1 00:00:01,280 --> 00:00:04,520 Speaker 1: Hey, welcome to Sign Stuff, a production of iHeartRadio I'm 2 00:00:04,519 --> 00:00:07,480 Speaker 1: More Hit Cham and today we're diving into the signs 3 00:00:07,520 --> 00:00:12,160 Speaker 1: of sports analytics. Can better mathematical bottles help your team win? 4 00:00:12,640 --> 00:00:15,760 Speaker 1: How exactly does it work? And is it ruining the 5 00:00:15,880 --> 00:00:18,840 Speaker 1: fun of the game for fans. We're going to be 6 00:00:18,840 --> 00:00:21,520 Speaker 1: talking to someone who ran the numbers for a Major 7 00:00:21,600 --> 00:00:26,040 Speaker 1: League Baseball team and who now publishes academically on the subject, 8 00:00:26,280 --> 00:00:28,120 Speaker 1: and he's going to step us through the history of 9 00:00:28,160 --> 00:00:31,960 Speaker 1: this phenomenon and how it's now spread into almost every sport, 10 00:00:32,320 --> 00:00:36,720 Speaker 1: including chess and video games. Now, I recorded part of 11 00:00:36,760 --> 00:00:40,040 Speaker 1: this episode during a visit to a sporting event, one 12 00:00:40,040 --> 00:00:43,520 Speaker 1: of the opening games for the LA Dodgers, who happen 13 00:00:43,600 --> 00:00:45,720 Speaker 1: to be the number one team in the world and 14 00:00:45,720 --> 00:00:48,640 Speaker 1: who have a whole staff of mathematicians working for them. 15 00:00:48,920 --> 00:00:52,120 Speaker 1: So gear up, lock in, and get ready to score 16 00:00:52,479 --> 00:00:56,200 Speaker 1: as we football tackle. The question is math We're winning 17 00:00:56,280 --> 00:01:10,280 Speaker 1: sports enjoy? Hey everyone, So I'm here at Dodger Stadium 18 00:01:10,360 --> 00:01:13,960 Speaker 1: for opening week to watch the LA Dodgers play the 19 00:01:14,160 --> 00:01:18,039 Speaker 1: Arizona Diamondbacks. Now, I'm not a huge baseball fan, but 20 00:01:18,400 --> 00:01:20,760 Speaker 1: from what I read, the Dodgers are favorite to win 21 00:01:21,000 --> 00:01:24,119 Speaker 1: according to the betting markets, with a probability of about 22 00:01:24,120 --> 00:01:28,040 Speaker 1: sixty seven percent, meaning that the Dodgers are favorite to win. 23 00:01:29,800 --> 00:01:32,280 Speaker 1: Of course, they have one of the best, if not 24 00:01:32,400 --> 00:01:41,160 Speaker 1: the best player ever. I'm talking, of course about show 25 00:01:41,200 --> 00:01:45,560 Speaker 1: Heyo Tani. Now, who actually determines that the Dodgers have 26 00:01:45,640 --> 00:01:48,200 Speaker 1: a sixty seven percent chance to win? How's that done? 27 00:01:48,320 --> 00:01:50,520 Speaker 1: Is that some guess or is there a lot of 28 00:01:50,560 --> 00:01:52,840 Speaker 1: maths behind it? That's what I want to find out. 29 00:01:53,200 --> 00:01:55,360 Speaker 1: But first I'm going to ask a few Dodger fans 30 00:01:55,360 --> 00:01:58,040 Speaker 1: here who they think are going to win. Who do 31 00:01:58,080 --> 00:02:00,800 Speaker 1: you think's gonna win? Dodgers or the Diamondbacks other Dodgers 32 00:02:00,800 --> 00:02:03,800 Speaker 1: for sure, Dodgers, Rogers. I think I'm sitting in a 33 00:02:03,880 --> 00:02:06,840 Speaker 1: Dodgers section. Can I ask you a question? Who do 34 00:02:06,840 --> 00:02:08,560 Speaker 1: you think is gonna win? Dodgers or Dynamits? 35 00:02:08,919 --> 00:02:11,440 Speaker 2: The Dodgers obviously, the Dodgers for sure. 36 00:02:12,000 --> 00:02:13,919 Speaker 1: This is kind of a silly question. Who do you 37 00:02:13,919 --> 00:02:14,560 Speaker 1: think is gonna win? 38 00:02:14,600 --> 00:02:16,160 Speaker 3: The Dodgers for the diamonback. 39 00:02:15,960 --> 00:02:18,079 Speaker 1: The Dodgers, of course, Dodgers. 40 00:02:19,600 --> 00:02:20,680 Speaker 2: And this is a silly question. 