state of the world. I wonder what would happen if you coached them to carry multiple states of the world?
  228. Like this mixed distribution should be really helpful to get away from being too sure of
  229. the state of the world. So I've run lots of experiments, behavioral experiments in the lab and in the field.
  230. I've never actually asked people like, does it help people better calibrate their uncertainty judgments by
  231. positing to them, you present them this idea of two states of the world, then ask
  232. them to make, draw confidence intervals or whatever. That's, we're going to do that and we're
  233. going to put it in our paper. It's a great idea. It's taken us 12 and a half years
  234. to come up with the collaboration. But there we go. That's a great idea for a study to
  235. run. Tej, which was the most impressive win of the weekend and why was it San Francisco?
  236. That was wildly impressive. There is this funny thing, and I don't know if you guys have talked about this
  237. in tennis before. I think it's really relevant in tennis whenever Sinner and Alcarez play, where it's like Sinner
  238. is usually the better player against every other opponent. But Alcarez has this advantage against Sinner when
  239. they go head to head, it seems like. It's very tough to actually model that out.
  240. I think the 49ers Rams have that sort of similar principle, where against most opponents over
  241. these past couple of years, the Rams have been the better team when healthy versus the
  242. 49ers. But when the 49ers and Rams play against each other, the 49ers and Kyle Shanahan seem
  243. to just have an advantage against Sean McVay. And it's extremely difficult to actually model that
  244. out when you look at the data. But last year, you had a Thursday night game where Kyle Shanahan took Mac Jones and
  245. all these backup and practice squad receivers and just went up and down the field on
  246. the Rams the entire game and beat them. You had this game in Australia where the
  247. Rams were three and a half point favorites pregame, and the 49ers made it look like
  248. they were favored by 14 points going into the game with how they played. It's just wildly impressive how Shanahan seems to
  249. keep evolving his offense and how they're able to game plan their defense to stop this
  250. Rams offense that does very well against every other team. What I loved about this game is all
  251. the jet lag analytics that have come out. Yeah, right. And right. I mean, like, how have we not?
  252. This is like this is something that's relevant to like almost everybody, anybody who flies.
  253. Like, how do we not know more about how people recover? Worse than that, we actually have some of
  254. the foremost experts in rhythms and jet lag here at Penn. We've never studied sleep.
  255. You've never studied jet lag. They're reluctant to come. We had one sleep expert once.
  256. I heard a podcast before the game last week. Someone give me a minute. I'll remember.
  257. And they had a sleep expert on and that person had a very strong opinion that the right thing to do was to go
  258. in last minute and not try to. That's right. Yeah, that was Ed Fang. Ed Fang's podcast.
  259. Oh, yeah. Yeah. So he brought someone. We've had Ed, Owen. Yeah. Yeah. From UCLA.
  260. I listened to that cause I worked. Maybe Michigan cause he's in Michigan. Cause I, yeah, no.
  261. I thought Ed's person was in Michigan. No, I thought it was UCLA. Ed lives in Michigan, but he brought someone
  262. on from UCLA. UCLA. There are a handful of sleep experts. We do have one here who I've been
  263. riding bikes to work with. A handful? There must be hundreds. Well, no, who actually expertise on circadian rhythms.
  264. That's the, that's the very specific. But the thing that I listened to that is that there's two competing things and we
  265. know that circadian rhythm matters. And that what the Rams are trying to do is sort of line up their, their
  266. preparedness to their natural body's rhythm. And that the San Francisco 49ers are going
  267. to come really early and try to line up to, to the new rhythm. And the research suggested that what the Rams
  268. was doing. Right. But there's another piece of it as well, which is the flight, which is horrendous.
  269. And being in your own bed has a huge amount of value. And, and, and, and that flight is a,
  270. it has another effect size. And what they know is that the circadian rhythm effect size is something.
  271. And the flight is something, but they don't know how to. There's got to be like an age component
  272. to the, I, for my own personal experience, there's an age component to that, like circadian
  273. recovery or whatever. And I mean, well, not Stafford of all people, like that guy should have been over
  274. there preparing like a week early. A part of, apparently the Rams sent like staff members on this journey multiple times to
  275. sort of check it out. Get out as if somehow that's the same thing as a staff members played a football
  276. game against the 49ers the next day. What an experiment. How are you feeling? And I remember, I remember I didn't once
  277. go to Australia for a conference. And I slept wonderfully on the plane. It was a 14, 15 hour flight.
  278. So I got seven or eight hours and I got there and I felt like crap. And I'm thinking of the rain.
  279. I just think to myself, this is idiotic. Well, by the way, sleep analytics, we, we,
  280. we occasionally go adjacent fields and do some analytics of something else like fire or whatever,
  281. or leaves. We did leaves one time, this time of year, traffic, traffic. So we need to get a sleep analytics
  282. person. Eric's been trying to get it. No, no. I was just going to ask. Is the, I, if I had thought about
  283. which was the most impressive game this week by a team, I watched last night's game
  284. last time. I checked weren't the Broncos like one bone Nick's injury away. They claim from going to the Superbowl last
  285. year. Oh yeah. No, I mean, I said claim, but then they got beaten by three touchdowns last night.
  286. Yeah. And so I don't know, to me, from what I expected to happen in the game, that one to me is at least as
  287. far away from expectation as the range, impressingly the chiefs look like the best team.
  288. And Oh God, that is, that is the best performance. Well, hold on. No, let's get the chiefs analysis since Eric
  289. put it on the table. And I would like real quickly before we go to best. Um, what can you talk about that game?
  290. And, and I did, I wasn't able to see it, but the bits I saw floating out from it, where they're running all over
  291. them and there's like three tight end sets and all this stuff, and that was some
  292. big surprise. So I'd like to hear about that as well. I did think that the chiefs were really
  293. impressive yesterday, but it wasn't for the reasons that you would normally expect from the chiefs,
  294. which I think might make it more scary for the rest of the league where it was the chiefs run game and defense that
  295. really dominated the game. And in Patrick Mahomes, his first game back from his injury, the past game wasn't impressive
  296. at all. And, but you had, you had this chiefs run game where they went out and they
  297. signed Kenneth Walker this off season because their running backs were the least explosive rushers in
  298. the league last year. And we talked about running back value all the time. And, you know, it doesn't matter as much
  299. as, as most other positions on the field. But I do think that when you look at which component the running back actually controls
  300. of their rush, it's the explosive rush ability. It's like, how well can you take what
  301. was blocked for an eight yard rush and turn it into a 20 yard rush? And Kenneth Walker appears to be really good
  302. at that. The chiefs running backs last year were not good at that. And so they, they had, they had this
  303. mentality to go out and find someone who could create explosive rushes for them.
