WEBVTT - March Madness Bracket Extravaganza 0:00:15.604 --> 0:00:28.764 Pushkin. Welcome back to Risky Business, our show about making 0:00:28.804 --> 0:00:31.404 better decisions. I'm Maria Kanakova and. 0:00:31.404 --> 0:00:33.844 I'm Nate Silver. Did you say our show or a show? 0:00:34.004 --> 0:00:36.804 I said our show, Nate, because it is our show. 0:00:37.364 --> 0:00:40.884 Sometimes decision say A you know, I like to play 0:00:40.884 --> 0:00:42.164 around with it the show. 0:00:42.204 --> 0:00:45.564 How about the show? So I'll tell you how to 0:00:45.604 --> 0:00:46.764 make better decisions? 0:00:47.204 --> 0:00:49.964 This is the show about how to make better decisions. 0:00:53.604 --> 0:00:55.444 What are we going to be talking about today. 0:00:55.284 --> 0:01:00.484 Night March madness, Maria, March madness. Indeed, it's New York, 0:01:00.684 --> 0:01:04.244 it's springtime. The weather is improving from mediocre to slightly 0:01:04.284 --> 0:01:06.044 go to the mediocre. And we're here to watch some 0:01:06.124 --> 0:01:07.724 basketball and talk about some basketball. 0:01:08.044 --> 0:01:08.324 Yep. 0:01:08.404 --> 0:01:12.004 And we're going to start off by talking about some 0:01:12.044 --> 0:01:16.084 of Nate's models, and then we each made brackets. And 0:01:16.444 --> 0:01:18.844 I'll also introduce a secret weapon that I used in 0:01:18.884 --> 0:01:23.404 making my bracket, a certain someone named Monty McNair. 0:01:23.564 --> 0:01:24.644 Oh well, so you cheated. 0:01:25.044 --> 0:01:26.324 Oh no, I want to not cheat. 0:01:26.444 --> 0:01:27.804 I do want to McNair. 0:01:30.124 --> 0:01:33.124 Nate, I did not cheat. When we don't know how 0:01:33.164 --> 0:01:35.764 to make a good decision, what do we do? We 0:01:35.844 --> 0:01:39.284 consult someone with the expertise and the knowledge to help. 0:01:39.724 --> 0:01:44.204 So I consulted with Monty McNair, who's the GM of 0:01:44.244 --> 0:01:47.604 the Sacramento Kings and something of a guru when it 0:01:47.644 --> 0:01:51.124 comes to making March Madness brackets. That full interview is 0:01:51.164 --> 0:01:54.524 actually going to be on push Can Plus highly recommend 0:01:54.524 --> 0:01:55.724 that you all listen to it. 0:01:58.244 --> 0:02:00.524 I feel like the politics that's a bit complicated where 0:02:00.524 --> 0:02:04.204 the woman has to ask it certified basketball expert. Well, 0:02:04.204 --> 0:02:06.444 I just kind of go the man just you know, 0:02:07.364 --> 0:02:10.044 uses a regression based model. 0:02:11.804 --> 0:02:15.844 So that's that's a great segue night, Let's talk about 0:02:15.844 --> 0:02:20.964 the regression based ELO model. So you released your Marsh 0:02:20.964 --> 0:02:23.484 Madness models. How did you you know, how did you 0:02:23.524 --> 0:02:25.684 make them? Kind of what's the thought process behind them? 0:02:25.884 --> 0:02:27.404 What goes into them? 0:02:27.764 --> 0:02:30.164 So some of this like actually literally dates back to 0:02:30.244 --> 0:02:34.084 two thousand and two or two thousand and three, right 0:02:34.084 --> 0:02:38.684 where I'm like a junior associate at KPMG. This is 0:02:38.684 --> 0:02:41.644 an accounting firm. I always was sensitive. I'm you know, 0:02:41.684 --> 0:02:44.244 it was a consultant at accounting firm, Maria, not an accountant. 0:02:44.364 --> 0:02:46.764 Just so you know, I always forget that you worked 0:02:46.764 --> 0:02:50.444 at an accounting firm. Every sing such a year, Oh yeah, right. 