41 00:02:22,880 --> 00:02:25,840 Speaker 1: Okay, Clearly I wasn't going to get an unbiased opinion 42 00:02:25,840 --> 00:02:28,480 Speaker 1: here among the Dodger fans, so it makes sense of 43 00:02:28,520 --> 00:02:31,400 Speaker 1: all of this. I reached out to doctor Ben Baumber, 44 00:02:31,800 --> 00:02:34,960 Speaker 1: a professor of statistical and data scientists at Smith College 45 00:02:35,240 --> 00:02:38,880 Speaker 1: and the former statistical analyst for the New York Mets. 46 00:02:39,200 --> 00:02:49,960 Speaker 1: So here's my conversation with doctor Ben Bomber. Well, thank you, 47 00:02:50,000 --> 00:02:51,440 Speaker 1: doctor Bomber for joining us. 48 00:02:51,560 --> 00:02:53,400 Speaker 2: Thank you so much for having me. It's a pleasure 49 00:02:53,440 --> 00:02:53,839 Speaker 2: to be here. 50 00:02:54,280 --> 00:02:55,960 Speaker 1: You mentioned you played for the Mets. 51 00:02:56,080 --> 00:02:59,160 Speaker 2: It works for okay, did not play for so. 52 00:03:00,800 --> 00:03:02,560 Speaker 1: Well, sort of, you're part of the team. Come on, 53 00:03:03,880 --> 00:03:04,040 Speaker 1: you know. 54 00:03:04,120 --> 00:03:06,280 Speaker 2: I started working for the New York Mets in two 55 00:03:06,280 --> 00:03:09,200 Speaker 2: thousand and four as a statistical analyst, and at that 56 00:03:09,320 --> 00:03:11,800 Speaker 2: time they had never had one before. I was working 57 00:03:11,840 --> 00:03:14,040 Speaker 2: on PhD in mass so I was the only person 58 00:03:14,160 --> 00:03:16,920 Speaker 2: doing it kind of at that pretty high technical level. 59 00:03:17,000 --> 00:03:19,240 Speaker 2: And I was able for a long time to do 60 00:03:19,320 --> 00:03:20,560 Speaker 2: both at the same time. 61 00:03:20,880 --> 00:03:21,440 Speaker 1: Wow. 62 00:03:21,560 --> 00:03:23,760 Speaker 2: But at the end of that process decided to do 63 00:03:23,840 --> 00:03:27,880 Speaker 2: something else. Instead of doing sports analytics for the Mets. 64 00:03:27,919 --> 00:03:34,200 Speaker 2: I do sports analytics for academia, the public journals. It's 65 00:03:34,200 --> 00:03:36,520 Speaker 2: a different league, Yeah, a different league exactly. 66 00:03:36,760 --> 00:03:40,920 Speaker 1: Awesome. Can you please tell us what exactly is sports analytics? 67 00:03:41,000 --> 00:03:44,920 Speaker 2: Sports analytics is the use of statistics and data to 68 00:03:45,280 --> 00:03:48,440 Speaker 2: think about sports, like to learn about sports, how we 69 00:03:48,520 --> 00:03:51,160 Speaker 2: might play sports better or more efficiently, or who the 70 00:03:51,160 --> 00:03:53,480 Speaker 2: better players are or who the better teams are. But 71 00:03:53,760 --> 00:03:57,040 Speaker 2: I think for centuries people have been watching sports and 72 00:03:57,120 --> 00:03:59,839 Speaker 2: trying to answer those questions for themselves. But I think 73 00:04:00,080 --> 00:04:02,800 Speaker 2: has changed when you talk about sports analytics is we're 74 00:04:02,840 --> 00:04:06,200 Speaker 2: actually recording the data about what's happening in those games 75 00:04:06,560 --> 00:04:09,520 Speaker 2: and them we're analyzing that data in order to inform 76 00:04:09,560 --> 00:04:10,320 Speaker 2: those questions. 77 00:04:11,560 --> 00:04:13,800 Speaker 1: Is there a moment in history we can trace this 78 00:04:13,920 --> 00:04:18,280 Speaker 1: idea too, or what is the historical origins of this idea? 79 00:04:18,400 --> 00:04:21,799 Speaker 2: Yeah? So actually there is a fairly specific origin story. 80 00:04:21,839 --> 00:04:24,600 Speaker 2: So you know, people have been playing sports going back 81 00:04:24,600 --> 00:04:28,160 Speaker 2: to ancient Greece or whatever. But in the United States, 82 00:04:28,200 --> 00:04:32,599 Speaker 2: like in eighteen seventy, people started playing baseball professionally. But 83 00:04:32,800 --> 00:04:36,039 Speaker 2: just imagine, like there's people playing baseball and if you 84 00:04:36,120 --> 00:04:39,000 Speaker 2: want to watch a professional baseball game, you have to 85 00:04:39,120 --> 00:04:42,920 Speaker 2: go to the gate, right, there's no TV. So there 86 00:04:42,960 --> 00:04:45,960 Speaker 2: was a person, a man named Henry Chadwick, had this 87 00:04:46,080 --> 00:04:49,960 Speaker 2: idea of like what if I recorded some statistics about 88 00:04:50,000 --> 00:04:53,400 Speaker 2: the game and then published it in the newspaper so 89 00:04:53,440 --> 00:04:55,480 Speaker 2: that people like, not only do they get the score, 90 00:04:56,000 --> 00:05:00,920 Speaker 2: they got a numerical summary of what happened in the game. 91 00:05:01,080 --> 00:05:03,400 Speaker 2: And he called it the box score. We still have 92 00:05:03,520 --> 00:05:06,080 Speaker 2: these today. You can pick up you know, USA today 93 00:05:06,120 --> 00:05:08,320 Speaker 2: and look at the box scores for the baseball games. 