  304. And so far in week one, that's paid off for them. We'll see if that continues throughout the rest
  305. of the year, because the Dolphins, or sorry, the Broncos had one of the worst tackling
  306. performances I've ever seen in a, in a football game yesterday. Uh, but it's like, it's definitely something to
  307. monitor for that. Mahomes still has retained his ability to be like an inch beyond the grasp of the
  308. nearest defender. Just, it's, it's magical. He like is, he's, he doesn't look like
  309. he's moving that fast, but he is just fast enough, just fast enough to get the job done.
  310. It's incredibly impressive. Okay. Adi, let's end with this question. Well, maybe not end, but close.
  311. We have to wrap soon. Adi had a question for Tesh. Well, I wanted to know what you, you
  312. got fed the best game by, by Cade. What would you really think was the most impressive victory in the first, in the first
  313. week? I thought it was the, the ones that we talked about, the 49ers victory was, was
  314. really impressive. Um, the, the, the Bills beating the Texans, I thought was also something that was notable
  315. where Josh Allen just like was incredible. Like this is a, this is a Texans defense that was expected coming into the year
  316. to be top three defense. They have this, uh, elite duo of, of pass rushers and Allen kind of just dealt
  317. with the pressure that was put on him, uh, by those pass rushers throughout the game
  318. was constantly pushing the ball downfield and, and, uh, successfully. And now we get to see him again
  319. on Thursday night, uh, which I think is, is really exciting because it's always fun when,
  320. when someone of that caliber plays in these, these primetime games, um, like it was, it
  321. was a, it was a very impressive performance, um, by, by the Bills offense. For what it's worth, Cade, uh, I asked,
  322. uh, chat GPT to just rank order the games by absolute spread deviation. Like how far the outcome was.
  323. You are correct. This week, week one 49ers Rams was the most deviant game. It was 23 and a half because it
  324. was San Francisco one by 20, but Rams were favored by three and a half. So it's 23 and a half Cardinals chargers
  325. was then 21 and a half. Cause it was a 12 point difference, but a nine and a half point spread bears.
  326. Panthers was 19. I mean, this, this scored so many points. And then Broncos chief was fourth, but you
  327. did pick the game that least has the largest deviation from the spread. It was obvious sitting there and that was,
  328. it was like a lifetime ago, but you were sitting there and just watching things happen
  329. that weren't, especially after all the hype, the preseason hype for the Rams, y'all, y
  330. 'all, y'all heard me on the text or that was taking a lot of satisfaction. And I'm never at an honors fan in
  331. my life, but I was really enjoying the hype being undermined. Okay. Last question going, we'll, we'll keep it pretty
  332. close. We're trying to, you know, pay attention to games. We have a slate coming up week two
  333. slate. It's just Thursday and Sunday. Well, Monday as well, Thursday, Sunday, Monday.
  334. But let me give you six games, pick one or two and make us smarter viewers of these games.
  335. Give us something to pay attention to, or you think is interesting or a question you
  336. have. These seem like the top games of the week. You already mentioned Detroit at Buffalo Thursday night.
  337. The early slots on Sunday, Minnesota, Chicago, Cincy at Houston. Houston gets another test at home against a
  338. good AFC team. And then three in the four o'clock slot, Jacksonville at Denver. Denver gets another run.
  339. Jacksonville looks super impressive. Jacksonville went flying up the FBI power rankings,
  340. just shocked like 12 slots or some crazy thing. Washington at Dallas, the two NFC East losers
  341. from last weekend. Somebody's going to come out of there, not Owen too. And then Seattle at Arizona.
  342. So which of those six or make us smarter about one or two of those games. I think the Bears Vikings game that you
  343. mentioned there is fascinating because you have this defensive coordinator and Brian Flores, who was just
  344. blitzing Jordan, he blitzed Jordan Love on 75 % of Love's dropbacks on Sunday. The really interesting thing about this game is
  345. when you look at how Kayla Williams did against the Blitz last season, he was very
  346. impressive. He ranked third in the league in EPA per dropback versus the Blitz last season.
  347. He basically never got sacked versus the Blitz last year, which, you know, is like the
  348. main function of why you send the Blitz, which was really good. And then Ben Johnson offenses, it's like generally
  349. done well when defenses send more than four pass rushers. And, you know, I think that the combination
  350. of Ben Johnson's scheme and what we saw from Kayla Williams last year versus the Blitz
  351. plus week one, Kayla Williams this year, where it looks like he maybe even took another
  352. step up in his game, makes that a super interesting test of going up against this
  353. very aggressive defense that we're going to see from the Vikings. Super helpful.
  354. Super helpful. How short are you on Denver after that showing last night and how up are you
  355. on Jacksonville? That's what intrigues me about that game. Denver's a two and a half point favorite
  356. at home. Jacksonville, Denver. Yeah, I think the FPI move is a bit strong there.
  357. Jacksonville, second half of last year into what we saw from them, week one has had
  358. a very good passing game basically ever since they trained for Jacoby Myers and Parker Washington
  359. has been incredible for them. And Brian Thomas Jr. bounced back a little bit after the Jacoby Myers trade as well.
  360. But I still am cautiously optimistic about the Broncos. I thought that it was a weird week
  361. one game where they missed a lot of tackles on defense and their offensive line broke
  362. down in pass protection. I still think that their roster on paper, from what you thought about it preseason, like
  363. looks pretty good. And this will just be a very interesting game to actually see if they're going to
  364. play closer to their preseason prior or what we saw from them in week one this year.
  365. Right. It's kind of a, it's kind of a great people are overreacting to week one game
  366. because you probably got the public maybe too strong on Jacksonville, too short on Denver.