0:02:50.684 --> 0:02:53.124 Such a weird part of the multiverse where I don't 0:02:53.164 --> 0:02:55.084 know how the fuck that happened. I don't know, but 0:02:55.244 --> 0:02:57.844 like it's because I parted too much ching college and 0:02:57.844 --> 0:03:03.244 then had to didn't really make plans anyway. So a 0:03:03.284 --> 0:03:05.764 friend who had an office pool had these complicated scoring 0:03:05.884 --> 0:03:07.684 rules where you get points for upsets, and so I 0:03:07.724 --> 0:03:12.164 took some other rating system I found and modeled that 0:03:12.204 --> 0:03:14.404 and did a little simulation. Actually it's not a simulations, 0:03:14.644 --> 0:03:18.444 it's not deterministic exactly, but like, there are only sixty 0:03:18.684 --> 0:03:22.244 seven games, so you can kind of explicitly do all 0:03:22.244 --> 0:03:24.884 the conditional probabilities instead of having to rely on a simulation. 0:03:25.084 --> 0:03:25.764 Right, just a. 0:03:25.684 --> 0:03:28.564 Technical detail, but it's evolved from there. I think when 0:03:28.564 --> 0:03:30.604 I was at the New York Times in twenty twelve, 0:03:30.644 --> 0:03:33.284 I did a big evaluation looking at all these different 0:03:34.924 --> 0:03:38.564 computer models, all these other different ways to predict the tournament, 0:03:38.604 --> 0:03:41.804 and found kind of very empirically which did best in 0:03:41.844 --> 0:03:44.564 predicting tournament games. So a couple of things that if 0:03:44.564 --> 0:03:47.204 you if you've ever made a model, might sound familiar, right. 0:03:48.324 --> 0:03:51.444 You know, first, almost always taking a consensus of indicators 0:03:51.964 --> 0:03:54.764 is better than any one indicator, right, So we wound 0:03:54.844 --> 0:03:57.164 up with a blend of like five different rating systems, 0:03:57.164 --> 0:04:00.644 all which do a little differently. Accounting for injuries is 0:04:00.764 --> 0:04:02.964 very important. You know, you don't want to ignore the 0:04:02.964 --> 0:04:05.604 fact that Cooper Flag, maybe the best player in the NCAAA, 0:04:05.604 --> 0:04:08.684 at least the best freshman the nca IS year, got 0:04:08.764 --> 0:04:12.404 hurt during a tournament. He seems to be helping now, 0:04:12.804 --> 0:04:14.604 but if he were to be heard, that would gravely 0:04:14.604 --> 0:04:15.764 impact Duke's prospects. 0:04:15.764 --> 0:04:16.604 So we account for that. 0:04:17.684 --> 0:04:21.044 And then also we found that, like there's reversion to 0:04:21.084 --> 0:04:24.284 the mean, meaning that you play roughly thirty three games 0:04:24.324 --> 0:04:26.964 or something in a college basketball regular season, not all 0:04:26.964 --> 0:04:29.484 that biggest samples some of us are against crappy competition, 0:04:29.724 --> 0:04:34.924 you know from you know, Southwestern State, University prep or whatever. Right, 0:04:36.124 --> 0:04:38.644 So we looked at pre season rankings as well as 0:04:38.724 --> 0:04:41.764 part of what we factored into the formula. Sometimes seems 0:04:41.764 --> 0:04:44.884 that underachieve have actually gotten a little bit unlucky, so 0:04:44.964 --> 0:04:48.324 believe it or not, giving human ratings some component actually 0:04:48.324 --> 0:04:51.444 helps a little bit, and this proved to be a 0:04:51.484 --> 0:04:53.884 popular feature. But like more recently, I began making my 0:04:54.004 --> 0:04:57.284 own rating system that kind of incorporates some of these 0:04:57.284 --> 0:05:01.804 different ideas. Right, So are you familiar with ELO ratings. 0:05:01.844 --> 0:05:03.604 I know some of our readers listeners might be. 0:05:03.604 --> 0:05:06.444 I should say, well, let's for people who are not familiar. 0:05:06.884 --> 0:05:11.804 Of course, I personally am very you're familiar with the writings. 0:05:12.164 --> 0:05:15.764 But for those of us who might not be, why 0:05:15.764 --> 0:05:17.044 don't we explain. 0:05:16.644 --> 0:05:17.324 It a little bit? 0:05:17.764 --> 0:05:20.244 So ELO was designed by let me look at the guys, 0:05:20.284 --> 0:05:22.844 it's like arpanned Elo, or let me look, it's Hungarian. 