94 00:05:08,240 --> 00:05:11,960 Speaker 1: Like how many runs, how many at that all of. 95 00:05:11,880 --> 00:05:14,279 Speaker 2: That, yeah, how many runs in total? And then for 96 00:05:14,320 --> 00:05:16,479 Speaker 2: each of the players in the batting order, how many 97 00:05:16,480 --> 00:05:18,240 Speaker 2: times did they come to bat, how many hits did 98 00:05:18,279 --> 00:05:20,560 Speaker 2: they get, how many runs did they drive in? For 99 00:05:20,640 --> 00:05:23,000 Speaker 2: the pitchers, how many innings did they pitch, how many 100 00:05:23,000 --> 00:05:25,120 Speaker 2: strikeouts did they have, how many runs did they give up? 101 00:05:25,720 --> 00:05:27,840 Speaker 2: And so for those of us who are like deep 102 00:05:27,880 --> 00:05:30,040 Speaker 2: in the weeds of baseball, like you can look at 103 00:05:30,040 --> 00:05:33,200 Speaker 2: a box score and basically reconstruct the entire game, like 104 00:05:33,480 --> 00:05:35,560 Speaker 2: two men on and two out or whatever, and this 105 00:05:35,600 --> 00:05:37,760 Speaker 2: person grounded into a double play and that end of 106 00:05:37,800 --> 00:05:38,599 Speaker 2: the inning, and then. 107 00:05:38,480 --> 00:05:40,960 Speaker 1: It is enough to sort of reconstruct some of the 108 00:05:41,080 --> 00:05:44,359 Speaker 1: drama of the sport. Absolutely absolutely, and they started with 109 00:05:44,440 --> 00:05:47,680 Speaker 1: baseball because I guess baseball was the as they say, 110 00:05:47,760 --> 00:05:50,839 Speaker 1: national past time. And absolutely so they started keep drugging 111 00:05:50,880 --> 00:05:51,560 Speaker 1: the box scores. 112 00:05:51,680 --> 00:05:54,840 Speaker 2: Right again, this was happening in the eighteen seventies. 113 00:05:55,880 --> 00:05:58,800 Speaker 1: So for the first time in known history, people started 114 00:05:58,800 --> 00:06:02,560 Speaker 1: to record data about sports events, mostly so fans could 115 00:06:02,640 --> 00:06:05,640 Speaker 1: follow along and know what happened in the game. And 116 00:06:05,680 --> 00:06:09,280 Speaker 1: this went on for over seventy years until in nineteen 117 00:06:09,360 --> 00:06:14,279 Speaker 1: forty a man named branch Ricky changed baseball and really 118 00:06:14,760 --> 00:06:16,080 Speaker 1: the world history. 119 00:06:17,920 --> 00:06:20,360 Speaker 2: So Bran Trickey, who was the general manager of the 120 00:06:20,400 --> 00:06:24,359 Speaker 2: Brooklyn Dodgers, made like two of the more important and 121 00:06:24,440 --> 00:06:27,400 Speaker 2: long lasting contributions to the way that baseball was played 122 00:06:27,400 --> 00:06:30,760 Speaker 2: in the United States. One was he signed Jackie Robinson, 123 00:06:31,080 --> 00:06:34,039 Speaker 2: and that broke the collar barrier major league baseball. 124 00:06:34,200 --> 00:06:34,880 Speaker 1: Wow. 125 00:06:35,000 --> 00:06:37,480 Speaker 2: The second was that he hired a man named Alan 126 00:06:37,600 --> 00:06:41,320 Speaker 2: Roth to be the first full time statistician to work 127 00:06:41,360 --> 00:06:44,560 Speaker 2: for a Major League baseball team. Really yeah, And there's 128 00:06:44,600 --> 00:06:48,880 Speaker 2: a great article. Life Magazine did a whole spread about 129 00:06:49,040 --> 00:06:51,040 Speaker 2: Alan Roth and what he was doing with the Brooklyn 130 00:06:51,080 --> 00:06:54,600 Speaker 2: Dodgers and all these equations written on the blackboard behind him. 131 00:06:55,279 --> 00:06:58,040 Speaker 1: What do you think with the mentality of that first 132 00:06:58,120 --> 00:07:01,520 Speaker 1: baseball owner who hired that that stition. I mean, obviously 133 00:07:01,560 --> 00:07:04,000 Speaker 1: he seemed to be sort of a groundbreaking type of 134 00:07:04,080 --> 00:07:07,200 Speaker 1: thinking person, you know, to hire Jackie Robinson, But what 135 00:07:07,320 --> 00:07:08,920 Speaker 1: do you think he was thinking at the time when 136 00:07:09,040 --> 00:07:10,880 Speaker 1: he hired the statistician, How do. 137 00:07:10,880 --> 00:07:14,040 Speaker 2: I win more games? You know, it's really pretty simple 138 00:07:14,280 --> 00:07:17,120 Speaker 2: because fundamentally it's just about like how do we do 139 00:07:17,200 --> 00:07:20,440 Speaker 2: this better? And in sports you have very clear outcomes. 