  367. Yeah. If you want to play the overreaction card, that would be the place to go. Okay, Tez, we'll let you go.
  368. Thank you for the time. Love the work that you're doing. Come back and join us again.
  369. And again, folks, you can catch him. The easy way is Twitter at Tez FB Analytics.
  370. Tez FB Analytics. Great follow up there. Thanks, man. Appreciate it. Thank you guys for having me.
  371. It was really fun to come on and to take a break from cloud coding and talk to you guys.
  372. That's great. That has been the first half of Wharton Money Bowl. Come back and join us after the break.
  373. Welcome back. Welcome back to Wharton Money Bowl. Welcome to the second half of this week's
  374. show on Wharton, the Wharton Podcast Network. This is Cade Massey hosting with the whole
  375. crew, Audie Weiner, Eric Bradlow, Shane Jensen. We are sitting in person in the Wharton
  376. Podcast Network studio, Huntsman Hall, Locust Walk, West Philadelphia. Been a while since we've been in here.
  377. We only get in here in person, same time, same place a couple of times a year. Always fun.
  378. Just off the Zoom lines with Tez Seth talking NFL. Good fun. Good fun. Open lines for the next half hour or
  379. so. Guys, we have lots we could talk about. We need to stop before we leave football.
  380. Stop on the way. With college football, week two in the books now, we're still kind of in that preseason
  381. stage. College, of course, doesn't have official preseason, but early weeks, especially these days, are pretty thin
  382. of big games. But y'all know there was a big one. There were a couple of great games, a
  383. couple of big outcomes anyway this past weekend, but Texas-Ohio State, game of the weekend,
  384. one of the games of the year going in, and then it lived up to its billing. I was fortunate.
  385. I was there. I wasn't going to let Ohio State, the number one team in the country, come to
  386. town and not go. And I was definitely questioning my life choices at halftime. Is it fair to say right now that
  387. from the college football playoff seeding point of view, that Texas now must, I don't want
  388. to say must, any reasonable person, they have a two game lead over Ohio State? As a starting place, maybe.
  389. As a start. All right. Now, if that is true, then the game has some serious meaning, especially given the structure,
  390. which is four teams get a bye. And that means Texas has a much better chance now than Ohio State of getting a
  391. bye. And I think they also get to pick where the game is. There's other benefits of being the higher seeded
  392. team. Was the game more important to Texas than Ohio State by that rationale? Because they have a greater competition, division, and
  393. everything. They were weirdly in kind of a similar situation. They have two of the, they might have
  394. the two toughest schedules in the country this year. They play this partly because they play each
  395. other out of conference and nobody else is playing big games out of conference. But they happen to both have really tough
  396. in-conference schedules. And so they have kind of the, the, the consequences are kind of symmetrical.
  397. Can I ask a question? I did not watch the game. It was Rosh Hashanah. So no chance for leaving.
  398. Shy that excuse, shy that excuse. I would be giving you. So obviously we know that Texas won in
  399. dramatic fashion, coming back with 21 points in the, in the fourth quarter. But what are the underlying metrics say?
  400. Which was the better team? They surprisingly close. It was a tale of two halves fully.
  401. I mean, the Texas Ohio State on the first and Texas on the second. If you look at, I saw Bill, our
  402. friend, Bill Connolly publishes, did, did I get, did we get beat that bad metric after
  403. every college football weekend? And it just shows the difference. And in this case, it's like success rate.
  404. So, you know, every play can be coded as successful. Like, did you advance the expected points of
  405. the drive or something? And so each team gets an success rate and most teams that win have a higher
  406. success rate than the team they beat, but not everybody. Reliably, there are some, and sometimes it can
  407. be kind of a big difference in the other direction. So it's always interesting to look at that.
  408. And I was interested to see what it said. And Texas, it wasn't quite 50, 50. It was, it was North of 50, 50
  409. for Texas. So when by that metric, Texas was ahead. So let me ask you another question.
  410. Just, it's an empirical question. Maybe there's some math one could do on this. Let's imagine, I tell you that two top
  411. teams play each other, Texas, Ohio State. Okay. The game essentially ends up equal.
  412. I mean, one team wins 24, 23. But you could win the first half by 21 or the second half by 21.
  413. Do you care? Like what I'm trying to get at here is maybe I'm not, I'm not going to
  414. use the word momentum, like they have momentum going into the next game. No, no, no.
  415. I'm asking is one half performance more diagnostic of performance than the other.
  416. I know what I would rather be Texas right now. Forget just that they won the game.
  417. Could you, when it mattered, when they had to perform, they way outperformed Ohio State.
  418. That's got to be worth something. You got, you could look at overtime performance
  419. and see if second half versus for like, like in these kind of tell two, which is more predictive, which is more, that is
  420. a momentum argument, but why, but it's like just a regime shift. So injuries would go that way, for example,
  421. right. It's still, it still makes the second half, but there is enough data in Eric's framework,
  422. which is, and I have never thought of it before. It's really interesting. What if you split halves and just ask
  423. which half was more diagnostic out of sample? Yeah. I'm going to guess it's first half because
  424. so much garbage time accumulates in the second half. So that, because people know the, one hypothesis,
  425. notice I conditioned on it being a close game, but you're right. No, no, no. This is so that, so this is, well,
  426. you could condition on it being close game. It's gotta be second half. Well, I mean, I agree.
  427. He's got a very strong argument where the unconditional correlation probably means first half.
  428. But I think, I think there's this one structural difference about the two halves is there's,
  429. there's plays are called on scripts typically in the beginning of a game. And so there's a, there's a kind of
  430. a famous phenomenon that happens where a team does really well first drive, or maybe even
  431. first two drives where they're on script. It's like they knew what place to call.
  432. And then it's like, okay, now let's see how you do. Once you have to adjust what you've revealed
  433. your, yeah. And they, they, the other team adjust to you and can you adjust to their adjustments?
  434. And so it's structurally a little bit of a different game in the second. How about, how about fatigue?
  435. Cause this is a big deal. So one of the things that the insiders in basketball explained to me was when I
  436. asked who won the, the, the championship was the Knicks that did well, or the Spurs
  437. that collapsed and the dialogue was the Knicks did well, but insider said the Spurs, they
  438. got exhausted. They over overplayed, they came out too hard and they just got too tired.
  439. Didn't they have a tougher champion conference championship? They did. Oh yeah. The Knicks ran, you know, ran roughshod.
  440. They had to beat the Thunder. So they were tired and they were, and they were fatigued and that, and that, that
  441. somehow has to, has to play out because the second half you're tired. There's no question.