0:05:22.844 --> 0:05:27.604 Probably the Hungarians are quite promiscuous within in kind of 0:05:27.644 --> 0:05:30.764 this era of intellectual history, are bad Elo a Hungarian 0:05:30.844 --> 0:05:36.404 American physics professor. But it's a formula where you are 0:05:36.724 --> 0:05:42.044 adjusting your rating constantly based on whether you beat your opponent, 0:05:42.164 --> 0:05:44.204 of course, and how good your opponent is. Right, So 0:05:44.284 --> 0:05:47.124 if I have an ELI rating of a nineteen hundred, 0:05:47.164 --> 0:05:48.364 and I don't know if that's good in chess, it's 0:05:48.364 --> 0:05:52.484 probably pretty good. Right, And you're a fourteen hundred, Maria, right, 0:05:53.524 --> 0:05:56.844 fifteen hundreds, average of four hundreds whatever, mediocre, I barely 0:05:56.844 --> 0:05:58.964 get any credit because you kind of already forecast to 0:05:59.044 --> 0:05:59.324 the win. 0:05:59.524 --> 0:05:59.764 Right. 0:06:00.004 --> 0:06:02.444 If I'm a nineteen hundred and I beat Magnus Carlson, 0:06:02.444 --> 0:06:05.804 who's a twenty one, I'm just guessing you might know. 0:06:06.364 --> 0:06:07.924 Then I get a lot of credit because I've beaten 0:06:08.324 --> 0:06:09.284 a stronger opponent. 0:06:09.644 --> 0:06:12.524 So twenty eight thirty three night as Magnus's. 0:06:12.564 --> 0:06:16.164 Twenty eighth ro' oh my, that's embarrassing. See in basketball, 0:06:16.164 --> 0:06:18.204 they only get to like twenty two hundred or something, right, 0:06:18.204 --> 0:06:21.284 And so I'm like, is Magnus Carlson better at chess. 0:06:21.084 --> 0:06:24.684 Tha twenty eight thirty three? His peak rating was twenty 0:06:24.724 --> 0:06:25.444 eight eighty two. 0:06:26.324 --> 0:06:28.084 Okay, that's pretty that's pretty good. I think I'm like 0:06:28.084 --> 0:06:33.284 a nine hundred or something. But so anyway, we adapted 0:06:33.284 --> 0:06:35.164 the ELO ratings. Actually other people have done this, but 0:06:35.164 --> 0:06:37.924 we adapted it first the NFL and then to college basketball. 0:06:38.764 --> 0:06:43.044 But there are a few twists in ELO. Originally just 0:06:43.084 --> 0:06:46.124 about the winner loss, right, we also count for the 0:06:46.124 --> 0:06:49.484 margin of victory. In basketball, we account for home court 0:06:50.044 --> 0:06:53.364 advantage in kind of an increasingly complicated way. We account 0:06:53.404 --> 0:06:56.564 for travel distance, so if you fly across the country, 0:06:56.604 --> 0:06:59.084 your jet lag and so forth. This actually shows up 0:06:59.124 --> 0:07:02.044 in in performance by the way different teams have different 0:07:02.044 --> 0:07:05.364 home court advantages. It turns out actually that having like 0:07:05.404 --> 0:07:07.524 a larger crowd actually helps. 0:07:07.564 --> 0:07:08.964 I was trying to model this the other day. 0:07:09.044 --> 0:07:13.164 You know you for every ten thousand fans show up, 0:07:13.844 --> 0:07:16.804 teams score roughly in additional point, that's right. 0:07:16.804 --> 0:07:18.844 That's absolutely fascinating, saying, yeah. 0:07:18.644 --> 0:07:20.724 So you're that's for I do that for women for men, 0:07:20.764 --> 0:07:23.724 if it's different women's because they have some teams are 0:07:23.844 --> 0:07:25.804 more popular than men's teams, and some like literally draw 0:07:26.244 --> 0:07:29.404 you know, twelve people, right, But yeah, the teams that 0:07:29.444 --> 0:07:32.204 actually have people watching do a little better in their 0:07:32.204 --> 0:07:33.884 home games controlling for other factors. 0:07:34.204 --> 0:07:37.044 So I love that you're able to account for these 0:07:37.084 --> 0:07:41.604 things that might seem kind of more ineffable, right, like 0:07:41.604 --> 0:07:45.284 like home court advantages, more psychological and more physical like 0:07:45.364 --> 0:07:50.444 injury risks. How much is like how much of model distinctions? 