140 00:07:20,480 --> 00:07:23,840 Speaker 2: It's wins and losses. And you know, brand Shreck he believed, 141 00:07:23,920 --> 00:07:26,560 Speaker 2: I'm sure he was correct at that time that like, 142 00:07:26,920 --> 00:07:28,880 Speaker 2: one way that I can win more games is by 143 00:07:28,960 --> 00:07:32,520 Speaker 2: understanding how baseball works better so that I can find 144 00:07:32,640 --> 00:07:36,960 Speaker 2: players who do things that help us win, especially when 145 00:07:37,200 --> 00:07:40,000 Speaker 2: those things are maybe overlooked by these other teams who 146 00:07:40,040 --> 00:07:43,000 Speaker 2: don't have this knowledge that I have acquired or developed. 147 00:07:43,320 --> 00:07:46,440 Speaker 1: Oh I see. He maybe asked like should I hire 148 00:07:46,440 --> 00:07:48,200 Speaker 1: this player or that player, or should we do these kinds 149 00:07:48,200 --> 00:07:50,080 Speaker 1: of places or that kind of play, And somebody told 150 00:07:50,120 --> 00:07:51,760 Speaker 1: him what to do, and he said, no, I don't 151 00:07:51,760 --> 00:07:54,600 Speaker 1: believe you, like, show me the data exactly. 152 00:07:54,920 --> 00:07:57,200 Speaker 2: And so just to give you a simple example, if 153 00:07:57,240 --> 00:07:59,840 Speaker 2: you've got to run around first base and second base 154 00:07:59,880 --> 00:08:02,200 Speaker 2: is open, the question is, you know, how do you 155 00:08:02,280 --> 00:08:06,720 Speaker 2: get that runner on first to second. One strategy is 156 00:08:07,160 --> 00:08:10,320 Speaker 2: to bond, which means they don't swing at the ball 157 00:08:10,320 --> 00:08:12,200 Speaker 2: and Triit just kind of hold the bat there and 158 00:08:12,240 --> 00:08:14,480 Speaker 2: try to like have the ball like fall kind of 159 00:08:14,480 --> 00:08:16,040 Speaker 2: fleck in front of the catcher. 160 00:08:16,200 --> 00:08:18,240 Speaker 1: Like they just tapped the ball. And it's almost like 161 00:08:18,280 --> 00:08:20,320 Speaker 1: a sacrifice play, right exactly. 162 00:08:20,400 --> 00:08:24,480 Speaker 2: It's called a sacrifice bunt. And so in some ways 163 00:08:24,560 --> 00:08:28,160 Speaker 2: you've gained something because you moved that runner across, but 164 00:08:28,280 --> 00:08:30,720 Speaker 2: in other ways you've lost something because now an out 165 00:08:30,760 --> 00:08:31,760 Speaker 2: has occurred. 166 00:08:31,840 --> 00:08:34,160 Speaker 1: Right right? Is it worth it? Is it actually a 167 00:08:34,200 --> 00:08:35,520 Speaker 1: good idea? Is the question? 168 00:08:35,920 --> 00:08:40,640 Speaker 2: Exactly? And so up until Alan Roth, basically people had 169 00:08:40,679 --> 00:08:43,160 Speaker 2: been trying to keep track of like, well, how does 170 00:08:43,160 --> 00:08:43,480 Speaker 2: it work? 171 00:08:43,640 --> 00:08:46,600 Speaker 1: I don't all, like, I've been watching baseball for twenty years, 172 00:08:46,640 --> 00:08:49,440 Speaker 1: and that's always a bad idea. 173 00:08:48,880 --> 00:08:52,640 Speaker 2: Exactly anecdotal evidence, you know. So because Henry Tradwick and 174 00:08:52,679 --> 00:08:55,520 Speaker 2: other people had, you know, been collecting data about baseball 175 00:08:55,640 --> 00:08:59,120 Speaker 2: for at that point already fifty or seventy years or whatever, 176 00:08:59,280 --> 00:09:02,600 Speaker 2: but we did thousands of games, people started to sort 177 00:09:02,640 --> 00:09:05,400 Speaker 2: of pull it apart and be like, does this actually 178 00:09:05,400 --> 00:09:07,960 Speaker 2: pay off? Do we actually score more runs if we 179 00:09:08,000 --> 00:09:11,160 Speaker 2: do this relative to the times that we don't do this. 180 00:09:11,880 --> 00:09:14,560 Speaker 2: So this was the type of analysis that people and. 181 00:09:14,600 --> 00:09:17,360 Speaker 1: Did that work. Did People were like, oh, my good 182 00:09:17,559 --> 