  442. And this is where Texas had an advantage because it was in Austin and it was hot.
  443. And so here comes Ohio state from, you know, the North and, but it's not Texas. And it's been, it's been a brutal, it's
  444. been a legit, tough August, September, and these guys, as long as they're working outside, they're
  445. going to be better conditioned than folks from Ohio and the place in the game. You'd expect that to matter is fourth.
  446. And those last three drives, I mean, this is one of the stories of the game is that those last three dives, Texas was
  447. moving them around in a way they weren't before I go ahead and then I have one thing I was saying, I'm confident that
  448. someone who builds models for predictive models for outcomes has looked at first and second half
  449. performance. And again, we're always on X or Twitter or wherever you want to call it at
  450. W money ball. If someone who's listening to our podcast wants to tweet at us what they have found,
  451. uh, that would be interesting to look at. Cause look, if I were someone that bet
  452. on sports for a professional living, trust me, it wouldn't take a lot of thinking to
  453. say, I wonder if first and second half performance does better out of sample. Like that, like it's, it's on the master
  454. list of things I would have looked at. How about that? Well, I don't know. I mean, we've, we don't, we've talked about
  455. this. We've I've heard you come up with ideas for 12 and a half years on the show and that's the first time you've said
  456. that. So sometimes there's, I mean, there's a lot to do over there. So sometimes they emerge slowly.
  457. Um, let me try one other thing on, on with you guys and y'all can shout me off on this.
  458. Um, Manning, everybody talks about Manning, you know, of course, and arch arch. And, um, it, I, as a, as just,
  459. I realized that I'm not a sophisticated football. I'm like a scout, you know, I'm, I'm
  460. just. I'd lifetime fan, but how about the following hypothesis for quarterbacks?
  461. Some quarterbacks have this thing where they're competitive enough or they're good enough when competition gets
  462. tough, that they, they have, they, there's a difference in the way quarterbacks react in the,
  463. in this is clutchiness. I suppose I'm talking, I guess I'm not about quarterback clutchiness and I'm asking whether it's
  464. a thing. And if it's a thing, would, how much just. Sheer quarterback reliability in normal circumstances.
  465. Would you give up for this quality that some quarterbacks seem to have, where they are
  466. gonna, they, especially maybe in college where there's a little bit more influence, a single player
  467. can have. So I'm talking about a lifetime of watching quarterbacks like, I mean, Sam Ellinger had this,
  468. he wasn't a very lauded quarterback, but he had this quality where like, man, I want
  469. that. If we're in a tough spot and need to drive and we want to touch down, I, I have a lot of confidence.
  470. I can get it out from the pros to my like this Baker Mayfield, Baker Mayfield's mean performance is pretty good, but down one
  471. score near the end of the game. I think most people, both true at Oklahoma, when he played there too, most people would
  472. take Baker Mayfield near the end of the game, more so than his, I'll call it his performance overall statistics would show.
  473. It's funny you say that because Sam and Baker played each other in one game. Sam's first game might've been Baker's last and
  474. Sam did really well under pressure and then Baker won the game. Okay. Okay. But so, so, so like, here's an example.
  475. Um, um, Chris Sims, when Chris Sims was at Texas, the other quarterback was major Applewhite.
  476. And major Applewhite didn't have the arm and the skills and the size that Sims did.
  477. And when Sims performed, he's like, you saw all that talent. And he, but the thing was kind of
  478. regardless of game situation, he looked the same and his, his completion percentage is the same.
  479. It's like, he's not moving the team any differently. You know, it's just kind of, you get
  480. this thing and major Applewhite, on the other hand, it's like this scrappy guy, but under
  481. pressure in the moment. He seemed to perform and just in Manning, if there's any truth at all to this,
  482. the Manning profile is the, the clutchy guy where, and he's already got like three top
  483. 10 wins and 14 starts or 15 starts or whatever it is, it's like he's already hitting numbers, the great Texas quarterbacks hit and
  484. he's going to fly past them if they have a good season. And I'm just wondering this, look, y'all
  485. shot me off of this, please. Like, is this a thing? Sorry, go ahead Shane. Just my reaction is kind of like, it's
  486. almost like the opposite. I think of how you think of Peyton, right? Peyton was an incredibly consistent, like performer, but
  487. like, you know, at least early on in his career, in the playoff, like, like in the big games, you kind of like almost
  488. saw a drop off, right? I don't know. Yeah. I mean, certainly his reputation in the early
  489. years. Was that what I was going to choke? What I was going to say is let's say what Kate is saying is true.
  490. Then I'm thinking now of the implication for team construction. So here's what you should do.
  491. You should have a high mean quarterback and a high variance quarterback. No, no.
  492. If you can move them in and out, I just, that was my quarterback sets. No, no. Is that what happens is you obviously play
  493. the high mean person. You hope that your way ahead, whatever, but if you're not, you then bring in the
  494. high variant person, because it doesn't matter whether you lose by 10 or you lose by
  495. 50, but you may come back and win, but you've just, I'm usually the one busting you for these bullshit psychological stories.
  496. And in this case, you're busting me because you just turned my clutchiness into high variance.
  497. So you're giving a parsimonious statistical model for my variance. That's not what I'm doing.
  498. But you're saying not high variance. You're saying, you're saying raised in the mean.
  499. Yeah. So, I mean, that's an obvious thing. If you have a particular quarterback who made,
  500. I mean, in football, high variance might be a particular way of playing maybe long passes,
  501. right? Call it the Jameis Winston. Right. So you're basically saying you have a guy
  502. who's good at get an EPA per play. That's fine. Another guy who's much lower on EPA per
  503. play, but high variance. And so you bring them. I love the, Eric, you don't ever hear
  504. that talk. Can we talk about take more chances, dial it up. I was actually quite serious that teams should
  505. have then two quarterbacks. That's not what I was talking about at all, but I like it as a model.
  506. Employees. James Winston is a backup. I don't think you need that because I don't think you need that.
  507. Because similar to basketball, when you're when you're down a situation, you employ a strategy that's
  508. high variance. So I think the the quarterback who has one way of playing can probably also play
  509. a different way. I don't know. Maybe the best way. Unless that's not true. Yeah, I think that's that.
  510. Maybe that's kind of another way of stating your. Well, another way that frame players do not
  511. play the same when it's. Yeah, I mean, you shouldn't. If you I mean, if you're a basketball
  512. team, you're down by 15. You better start shooting threes. Whether you're good at it or not, you
  513. do it. Yeah, yeah, yeah. And similarly, certainly Brady was a person who like only had like he had the high
  514. mean. But like, I don't think if you asked him to play high variance, he would be
  515. his natural role, certainly, especially with quarterbacks. You want you want gunslingers, right?