0:07:50.444 --> 0:07:53.204 And like your model versus other models is kind of 0:07:53.884 --> 0:07:56.004 a little bit more subjective, like as you said, the 0:07:56.124 --> 0:07:59.924 human touch versus kind of the big data crunching. I'm 0:07:59.924 --> 0:08:03.964 just curious, from like a from a psychologist's standpoint, what 0:08:04.044 --> 0:08:05.004 the balance is there? 0:08:05.764 --> 0:08:08.244 So are you saying when I make a bracket or like, when. 0:08:08.124 --> 0:08:10.604 I'm billion, you're actually building the model. 0:08:12.564 --> 0:08:15.724 Look, inherently models, you have a lot of different choices 0:08:15.724 --> 0:08:19.324 you can make. But like, because you have like i'd say, 0:08:19.364 --> 0:08:24.004 this is like less subjective than some of the other 0:08:24.124 --> 0:08:26.404 models I've built because you have a lot of data 0:08:27.244 --> 0:08:31.844 and elo is relatively simple. Right, there are only so 0:08:31.924 --> 0:08:35.084 many knobs you might have to press, versus more open 0:08:35.164 --> 0:08:36.644 ended problems. 0:08:36.724 --> 0:08:36.884 Right. 0:08:36.884 --> 0:08:41.964 But yeah, in general, like with politics election models that 0:08:42.284 --> 0:08:44.084 you know, those get to be a little bit more 0:08:45.244 --> 0:08:47.724 art than science. Not in the sense because people mistake this. 0:08:47.844 --> 0:08:49.844 Not in the sense that I am like subjectively changing 0:08:49.884 --> 0:08:52.364 out once the model is designed, but like, but the 0:08:52.444 --> 0:08:55.644 design questions are harder. And also, you know, so one 0:08:55.644 --> 0:08:58.444 thing you look at in basketball too, is like how 0:08:58.484 --> 0:09:01.324 do the parameters of the game change over time? 0:09:01.604 --> 0:09:05.084 So one thing we do is like, in. 0:09:05.044 --> 0:09:10.404 Between seasons, how you did last season? Provide some information obviously, right, 0:09:10.844 --> 0:09:14.604 you know Connecticut was a national champion last year, had 0:09:14.644 --> 0:09:16.964 a very good year. Right, you wouldn't just toss them 0:09:17.004 --> 0:09:21.524 in a hat versus you know, writer or Canisius or 0:09:21.564 --> 0:09:22.484 one of these small schools. 0:09:22.524 --> 0:09:24.364 I now have beamed him and I had, you. 0:09:24.284 --> 0:09:28.364 Know, Youngstown State, Texas, Rio Grand Valley. You know, you 0:09:28.564 --> 0:09:32.044 you you have some information from the prior season. But 0:09:32.084 --> 0:09:34.324 because there's so much more player turnover now the best 0:09:34.364 --> 0:09:39.924 players go pro after one year, it's less continuity than 0:09:39.964 --> 0:09:41.764 there used to be, right, and so and so you 0:09:41.764 --> 0:09:43.924 should always be asking questions, right, and like you're billion 0:09:43.924 --> 0:09:45.404 a model, part of what you're doing is saying, Okay, 0:09:46.564 --> 0:09:50.364 here's this data. It's been well calibrated on literally what's 0:09:50.524 --> 0:09:53.724 you know, seventy five years of college basketball history, right, 0:09:53.804 --> 0:09:57.884 but let's make sure that it's performing as advertised on 0:09:58.724 --> 0:10:02.164 the past couple of years, right, because after all, if 0:10:02.164 --> 0:10:04.084 you're then adapting this to make predictions, and that's kind 0:10:04.084 --> 0:10:05.884 of what people are going to care about, whether you 0:10:05.884 --> 0:10:08.484 Rectro actually got some game right in nineteen seventy two 0:10:08.884 --> 0:10:14.484 or whatever. And what we found is that the model 0:10:14.524 --> 0:10:16.084 had I know, I'm saying we because it's just me 0:10:16.124 --> 0:10:18.684 at this point. What I found is that the model 0:10:18.724 --> 0:10:25.644 had become overconfident in the in recent years, right, because 0:10:25.724 --> 0:10:28.084 it was it was