00:09:18,160 Speaker 1: after that. 183 00:09:18,600 --> 00:09:21,160 Speaker 2: I mean, the Dodgers won a lot of games. Yeah, 184 00:09:21,320 --> 00:09:24,800 Speaker 2: that's a good research question actually, But certainly the Dodgers 185 00:09:24,800 --> 00:09:26,959 Speaker 2: were a very good team through the late forties and 186 00:09:27,320 --> 00:09:30,320 Speaker 2: through the fifties. You know, fifty five the Dodgers finally 187 00:09:30,320 --> 00:09:33,000 Speaker 2: beat the Yankees, and yeah, the Dodgers today are the 188 00:09:33,200 --> 00:09:35,319 Speaker 2: dominant team in Major League Baseball for sure. 189 00:09:35,760 --> 00:09:39,600 Speaker 1: Yeah. Yeah, so it worked back then. Did that cause 190 00:09:39,920 --> 00:09:42,200 Speaker 1: other teams to start looking at stats? Also? 191 00:09:42,600 --> 00:09:45,040 Speaker 2: I think the short answer is it doesn't appear to 192 00:09:45,120 --> 00:09:48,840 Speaker 2: be the case. I see, there's not a lot of 193 00:09:49,040 --> 00:09:53,880 Speaker 2: historical record for people being like full time employees statistical 194 00:09:53,960 --> 00:09:56,960 Speaker 2: analysts for major League Baseball teams. Not so much between 195 00:09:57,000 --> 00:09:59,520 Speaker 2: the forties and the eighties, nineties, two thousands. 196 00:09:59,600 --> 00:10:04,800 Speaker 1: Yeah, yes, despite the then Brooklyn Dodgers having a math 197 00:10:04,920 --> 00:10:08,120 Speaker 1: person on the team and having a winning streak, people 198 00:10:08,160 --> 00:10:11,679 Speaker 1: in sports were still not convinced. We'll get into what 199 00:10:11,880 --> 00:10:14,319 Speaker 1: that was a little later in the program. But all 200 00:10:14,360 --> 00:10:18,679 Speaker 1: of that changed with the publication of a book called Moneyball. 201 00:10:19,200 --> 00:10:21,320 Speaker 1: You might have heard of it or maybe seeing the 202 00:10:21,400 --> 00:10:24,440 Speaker 1: movie based on it, starring Brad Pitt. It's about a 203 00:10:24,440 --> 00:10:28,120 Speaker 1: struggling baseball team. The two thousand and two Oakland A's 204 00:10:28,640 --> 00:10:31,880 Speaker 1: who were able to get to the World Series playoffs 205 00:10:31,920 --> 00:10:35,280 Speaker 1: despite having a third of the money that other larger 206 00:10:35,400 --> 00:10:38,240 Speaker 1: teams had, and they did it by you guessed it, 207 00:10:38,720 --> 00:10:42,640 Speaker 1: using math. It all started with an amateur baseball fan 208 00:10:42,800 --> 00:10:45,480 Speaker 1: in Kansas named Bill James. 209 00:10:46,840 --> 00:10:49,839 Speaker 2: Bill James is known as kind of like the godfather 210 00:10:50,000 --> 00:10:53,240 Speaker 2: of baseball analytics. Uh huh. Literally a night watchman out 211 00:10:53,240 --> 00:10:56,040 Speaker 2: a pork and beans factory in Kansas, uh huh. You 212 00:10:56,040 --> 00:10:59,160 Speaker 2: know it's a very smart guy, loved baseball, but somehow 213 00:10:59,440 --> 00:11:02,920 Speaker 2: he was the person who probably did the most to 214 00:11:03,160 --> 00:11:08,600 Speaker 2: popularize the ideas in baseball analytics. So he started looking 215 00:11:08,600 --> 00:11:11,240 Speaker 2: at data, and he started writing about it, and then 216 00:11:11,320 --> 00:11:15,920 Speaker 2: he started publishing these newsletters that contained like tons of ideas, 217 00:11:16,080 --> 00:11:19,400 Speaker 2: some of which became these kind of revolutionary ideas like 218 00:11:19,480 --> 00:11:23,720 Speaker 2: what it's not just the data being valuable for itself 219 00:11:24,000 --> 00:11:26,840 Speaker 2: or like in its own right, It's that the data 220 00:11:26,960 --> 00:11:30,480 Speaker 2: is a mechanism through which these very creative, intelligent people 221 00:11:30,640 --> 00:11:34,439 Speaker 2: were like learning new things about the game. It's those ideas, 222 00:11:34,440 --> 00:11:37,000 Speaker 2: like that's what led to moneyball. It wasn't like Billy 223 00:11:37,040 --> 00:11:38,640 Speaker 2: being woke up one day and was like, hey, we 224 00:11:38,679 --> 00:11:40,840 Speaker 2: should look at data. It's like he read Bill James, 225 00:11:40,840 --> 00:11:43,520 Speaker 2: and like Bill James