  516. So you can't be asking the asking the conservative guy to be the gunslinger may not
  517. be the best strategy. No, I was only going to say that another way to interpret what you're saying, other
  518. than high variance, would be to say the distribution of the player's ability conditional on it
  519. being either a high leverage, high intensity situation is varying. And so you're saying you're you're modeling against
  520. a unidimensional ability model where the person behaves like this ability under all situations.
  521. And so and by the way, I don't disagree with that as a premise. I, in general, since I worked at ETS
  522. for two years, I, in general, do not believe in single ability models in general.
  523. They don't, despite the fact that's what the S.A.T. and these standardized tests purport
  524. to measure. I just do not believe in them. So I do believe that there is something
  525. I could condition on that someone's ability would vary. That could some if you told somebody in
  526. English it could represent clutch ability in some way. I do believe that. I mean, that's the thing that we don't.
  527. I mean, yes, I didn't weigh in, but my you were rubbing your eyes and avoiding eye contact.
  528. My priors would be that these things aren't real. Yeah. But but the ability thing is real.
  529. I mean, there are definitely quarterbacks who have different styles and one style is high variance.
  530. And that's that's I would say I might be real. These things are not measurable at the individual
  531. level. That's the S.A.T. Is that what you mean by real stock? That's what I mean.
  532. Yeah. Obviously, something is real. The question is, is detectable and measurable.
  533. Yeah. Yeah. You're not saying it couldn't be real. You're just saying it's not measurable.
  534. It's not measurable. And that's another way. Probably not real. It's another way of saying it's probably small.
  535. That's the other thing. People miss that when someone says we haven't found it. There's a part of that that is saying
  536. if it's Israel, it ain't big because otherwise you'd have found it. Yeah. Yeah. All right.
  537. Let's talk tennis because we're wrapped up on Sunday. That's what's there of. There's a count.
  538. He had an easy draw. Is that fair? This is me not watching. Well, we're starting to get into that gray
  539. area because we can't. Unlike the French where he did win. But remember, Algaraz didn't play and Sinner melted
  540. in two sets the love five one and then went into like heat stroke and got eliminated in like round two.
  541. Algaraz did play in this tournament. Now, I understand he was coming back from injury.
  542. He lost to Shelton. Shelton beat him in five sets. So, you know, Djokovic also played in the
  543. tournament. Now he was eliminated early in the tournament. At some point, you have to give the
  544. managed. Well, Algaraz wasn't the real Algaraz. Yeah, I mean, you can only play the
  545. players. They are not going to show up on your bracket. I've not done an analysis of Djokovic's 24
  546. majors. Nadal's 22, Federer's 20. Look, Federer's stocked up seven or eight majors
  547. before Nadal and Djokovic even started playing. So this conditional argument is a tough one.
  548. Zverev has two majors now. He has two. He also made the final of Wimbledon and lost to Sinner in a tight match.
  549. He also made the semi-finals of the Australians. Let's go over his year. Semi-finals Australian, won the French, finals of
  550. Wimbledon and won the US Open. That's a good year. That's not bad. And so that's an incredible year.
  551. But you know, they asked him the question that Cade just posited. It was a great interview.
  552. They asked him, are you the best player in tennis now? And he answered very honestly.
  553. He said, yes, right now I'm the best. Cause Sinner's injured and Algaraz is coming back
  554. from injury. But if you were asking me if those two are fully healthy. No, he didn't say he can't win.
  555. He just said, let's be real about it. He goes, no, I'm not the best. And by the way, I started to think
  556. about his career. So he went 0-3 to start his career in majors, major finals. He's now 2-3.
  557. The greatest example of this, the greatest example, someone I followed a lot when I was
  558. younger, it was in the McEnroe era, Ivan Lendl was 0-4, lost his first four major finals.
  559. Shockingly, the number he went to, he went to 19 major finals. He went 8-7 though, after his original
  560. 0-4 and ended up, not that it's not respectable, he won eight, but 8-11. I would not be surprised if Zverev, he's
  561. not catching Sinner, Algaraz, any of that. He can end up with four or five majors in his career, which, you know, there's
  562. a lot of great tennis players that didn't win four or five majors. He's now got two and there's no reason
  563. why, by the way, why can't he win the Australian? Why can't he, he could win any of
  564. them next year. So how serious a contender for future championships are, is Ben Shelton?
  565. Hmm. Great question. It's a great question. I think the way, I think where Ben Shelton has to, I've watched every, I watched
  566. a lot of tennis. I love watching the US Open. We are not surprised. I did, I watched a lot of it.
  567. While watching every Yankee game along the way. The problem he has right now. Oh, but he has three TVs, you know.
  568. Yeah. Well, that's it. Multi-screening. He has to improve from the back of the court.
  569. So his serve is one of the best serves in tennis. That part's great. His service return has improved, but he can't
  570. win enough long rallies. So then what he has to do, he has to force the issue on balls that,
  571. you know, a Djokovic during his prime, even Zverev today, certainly Sinner, Algaraz.
  572. Doesn't have to be aggressive with, he has every shot that has a minor possibility of
  573. him hitting a winner. He has to go for it because he's not going to win enough. Nine, 10, 11 shot rallies.
  574. He's not. So now the question is, I don't know whether that's a skill where you can improve
  575. enough to match those other players say, well, he's only 23. Yeah. Well, okay. But even playing professional tennis for five years,
  576. if he knew that that wasn't one of his strengths, why hasn't he improved what we
  577. would call what they call the neutral ball? He doesn't win enough neutral balls.
  578. Does that, is he going to get better than Sinner, Algaraz, Zverev and others? Can you norm a little bit like in
  579. this particular, like, you know, skill where were Sinner and where was Djokovic at his age?
  580. Were they beyond like, right? I mean, like that's certainly. So here's what I can say.
  581. So Sinner has five majors. He's only a year older than Shelton. Algaraz is younger than Shelton and he's got
  582. seven majors, um, Djokovic and Nadal, Nadal had more majors than anybody by this age.