overconfident early in the season before 0:10:28.084 --> 0:10:31.764 catching up. And that was because it you know, overweighted 0:10:31.764 --> 0:10:33.724 the impact of last season when if you have a 0:10:33.764 --> 0:10:35.964 good player, it's going to the NBA anyway, and so 0:10:36.044 --> 0:10:38.804 and so it doesn't help quite as much and so like, yeah, 0:10:38.844 --> 0:10:42.524 so this is where like domain knowledge helps, right, because 0:10:42.524 --> 0:10:44.524 you could just say, well, it's a fairly large sample, 0:10:44.604 --> 0:10:46.524 but you know, it's been a weird couple of years 0:10:46.564 --> 0:10:48.764 because of COVID and new rules and stuff like that. Right, 0:10:48.764 --> 0:10:51.044 But it's like, Okay, we have a good hypothesis for why, 0:10:51.724 --> 0:10:55.044 for why there's more mean reversion now. But by the way, 0:10:55.084 --> 0:10:57.684 this has actionable implications. 0:10:57.804 --> 0:11:00.284 Where Okay, so. 0:11:02.004 --> 0:11:05.404 For our tournament model, which is related to this ELO model, 0:11:05.444 --> 0:11:09.884 but we combined with other systems. I found myself I 0:11:09.924 --> 0:11:14.804 do literally make bets based on these, right, And I 0:11:14.844 --> 0:11:20.284 found myself betting a lot of underdogs to cover the 0:11:20.404 --> 0:11:24.164 spread in the men's game, but not in the women's 0:11:24.204 --> 0:11:29.284 game interestingly, right, because women you can't join the WNBA 0:11:29.524 --> 0:11:32.484 until you're age twenty two, I think, which seems fucking 0:11:32.564 --> 0:11:35.684 insane to me, Like, how was that legal? By the way, 0:11:35.684 --> 0:11:36.844 that is very strange. 0:11:36.844 --> 0:11:37.684 I don't understand. 0:11:38.204 --> 0:11:42.764 Yeah, the politics THEWWA, you know, it's a whole other segment. 0:11:42.844 --> 0:11:44.964 But yeah, the teams are still as strong as ever, 0:11:45.044 --> 0:11:47.804 and so you have less reversion from year to year. 0:11:47.844 --> 0:11:49.644 So basically for lots of reasons, and the fact that 0:11:49.764 --> 0:11:53.324 like you know, for a long time in the women's game, 0:11:53.844 --> 0:11:58.604 they were like, you know, somewhere between a half dozen 0:11:58.644 --> 0:12:01.844 and two dozen schools. It took women's basketball really seriously, 0:12:01.964 --> 0:12:04.044 and then some where it just like you go out 0:12:04.044 --> 0:12:06.124 with a clipboard and say, hey, who wants to play? Right, 0:12:07.204 --> 0:12:10.924 And so it's almost like professional versus amateur. Now it's 0:12:10.924 --> 0:12:13.164 become much more popular. There's more money in the sport, 0:12:13.244 --> 0:12:16.804 more young women are playing the sport, right, But still 0:12:16.844 --> 0:12:20.044 it's more it's much more lopsided, right, And so you know, 0:12:20.364 --> 0:12:23.324 so I found and now I'm giving away and replaces 0:12:23.364 --> 0:12:25.924 my bet. So too late, right, But yeah, so I've 0:12:25.964 --> 0:12:29.084 ad also fitting a lot of favorites in women's basketball 0:12:29.124 --> 0:12:32.284 and a lot of underdogs in men's basketball, because there's 0:12:32.324 --> 0:12:35.404 been more mean reversion in men's basketball. 0:12:36.724 --> 0:12:38.444 And I don't think that. 0:12:38.564 --> 0:12:42.804 The models are that the odds are accounting for this enough, 0:12:42.884 --> 0:12:44.724 although you know, we'll see I'm putting my money where 0:12:44.724 --> 0:12:45.884 my mouth is at least. 