was the one who was showing 226 00:11:43,600 --> 00:11:46,120 Speaker 2: sort of like what this data could be, like how 227 00:11:46,120 --> 00:11:48,120 Speaker 2: it could inform your understanding of the game. 228 00:11:49,160 --> 00:11:52,680 Speaker 1: Okay, here's one example of a Bill James idea. For 229 00:11:52,800 --> 00:11:56,600 Speaker 1: most of baseball history, people cared about a batters RBI 230 00:11:57,240 --> 00:12:00,439 Speaker 1: or runs batted in. It's a measure of how often 231 00:12:00,440 --> 00:12:03,600 Speaker 1: a team scores a point whenever a batter goes up 232 00:12:03,679 --> 00:12:05,960 Speaker 1: to hit the ball. If you look up the list 233 00:12:06,000 --> 00:12:09,320 Speaker 1: of the people with the top career RBI numbers, you'll 234 00:12:09,320 --> 00:12:13,520 Speaker 1: see names like Hank Aaron, Babe Ruth, Alex Rodriguez, or 235 00:12:13,640 --> 00:12:17,559 Speaker 1: a rod Ty Cobb. You know, legends. And so the 236 00:12:17,640 --> 00:12:21,200 Speaker 1: higher your RBI, the better people thought you were, and 237 00:12:21,240 --> 00:12:23,880 Speaker 1: the more money someone would have to pay you to 238 00:12:23,960 --> 00:12:27,800 Speaker 1: be on their team. But actually, Bill James and then 239 00:12:27,840 --> 00:12:30,760 Speaker 1: the Oakland A's figure it out that's not the best 240 00:12:30,800 --> 00:12:34,440 Speaker 1: statistic to be focusing on because it kind of depends 241 00:12:34,520 --> 00:12:37,800 Speaker 1: on your teammates. If you have good teammates that got 242 00:12:37,840 --> 00:12:39,880 Speaker 1: on base by the time you went to bad, you're 243 00:12:39,920 --> 00:12:43,080 Speaker 1: going to have a higher RBI. But looking at the 244 00:12:43,160 --> 00:12:46,280 Speaker 1: data more closely, it turns out a better measure of 245 00:12:46,320 --> 00:12:50,199 Speaker 1: how good a batter is is something called runs created, 246 00:12:50,480 --> 00:12:53,840 Speaker 1: which is computed using a different formula that doesn't depend 247 00:12:53,840 --> 00:12:57,600 Speaker 1: as much on what your teammates do. And because nobody 248 00:12:57,640 --> 00:13:00,480 Speaker 1: else was looking at the statistic, the oakland As were 249 00:13:00,480 --> 00:13:05,800 Speaker 1: able to get a good team for less money. And 250 00:13:05,880 --> 00:13:08,640 Speaker 1: it really was then Moneyball, which is based on the 251 00:13:08,679 --> 00:13:11,760 Speaker 1: work in the Oakland A's that really got people thinking like, 252 00:13:11,960 --> 00:13:13,880 Speaker 1: oh my goodness, we should totally do this. 253 00:13:14,360 --> 00:13:18,400 Speaker 2: Yeah. Things moved quickly after the publication Moneyball around the 254 00:13:18,440 --> 00:13:21,720 Speaker 2: time early aughts mid ots, when I was working for 255 00:13:21,760 --> 00:13:24,520 Speaker 2: the Mets, you know, that's when things spread throughout baseball. 256 00:13:24,679 --> 00:13:27,000 Speaker 2: So two thousand and three, two thousand and four, I 257 00:13:27,080 --> 00:13:30,959 Speaker 2: hamd full of teams have some person who's doing statistical 258 00:13:31,000 --> 00:13:34,680 Speaker 2: analysis full time. Most teams don't. But by the time 259 00:13:34,720 --> 00:13:38,400 Speaker 2: I left in twenty twelve, more than half the teams, 260 00:13:38,400 --> 00:13:41,120 Speaker 2: maybe three quarters of the teams had at least somebody 261 00:13:41,240 --> 00:13:46,000 Speaker 2: doing something. And now everybody and the Dodgers have like 262 00:13:46,240 --> 00:13:50,040 Speaker 2: a thirty person analytics staff, you know, with like multiple 263 00:13:50,080 --> 00:13:52,600 Speaker 2: people with PhDs and statistics and stuff like that. 264 00:13:52,640 --> 00:13:56,439 Speaker 1: So what thirty person staff just looking at the numbers? 265 00:13:56,559 --> 00:13:57,439 Speaker 2: Yeah? 266 00:13:57,520 --> 00:14:01,440 Speaker 1: Wow? And part of the story here is that amount 267 00:14:01,480 --> 00:14:04,880 Speaker 1: of data that is tracked in sports has also exploded 268 00:14:05,000 --> 00:14:08,120 Speaker 1: in the last ten years. It started with fans keeping 269 00:14:08,160 --> 00:14:11,880 Speaker 1: track of more detailed baseball statistics like play by play data, 270 00:14:12,200 --> 00:14:14,360 Speaker 1: who was on base, when someone hit a home run, 271 00:14:14,400 --> 00:14:17,520 Speaker 1: and where was the ball hit. For many years, this 272 00:14:17,679 --> 00:14:22,320 Speaker 1: was published in books and eventually websites for basically baseball nerds. 