  583. So I'm just saying what we can already, I think eliminate relates to our first discussion
  584. with Tez, Tez about what you can learn, I think we can already say with some reasonable, like Spence Shelton's unlikely, not impossible, he's
  585. unlikely to be an all-time great, I just, I'm asking will it be one of those four or five every year is contending
  586. in all the. I think his best case scenario is he ends up like a Andy Murray, Stan Verinka,
  587. which is, he ends up with two or three majors. I was going to say, I was going to ask you whether you expect a number
  588. of majors. I would be, I would say, you know, his confidence interval has to have a number.
  589. We have, can't overestimate certainty here. It's gotta be wide. I would take the under though.
  590. If you asked me whether he's going to win more than three majors, I think I would take the uncertainty.
  591. I would take the under, I think there's more than a 50% chance. He ends up with three or fewer majors
  592. in his career. I do. What were his odds of beating Zverev? Uh, Zverev was minus, I think about 160
  593. to start the match. Um, and so what's that about? 60%, 65%. Um, and again, cause let's remember when Sinner
  594. and Alcaraz come back healthy, then the problem you have is you have to beat maybe
  595. two of them or even possibly three of them to win the major. And that's, he just doesn't have the consistency
  596. to do it. We should also comment by the way, on the women's side, which was a big deal.
  597. You know, it was just as big as the men's side. Um, Elena Rabakina won the major.
  598. Well, she won. She's now the number one, Rabakina is how it's pronounced. Yeah. It's pronounced Rabakina.
  599. She's now the number one player in the world. She now has three majors. Also Sabalenka, who she beat, she's got a
  600. problem. I mean, she's now, she won like her first four majors. She's now own four in a row in
  601. major finals. She holds none of the majors right now. The U S open was the only one.
  602. As a matter of fact, of the last eight majors, she just has the last year's U S open.
  603. She hasn't won any of them. And for someone who was number one in the world for two years, and she had
  604. one major, she's underperforming what her ability would say. And so I'm starting to get to a
  605. point where she's going to win more majors. But I want to give you an analogy. She's got the same number of majors as
  606. Naomi Osaka. Matter of fact, she's won the exact same major record as Naomi Osaka.
  607. Two Australians and two U S opens. She's never won Wimbledon. She's never won the French.
  608. So she has only hardcore majors. If you would ask me at age 23, how many majors Naomi Osaka would have ended
  609. up with? I might've given you the upper bound is nine, 10. I'm pretty confident.
  610. She's not going to win another major. I think Sabalenka will, but is she going to win eight, nine?
  611. Now I'm not as convinced. I think she's having a trouble. Look, she obviously is really talented.
  612. She's getting to the later stage of tournaments, but she's not winning them. Eric, can you tell us a little bit
  613. more about Ravachana because she's the new biggest thing, right? And she's, and she's number one.
  614. She, and how, when is she to win these three? And what are your forecasts for her?
  615. She won her first major, as I remember in 2022. Oh, that long ago. She did. And, um, then she won in 2025.
  616. She won a major last year. I think she won Wimbledon, not this past Wimbledon, that was Noskova.
  617. It was the year before that. And now she's won obviously the U S open this year.
  618. She look, I think any person that watches tennis would say this. She's got the most, um, she hits the
  619. ball harder and cleaner than anybody in women's tennis. So, and, and she has a great serve.
  620. So she's going to be extraordinarily effective. She's 27 for the next, at least four
  621. or five years on hard courts. She has to be the favorite. Is it common for a women's tennis player
  622. to get that much better? It seems like she's gotten better over five years, a five year span between her first
  623. and third or to, to the lay outside watcher. That feels like a long growth. Yeah, I would agree with you.
  624. I, if you had asked me two, three years ago or four years ago, when she won her first major, how many majors she'd
  625. buy have by this point? I'd be like, wow, looking at her game, I would have guessed a lot more.
  626. I would have guessed a lot more. Three seems low for her given what I saw. And she, she blew through everybody.
  627. I mean, she, she defeated a very hot and well-playing Coco golf. I mean, she, she had the tough road.
  628. But what, what is it? What, what if our model of women's tennis is just wrong? What, what if it's more like golf where
  629. you have some of these regime shifts, players get hot, but then they shift back down.
  630. And in the meantime, their latent talent is actually pretty close to each other.
  631. There's not, there's not real standouts. They just seem in these regime shift moments
  632. to be standing out. In fact, they're all just jumbled together. And so it's a jumbled together, bunch of
  633. people equally skilled with some regime shifts. And I, let me just say, I think that's a fair comment over the last 10
  634. years. I mean, but Serena in a totally different league, I think over the last 10 years,
  635. that's why you've seen, we've talked about this on the show many times in women's tennis,
  636. there's six to 10 women that can win at any major, which, which makes your theory,
  637. right, which is there's a bunch of highly clustered, very accomplished players who on any given
  638. tournament could be the one who pops up. Okay. So that's not on the men's side, but
  639. on the women's, that's a simpler model than I said, which is better. But it does feel like there are these
  640. little runs that go on in the way that golf, it does feel like there's this. Yeah.
  641. And like, you're almost arguing for like a hidden Markoff, like versus not, which is more
  642. complicated. I don't love complicated models, but the only problem with tennis, unlike golf.
  643. Is first of all, there's a lot of, there's more spacing between the tournaments, which allows,
  644. you know, in some sense, she's it's September. She just won the U S open. They're Australians, not for four months.
  645. Okay. Number two, remember they play on three different surfaces. So now I back to my non-unidimensional
  646. ability model. I think we'd all agree in tennis. Like someone, the greatest thing about Nadal, I've,
  647. I've said this a million times, takeaway is 14 French opens and he's still won more
  648. majors than John McEnroe he's Jimmy Connors and he has 14 French opens. So that's incredible that he still has eight
  649. other majors, but we agree there's differences by surface. So not only is there spacing, but there's
  650. three different surfaces. And someone even argued the Australian and the U S open while they're both hard courts
  651. aren't even. So you could argue there's four surfaces spread over time, which is a little different than
  652. golf for sure. For sure. It, you'd have to say the, the null would be forget the regime shift Markov model
  653. thing. It just feels like into it. So here's what that tells me. My perception that there are such things is
  654. just randomness. I'm over-interpreting, but let me, well, let me just say, by the way, I'll just
  655. made a lot of, um, I want to make, I want to make this statement that I've made.