0:12:45.924 --> 0:12:47.044 Yeah, they're just settle things too. 0:12:47.084 --> 0:12:54.204 Like you know, big teams that draw huge audiences and 0:12:54.444 --> 0:12:57.044 have very enthusiastic fans and good in game presentation and 0:12:57.164 --> 0:12:59.604 nice locker rooms, right, they have a bigger home court 0:12:59.644 --> 0:13:02.364 advantage which makes their rating look better. But you're playing 0:13:02.404 --> 0:13:05.564 the NCAA tournament on neutral sites instead, so there have 0:13:05.564 --> 0:13:08.204 actually been quite a few upsets in the tournament in 0:13:08.284 --> 0:13:11.204 recent years. So yeah, all this is kind of you know, 0:13:11.324 --> 0:13:13.204 so that's why it's like, this is what I hate 0:13:13.604 --> 0:13:17.124 when like academics are like, well you a sports model, 0:13:17.204 --> 0:13:19.324 election Wedell, this isn't not real science, Like it's actually 0:13:19.724 --> 0:13:22.564 better science. Sure, because you're like, now I sound like 0:13:22.564 --> 0:13:27.284 a fucking asshole, but no, but you're like you it's 0:13:27.324 --> 0:13:29.964 it's still like hypothesis driven, right, you're looking at the 0:13:30.004 --> 0:13:30.404 real world. 0:13:30.484 --> 0:13:31.604 You're just actually willing. 0:13:31.324 --> 0:13:34.084 To test these ideas in the form of of making predictions, 0:13:34.124 --> 0:13:35.844 in some cases even betting on these predictions. 0:13:36.284 --> 0:13:39.044 No, that that makes perfect sense. And do I understand 0:13:39.044 --> 0:13:41.484 correctly that you actually have two versions of your model, 0:13:41.484 --> 0:13:44.844 one that's elo based and one that's a Bayesian elo version. 0:13:45.444 --> 0:13:47.564 Is that right? Abazing just because I'm a Doric, right, 0:13:47.604 --> 0:13:49.844 But so in the Baysing version, like I said before, 0:13:52.004 --> 0:13:55.044 you uh, the new season begins in the team is 0:13:55.204 --> 0:13:57.084 kind of the same and kind of a different team. 0:13:57.324 --> 0:13:57.524 Right. 0:13:58.964 --> 0:14:02.884 In the original version, it reverts to the mean based 0:14:02.924 --> 0:14:04.684 on the other teams in your conference, right, because there 0:14:04.724 --> 0:14:08.124 are clearly differences between like you know, whatever, the ACC 0:14:08.364 --> 0:14:12.444 and the IVY in the Basing version, then we give 0:14:12.484 --> 0:14:14.084 weight to the human poles. Instead, we look at the 0:14:14.084 --> 0:14:16.964 preseason top twenty five. We assume that these people are 0:14:17.004 --> 0:14:20.404 evaluating recruitment and player aging and Injuriason all this stuff, right, 0:14:20.444 --> 0:14:23.204 and so we revert based on that instead. It turns 0:14:23.204 --> 0:14:27.204 out actually that like you do best with a blend 0:14:27.564 --> 0:14:30.244 of you know, because the humans can add or subtract, 0:14:30.644 --> 0:14:33.764 they can introduce biases that aren't helpful, right, So it's 0:14:33.804 --> 0:14:35.684 really it's really kind of a what we call base 0:14:35.764 --> 0:14:39.204 is really like a blend of the original version and 0:14:39.364 --> 0:14:45.124 kind of a pure Bayse version. So that accounts for 0:14:45.204 --> 0:14:49.284 like preseason makings in the human poles, and that has 0:14:49.324 --> 0:14:50.284 interesting implications. 0:14:50.684 --> 0:14:54.244 Do they have the same results this year or not? 0:14:54.404 --> 0:14:55.684 Do your two models differ? 0:14:56.844 --> 0:14:58.964 Yeah, let me show you it's actually not? Or is 0:14:59.004 --> 0:15:01.044 it pretty similar that different? 0:15:01.124 --> 0:15:01.644 I mean, for. 0:15:03.404 --> 0:15:06.524 So the Basing version has Duke as the best team 0:15:07.484 --> 0:15:10.764 and the pure version has Floored as the best team. 0:15:10.804 --> 0:15:12.764 Now these are all within like a few points. It's 0:15:12.764 --> 0:15:15.884 a very minor difference, right, But like you're counting for 0:15:15.964 --> 0:15:18.084 like that you're basic kind of counting for like the 0:15:19.124 --> 0:15:24.124 historical reputation of Duke that gets boosted in these human polls. 