273 00:14:22,680 --> 00:14:26,160 Speaker 1: But now it's a full blown industry. There are companies 274 00:14:26,200 --> 00:14:28,760 Speaker 1: that gather data about games and then sell it to 275 00:14:28,800 --> 00:14:32,080 Speaker 1: the teams for their analytics, and it's getting more and 276 00:14:32,200 --> 00:14:36,120 Speaker 1: more high tech. The LA Dodgers, for example, have cameras 277 00:14:36,160 --> 00:14:38,800 Speaker 1: that tract the movement of every player on the field 278 00:14:38,880 --> 00:14:41,840 Speaker 1: during every game, so they have data now on how 279 00:14:41,880 --> 00:14:44,280 Speaker 1: far players were when they made a good catch, or 280 00:14:44,480 --> 00:14:47,720 Speaker 1: where the shortstop should stand to have the optimal chance 281 00:14:47,800 --> 00:14:52,320 Speaker 1: of making a double play. Nowadays, teams can even buy 282 00:14:52,520 --> 00:14:57,320 Speaker 1: biomechanical data about their players, how their bodies move at 283 00:14:57,400 --> 00:14:59,920 Speaker 1: three hundred times per second when they throw a pitch, 284 00:15:00,080 --> 00:15:04,520 Speaker 1: sure swing their bats during live games. Also, they can 285 00:15:04,560 --> 00:15:08,160 Speaker 1: figure out how their players can play better. But here's 286 00:15:08,160 --> 00:15:12,080 Speaker 1: the question. Does all of this actually work? Is it 287 00:15:12,160 --> 00:15:15,680 Speaker 1: actually making teams better or is it that some people 288 00:15:15,720 --> 00:15:20,320 Speaker 1: say ruining the sport. When we come back, we'll answer 289 00:15:20,320 --> 00:15:23,080 Speaker 1: that question and we'll talk about how this push for 290 00:15:23,200 --> 00:15:29,280 Speaker 1: more math has spread to other sports like professional basketball, football, 291 00:15:29,520 --> 00:15:52,560 Speaker 1: and even chess. So stay with us. We'll be right back. Hey, 292 00:15:52,600 --> 00:15:56,520 Speaker 1: welcome back. So we're talking about sports analytics or how 293 00:15:56,600 --> 00:15:59,360 Speaker 1: math is being used in sports. And I'm recording this 294 00:15:59,400 --> 00:16:06,640 Speaker 1: from Dodger Stadium. It's the bottom of the fifth inning, 295 00:16:07,120 --> 00:16:10,360 Speaker 1: so about halfway through the game, and the score is tied. 296 00:16:10,720 --> 00:16:13,600 Speaker 1: The Diamondbacks took an early lead in the game, but 297 00:16:13,680 --> 00:16:16,400 Speaker 1: then the Dodgers had an amazing third inning with two 298 00:16:16,440 --> 00:16:20,280 Speaker 1: home runs with people on base, but then the Diamondbacks 299 00:16:20,280 --> 00:16:23,720 Speaker 1: caught up on the fourth inning. So it's a closed game, 300 00:16:24,160 --> 00:16:27,200 Speaker 1: and the under fans around me are not as confident 301 00:16:27,240 --> 00:16:31,680 Speaker 1: as they were earlier. Okay, so far, we've talked a 302 00:16:31,680 --> 00:16:34,560 Speaker 1: little bit about the history of sports analytics, how it 303 00:16:34,640 --> 00:16:37,520 Speaker 1: started in baseball, and now we're going to talk about 304 00:16:37,520 --> 00:16:41,920 Speaker 1: this idea of using statistics and math to play better 305 00:16:41,960 --> 00:16:45,880 Speaker 1: at games has spread to other sports, including, if you 306 00:16:45,960 --> 00:16:56,480 Speaker 1: believe it, chess and the sports. Here's doctor Ben Bauer. Okay, 307 00:16:56,640 --> 00:16:59,800 Speaker 1: so you're saying now it's pervasive in baseball, would you 308 00:16:59,840 --> 00:17:02,760 Speaker 1: say every team now has a sort of a statistics team. 309 00:17:03,000 --> 00:17:06,640 Speaker 2: Yeah, for sure. Every team has a dedicated analytics presence 310 00:17:06,720 --> 00:17:07,399 Speaker 2: for sure. 311 00:17:07,359 --> 00:17:10,120 Speaker 1: And just for baseball, does it work, Like it must 312 00:17:10,200 --> 00:17:12,040 Speaker 1: be worth it for them to hire thirty people. 