  656. I've missed your silent judgment. No, but I'm, I'm, I'm agreeing, but let me just say, I'll make a statement about
  657. my home field of marketing, which I, which relates entirely to your point, Kate, we've spent
  658. as Bayesians, we've spent so much time discussing heterogeneity, meaning my preferences are different than yours,
  659. we've spent much less time talking about within person variation that we move between States.
  660. And my view is the future of marketing, because where you have longer and longer time
  661. series data at the individual level is going to allow us to measure that. And I actually think there's greater, once you
  662. account for within state variance, I'm not sure you need as I know this by the law of total variance.
  663. You don't need as much heterogeneity. And so there's a whole bunch of projects I'm working on related to that, but I'm
  664. totally with you that within state variability is not studied enough. Great. Okay.
  665. We've only got a few minutes, but we have to talk a little bit of baseball. I see that the Yankees have clenched at
  666. least a wild card and maybe a wild card spot. Is that right? Only a wild card spot.
  667. Tampa Bay is about three and a half games ahead. And they're AAL wet. Hold on. So real quickly, context.
  668. It's two weeks. So 13 games, 12, 13 games, 12, 12, 13 games left. There may be a winner.
  669. We always talk about this at the beginning of the season, how many hundred game winners
  670. are going to be. We were so certain the Dodgers are going to make it. They're, they're a long shot.
  671. Tampa Bay's are right. We're Brewers are the best chance to make it. So I would estimate.
  672. Uh, one. It's worth noting about the Braves, aren't they? The Braves are a long shot to make
  673. a hundred. It's worth noting that again. There are 59 losses. The Brewers are the Brewers only need seven
  674. out of 12. Yes. They're, they're the best. The only good teams in the American league
  675. are all in the ALEs. Yep. It may turn out as you've seen Shane, all the playoff teams.
  676. I mean, not all, not the division, Toronto might do three, all three wild cards may
  677. come from the ALE. Oh yeah. The AL West division leader, I think is below 500 or 500 right now.
  678. Cleveland, I believe is ahead of Toronto. Yeah. But they're both right at five. That's amazing.
  679. You're basically saying the Jays were one game below 500 might make it. Yeah. Might make it.
  680. It's them or the guardians are one game above 500. That's amazing. Well, the Rangers are also, they're only two
  681. games below 500. Yeah. Okay. And Vladi Jr. Got his first home run at home. So I put in the rundown, he's has
  682. almost 200 career home runs. He's this is his 68th home game of the season. And he had his first home run at
  683. home. We don't understand how that, how that can happen. Is that even what, what about the catcher
  684. out in the memory? Yeah. You're Kyle Riley. I said, I think in the bottom 10 batting averages all time now for a season.
  685. Oh my, that, that boy. He had 60 home runs last year. 60 or was it 61? Yeah. And he has had one of the most
  686. absolutely catastrophic second, second follow-up seasons. Is there an explanation?
  687. Is it something physical or is this something? It often is that doesn't get revealed until
  688. later. And he is a catcher and catchers are catching. Yeah. That would be kind of given that he's
  689. a catcher. I think that would probably be today's game. It's harder and harder. They throw so damn hard in a minute
  690. or so. You could, this does not, this relates though to your new England Patriots kicking thing.
  691. So if there was some sort of injury, then I'm about to look it up, but like how many home runs does Vlad Guerrero
  692. have on the road this year? And so the answer is, I can tell you right now, I'll tell you in one
  693. second, it's some, not zero number. It's not a lot. It's like, well, it's might be 16 or
  694. 17. Like, right. That's nuts. That's a big difference. So if you do a Fisher's exact test
  695. on this and you're going to be like, I'm going to use my prepositions correctly.
  696. Is SkyDome a place, a tough place to hit home runs? No, not necessarily. No, not at all.
  697. He certainly hasn't struggled in previous years. Actually, shockingly, he's having a really, he's having
  698. a bad season. He hasn't had that. He only has nine home runs for the entire season, but he does have eight on
  699. the road, but yes. Yeah. But that's not even, that doesn't even rise to craziness.
  700. Okay. No. Eight and one. Yeah. If you just throw down nine home runs and eight goes in one, he's just hasn't
  701. been playing guys. Anything interesting is he's such a great player. Anything interesting on the NL side are Philadelphia's
  702. local Philadelphia Phillies. Nobody's clenched over there, except the Brewers and
  703. Dodgers have clenched their divisions. The Phillies have the top two pitchers in
  704. the NL in terms of at least baseball reference war. It's amazing that they are. I've still trailed the Braves.
  705. But didn't they just come out of the Braves and Lozardo? Yeah. They're excellent pitchers.
  706. They went down to Atlanta and lost two out of three or something. They did. Yeah.
  707. Yeah. But by the way, the thing that's plagued the Phillies for last bunch of years, their
  708. bullpen is bad. And so Doran blew a game into the Braves. And then I forget who else, they were
  709. leading late eighth inning and later in two at all. They could easily have swept the Braves.
  710. Was this surprising at the deadline? No. The team is well resourced as they are.
  711. I mean, Doran was the move last year. That was the move. Brian, try and help this.
  712. And I think he has, I mean, he's probably the best closer they've had in like five or six years.
  713. Right. I would agree with that. Um, but you know, it's just still, you know, I mean.
  714. It's their middle relief guys or. Every, yeah. Never. No, I understand. Right. So what, look, this comes to an interesting.
  715. Every time I see Ryan Kirkering, I'm like. Right. Well, that's the guy that's really bad.
  716. So this is, look, this is the question. You're asking another strategy question, Kate.
  717. If I have Wheeler, Lizardo, um, Sanchez, even Nola, it goes back to what Adi said
  718. from the beginning. Let's imagine they're pitching six innings and they're at 90 pitches.
  719. It's in the post-season. You go. See, this is my point. You know that this is an on day
  720. for that pitcher. Trust me, any Phillies reliever. You do not know. They should pitch these guys, 110, 120 pitches
  721. in the postseason. And if they don't, then Mattingly is out of his mind. Well, listen, we saw that last year.
  722. I mean, Yamamoto pitched all this for the Dodgers. Double what he normally goes.
  723. Not double, but much later in the game. Do it! But then in the final series, but in
  724. the earlier, in the playoffs as well. Yeah. Well, they also had very high quality pitching
  725. performances from their starters. So they really didn't like, you know, you absolutely have to leave them in.