0:15:24.204 --> 0:15:24.604 Got it. 0:15:24.964 --> 0:15:28.284 Well, that's all super interesting, Nate. So let's see what 0:15:28.364 --> 0:15:31.724 the results show when it comes to actually choosing brackets. 0:15:31.724 --> 0:15:34.004 So you got to use all of your models. I 0:15:34.084 --> 0:15:36.844 did not, but I got to use a Monte McNair. 0:15:37.404 --> 0:15:40.964 So I got to use a human and after the breaks, 0:15:41.044 --> 0:15:58.684 let's see how it all turns out. Nay, I have 0:15:58.724 --> 0:16:02.324 a confession to make. I have never filled out an 0:16:02.444 --> 0:16:05.524 NC double a bracket until now. 0:16:05.764 --> 0:16:06.004 Ria. 0:16:06.324 --> 0:16:08.404 This is my first ever bracket and. 0:16:08.324 --> 0:16:09.884 We're still allowing on a show. 0:16:10.084 --> 0:16:14.444 Come on, So this is my first ever bracket. 0:16:14.444 --> 0:16:16.004 But how can you never filled out a bracket? 0:16:16.084 --> 0:16:21.844 I've never filled out a bracket, Nate, I confess, I confess. 0:16:22.524 --> 0:16:24.124 You know I've never done yoga. 0:16:24.324 --> 0:16:26.604 Oh there you go, Nate. How could you have never 0:16:26.644 --> 0:16:27.204 done yoga? 0:16:28.364 --> 0:16:31.724 It feels never the gender stereotype being here. I don't know, 0:16:31.804 --> 0:16:33.644 but like it's like something you think you would have 0:16:33.604 --> 0:16:34.724 done once. 0:16:34.804 --> 0:16:37.844 Right, Okay, Nate, I will I will ever do yoga 0:16:37.884 --> 0:16:40.244 with you. And now that I filled. 0:16:40.044 --> 0:16:43.364 Out our bracket, okay, yeah. 0:16:42.804 --> 0:16:46.524 So so I actually you have your bracket right, and 0:16:46.564 --> 0:16:50.204 it's probably a normal bracket. I have two brackets, so 0:16:50.364 --> 0:16:53.884 Here's here's what I did. So first, I know nothing 0:16:54.044 --> 0:16:57.084 you know about NCAA basketball. I never filled out a 0:16:57.084 --> 0:17:00.124 bracket before. So I did a bracket based on vibes. 0:17:00.724 --> 0:17:03.764 So I didn't even look at the at how teams 0:17:03.844 --> 0:17:06.284 were ranked, didn't look at their record, didn't look at 0:17:06.324 --> 0:17:09.924 any data whatsoever, and just like went on a totally 0:17:10.004 --> 0:17:12.884 vibe based like who do I like? What sounds better? 0:17:13.244 --> 0:17:15.684 What state do I like better? Where's the weather better? 0:17:16.044 --> 0:17:19.644 Just just random things like that, my ViBe's bracket night. 0:17:19.724 --> 0:17:21.684 Do you want to guess who won at the end? 0:17:22.204 --> 0:17:23.524 Who I have winning the whole thing? 0:17:25.204 --> 0:17:26.284 H Yale? 0:17:27.244 --> 0:17:29.044 No, I'm a Harvard grad. 0:17:29.204 --> 0:17:30.804 Okay, it doesn't make sense. I don't. 0:17:31.484 --> 0:17:37.884 Uh, who would This is a weird what's it like? Laden? 0:17:38.644 --> 0:17:40.844 Like? What would Maria think? Is exactly? 0:17:40.884 --> 0:17:43.004 This is a perfect psychology experiment. 0:17:44.004 --> 0:17:45.444 Uh, give me give me a hint. 0:17:46.324 --> 0:17:49.124 You weren't You weren't that far off when you said 0:17:49.244 --> 0:17:51.564 Yale in terms of the types of things I was 0:17:51.604 --> 0:17:54.724 thinking about. But it was a place. It's so it 0:17:54.804 --> 0:17:57.884 was an emotional based decision, all right. 0:17:57.924 --> 0:18:00.244 Okay, So one thing is, are you enough of a 0:18:00.244 --> 0:18:02.804 college basketball fan to know that you're supposed to hate Duke. 0:18:03.524 --> 0:18:06.644 Yes, okay, okay, because. 0:18:06.404 --> 0:18:09.644 That's you know, that's like the point he headad like 0:18:09.684 --> 0:18:11.804 by the way I was out, Uh, I don't know. 0:18:11.884 --> 0:18:17.324 I was out with my partner last Saturday, and we'll 0:18:17.324 --> 0:18:19.444 at least is maybe really too much? Like why is 0:18:19.484 --> 0:18:26.324 everyone in this bar so so white? Because it's New 0:18:26.404 --> 0:18:28.844 York and it's a big bar and usually and it's 0:18:28.884 --> 0:18:30.124 because the New Game was fun? 0:18:32.644 --> 0:18:35.524 That's funny. But okay, so, so do you give up? 0:18:35.844 --> 0:18:38.924 Should I just tell you? Let me give give me 0:18:38.964 --> 0:18:39.964 one more, one more guess? 0:18:39.964 --> 0:18:44.844 Okay, Saint John's no is there a New York guess? 0:18:45.084 --> 0:18:46.564 Or not? Maybe? I think, Yeah, I guess you're not 0:18:46.604 --> 0:18:48.204 really like it. I mean, you were live in New York. 0:18:48.244 --> 0:18:50.844 You don't like it New York born and bread. 0:18:50.844 --> 0:18:51.484 Okay, I give up? 0:18:51.564 --> 0:18:55.604 All right? So I had you're gonna laugh. I'm curious 0:18:55.604 --> 0:18:58.084 to hear your reaction. I had U c l A 0:18:58.164 --> 0:19:01.204 winning because I felt bad for LA and the wildfires, 0:19:01.204 --> 0:19:03.204 and I felt like they needed something good to happen 0:19:03.244 --> 0:19:03.524 to them. 0:19:03.804 --> 0:19:08.284 Okay, you know, I'm generally pro you. So I like 0:19:08.324 --> 0:19:09.124 that color mix. 0:19:09.524 --> 0:19:13.844 Yeah, there was just UCLA had good vibes and I figured, 0:19:13.884 --> 0:19:16.164 you know, what let's have them win. So I had 0:19:16.284 --> 0:19:18.764 UCLA beating Duke in the final four. 0:19:19.924 --> 0:19:23.884 What seedd UCLA even this year there are seven seed probabilistically, 0:19:23.964 --> 0:19:26.844 let me let me look up the old silver bulletin 0:19:26.964 --> 0:19:31.364 forecast and see what it says. So you have an 0:19:31.444 --> 0:19:36.764 zero point three percent chance of being right? All right, yes, 0:19:36.884 --> 0:19:38.684 but getting it it's like getting aces right. 0:19:39.004 --> 0:19:43.444 Yeah, absolutely so then okay, So I showed Manty this 0:19:43.604 --> 0:19:45.564 vibees based bracket and I was like, will you help 0:19:45.604 --> 0:19:48.124 me correct it? And he was like, dude, we need 0:19:48.164 --> 0:19:52.004 to start from scratch. So, for listeners who don't know, 0:19:52.444 --> 0:19:56.564 Monty McNair is the GM of the Sacramento Kings and 0:19:56.804 --> 0:19:59.724 he went to Princeton and he did his thesis on 0:19:59.804 --> 0:20:03.604 March Madness and on building brackets. So this is actually 0:20:03.644 --> 0:20:07.924 what he studied academically. And about eight or nine years 0:20:07.924 --> 0:20:13.004 ago he came in second in the Cagle bracket contest. 0:20:13.044 --> 0:20:16.964 So this is someone who is really really good at 0:20:17.004 --> 0:20:21.164 making brackets. And I asked Monty kind of how he 0:20:21.244 --> 0:20:23.524 would go about doing this. And Monty has his own 0:20:23.564 --> 0:20:27.444 models as well, by the way, So let's actually listen 0:20:27.524 --> 0:20:30.324 to a brief clip of that conversation. And then we'll 0:20:30.364 --> 0:20:34.284 talk about what the resulting bracket looks like, and we'll 0:20:34.284 --> 0:20:35.364 talk about your bracket night. 0:20:35.524 --> 0:20:37.964 I would start with a clean slate, no offense to 0:20:38.004 --> 0:20:42.164 your ECLA Champion bracket, and I think first we need 0:20:42.164 --> 0:20:42.764 to set our. 0:20:42.644 --> 0:20:45.484 Goal, because that's going to dictate. You know, I'll try it. 0:20:46.124 --> 0:20:49.124 I'll try not too many tortured poker analogies, but I. 0:20:49.284 --> 0:20:53.444 No, let's do tortured poker analogies. We live for tortured 0:20:53.484 --> 0:20:54.844 poker analogies on the show. 0:20:54.964 --> 0:20:59.364 Yes, so head to head cash game versus a World 0:20:59.404 --> 0:21:02.004 Series of Poker tournament, you're going to do different things. 0:21:02.324 --> 0:21:04.124 I think in your case, really you want to be 0:21:04.324 --> 0:21:04.964 one person. 0:21:05.164 --> 0:21:08.084