  726. Yeah. You, you do. Particularly when you don't have a weak bill. A trick, a trickier, um, situation would be
  727. for the Yankees who have a very strong bullpen and they've been pulling their starters very
  728. early, all year round. And they've been, and the relievers have, have really come together in the last month, that's
  729. been the backbone. The Yankees have scored so little, so little, and they're winning games by, if you look
  730. at how many games they've won where they've scored fewer than two runs or two runs
  731. or fewer, it's like more than half their game. It's like, it's ridiculous. You're supposed to lose fast and it's their
  732. middle relief. That's done really well. And they were not great all season. They've really put it together.
  733. What difference does it make the judges back? Well, a huge amount. If judges judge, I mean, the thing is,
  734. is that he, I think he rushed his return. He has looked smart. I mean, he's struck out like nine times
  735. in the first 14 at best. Right. Um, he connected yesterday. He has two singles, a walk.
  736. He does not look like the judge that you're going to need. The idea that he feels like he needs
  737. to get back to minor leagues. He didn't do any rehab. Literally none. And he seems to have a super linear
  738. effect on that lineup. Yeah. Like at least in past, what does it mean? Super linear in that, like somehow the hitters
  739. around them get better interaction effect at sea, like, like that, that offense seems more tied
  740. than it's even given his obscene sort of like, so you didn't mean very linear. You met, um, interactive, interactive.
  741. Interactive. Yeah. Okay. No, that's what I'm learning. I'm learning a term. No, no.
  742. So, so I mean, the thing about judges also, he, he's, he, the pitchers fear him.
  743. So there's more walks. It just having him in the, in the lineup just create something that seems to help
  744. the Yankees. No, no, it makes sense when you got that guy, but is it, why didn't he
  745. just, did he need some games before the playoffs started? Is that, I think so. I think he did.
  746. He wanted, he wanted to pay some actual major league pitching, not go to the, to
  747. the AAA league. They're in a divisional race right now. These are important. He said he did nothing.
  748. That was shocking. It's like nothing. Okay. All right, guys. Well, we got too late, two weeks left
  749. in the baseball season. Um, we have both football, uh, NFL and college off the ground.
  750. It's a fun time of year. It's ramping up to maybe one of the best. And actually hockey, hockey's going to start this
  751. month, the end of September. I do have one, I do have one quick, I know we're late on time, but
  752. just one quick college question to ask you, Cade. So weren't we all over Michigan a week
  753. ago after they should have lost or could have lost to Western Michigan. Now all of a sudden they win and
  754. I see them ranked number 19. So I was just asking you, does that make any sense?
  755. Like, like it just seemed like I understand they beat Oklahoma, right? They beat Oklahoma.
  756. Nice win. Very nice win. But they almost lost. I mean, if the clock hadn't been, they
  757. could have easily lost to Western Michigan. So does it surprise you that they were
  758. ranked number 19? Well, what you want to know is what were the priors before they almost lost to
  759. Western Michigan, right? So what do we think had been coming into the season was Western Michigan, the aberration,
  760. right? And you have to order the two games differently. Would you feel differently about Michigan?
  761. Oh, a hundred percent, but you know, college football has games differently, they wouldn't be ranked.
  762. College football has a big standard deviation. People forget that. That's weird.
  763. That's so you should draw a lesson from these post games real quickly. Guys, a buddy of mine went Patrick Willis,
  764. a buddy of mine went to the game with me on Saturday night, and I think he had the best fan experience a person
  765. can have at a sporting event for the following reason. Eric just brought it up. Michigan played Oklahoma earlier in the day, Texas
  766. played Ohio state. Two of the bloodiest blood rivals in college football are Texas, Oklahoma, and Michigan Ohio state.
  767. Like literally they hate each other. Both of them, Michigan Ohio state is frankly
  768. another category above, but they hate each other. So there was this meme going around on
  769. Instagram about, it was, it was an office sketch about, um, Texas leaning into Michigan saying,
  770. do you want to be in a coalition with me? And they turn around and says, absolutely.
  771. And so Michigan was not supposed to win that game. No one gave Michigan a chance because they
  772. looked so bad. Oklahoma is supposed to be strong. They take out Oklahoma early in the day.
  773. And then my buddy wears Michigan gear to the Texas game. And so everybody is giving him love all
  774. day long for beating Oklahoma. Meantime, he's rooting his heart against Ohio state
  775. as I was. And he, people wanted to take pictures with him after the game. People were giving him high five before and
  776. after the game. Because he was in Michigan, people were shouting, even if it's not your team, you must
  777. go to a game with a rooting interest. It's really hard to go to sporting events.
  778. If you have no rooting interest, this Michigan guy was rooting against Ohio state.
  779. That's a rooting interest. There was a guy that betting is serving his forces creates a rooting interest, but also
  780. sports sport. It's good to hate watch. Yeah. You work hate out of your system, kind
  781. of an innocent way, but I feel like I really kind of foster the heat in my system through sports.
  782. It's been really nice for me, actually. All right, guys, let's leave on that note.
  783. Warton Moneyball. Thank you guys for listening for the whole crew here. Audie Weiner, Shane Jensen, Eric Brettel.
  784. This has been Cade Massey, Aaron Tran pulling double triple duty in the studio today.
  785. Many thanks, Jacob Grodnick on the board as usual, coming in remotely and Deep Patel, who
  786. I think might still be on is his second martini back there. We've still got some bonbons.
  787. What's going on? Deep Patel boss lady in here. Thank you guys for listening. Come back and join us next time.
  788. Between now and then enjoy your sports.

Brief Summary

Tej Seth discusses NFL Week 1 analysis, college football matchups, and MLB playoff races.

Key points

  • Tej Seth Joins the Show. Tej Seth, a rising football analyst, shares his insights on NFL trends and performances.
  • Ravens' New Offensive Coordinator. Tej discusses the impact of the Ravens' new offensive coordinator on their gameplay.
  • Impressive 49ers Victory. The 49ers showcased dominance against the Rams, surprising many with their performance.
  • Josh Allen's Incredible Performance. Josh Allen led the Bills to a notable victory against the Texans, showcasing his skills under pressure.
  • Texas-Ohio State Showdown. Texas staged a thrilling comeback against Ohio State, winning with a 21-point fourth quarter.
  • Quarterback Clutchiness Debate. A discussion on whether some quarterbacks perform better under pressure, exploring the concept of 'clutchiness'.

Episode highlights

Tension

45/ 100Some tension

Related episodes