313 00:17:12,240 --> 00:17:14,760 Speaker 2: So one thing to keep in mind is that statistical 314 00:17:14,760 --> 00:17:18,040 Speaker 2: analysts make a tiny fraction of what players make, right, 315 00:17:18,359 --> 00:17:21,720 Speaker 2: And so it's like, if you can make better decisions 316 00:17:22,400 --> 00:17:26,280 Speaker 2: about even one player, well, like you can pay a 317 00:17:26,280 --> 00:17:28,920 Speaker 2: lot of analysts to help you make that better decision, right. 318 00:17:31,000 --> 00:17:34,840 Speaker 1: I think what you're saying is that statisticians need agents, Yeah, 319 00:17:35,040 --> 00:17:38,560 Speaker 1: and I think we do. They seem very valuable exactly 320 00:17:39,200 --> 00:17:42,200 Speaker 1: free agency. But then the other part is you're working 321 00:17:42,200 --> 00:17:45,640 Speaker 1: in the zero sum system, right, because games are either 322 00:17:45,680 --> 00:17:48,720 Speaker 1: one or lost. So if my team gets better at 323 00:17:48,760 --> 00:17:51,520 Speaker 1: analytics and then we play better baseball, like, we're going 324 00:17:51,560 --> 00:17:54,000 Speaker 1: to win more games in the short term, But then 325 00:17:54,080 --> 00:17:56,400 Speaker 1: eventually other teams are going to figure out what we're 326 00:17:56,440 --> 00:17:58,679 Speaker 1: doing and they're going to catch up to us, and 327 00:17:58,720 --> 00:18:00,960 Speaker 1: so you have this kind of like cat and mouse game. 328 00:18:01,760 --> 00:18:05,520 Speaker 1: And you know, so when you say, like does analytics work, yes, 329 00:18:05,600 --> 00:18:08,679 Speaker 1: it works, but like there is an aspect of like 330 00:18:08,880 --> 00:18:11,520 Speaker 1: what works only works for a short period of time 331 00:18:11,720 --> 00:18:15,600 Speaker 1: before everyone else catches on, and then you have to 332 00:18:15,920 --> 00:18:18,160 Speaker 1: like figure out the next thing I see, And at 333 00:18:18,160 --> 00:18:21,480 Speaker 1: that point you're kind of locked into having a statistic 334 00:18:21,520 --> 00:18:24,159 Speaker 1: team because if you didn't, then you would fall behind 335 00:18:24,240 --> 00:18:24,840 Speaker 1: everyone else. 336 00:18:25,040 --> 00:18:27,760 Speaker 2: That's exactly right, and that's certainly what happened through the 337 00:18:27,800 --> 00:18:29,920 Speaker 2: two thousands and the two thousand tens. Wow. 338 00:18:29,960 --> 00:18:32,480 Speaker 1: Yeah, okay, so that's baseball. But now it's gone on 339 00:18:32,560 --> 00:18:35,200 Speaker 1: to other sports after the success in baseball. 340 00:18:35,320 --> 00:18:38,399 Speaker 2: Absolutely, yeah, I would say the spread is not as 341 00:18:38,640 --> 00:18:41,480 Speaker 2: wide or as deep as it is in baseball, in 342 00:18:41,560 --> 00:18:46,280 Speaker 2: part because baseball is kind of fundamentally different than other 343 00:18:46,359 --> 00:18:50,520 Speaker 2: sports in the way that baseball has very discrete actions. 344 00:18:50,760 --> 00:18:54,560 Speaker 2: So it's just like the nature of the game is different. However, 345 00:18:54,960 --> 00:18:58,440 Speaker 2: yes it has spread to other sports, and yes NBA teams, 346 00:18:58,600 --> 00:19:01,719 Speaker 2: NFL teams, and NHL teams in soccer leagues in Europe, 347 00:19:01,720 --> 00:19:05,560 Speaker 2: and yes there are people doing statistical analysis for all 348 00:19:05,600 --> 00:19:09,240 Speaker 2: those teams at some level. And so in basketball you 349 00:19:09,359 --> 00:19:12,520 Speaker 2: have seen things like the way the teams are shooting 350 00:19:12,560 --> 00:19:15,440 Speaker 2: three pointers these days. I mean, when I was growing up, 351 00:19:15,880 --> 00:19:18,960 Speaker 2: our whole offensive strategy was let's get the ball into 352 00:19:19,000 --> 00:19:21,760 Speaker 2: the posts so that the tall players can put it 353 00:19:21,800 --> 00:19:24,600 Speaker 2: in the basket, because that's the best way for us 354 00:19:24,640 --> 00:19:25,720 Speaker 2: to score points. Right. 355 00:19:25,920 --> 00:19:26,040 Speaker 1: Uh. 356 00:19:26,840 --> 00:19:30,960 Speaker 2: Now, what they're doing, and this is definitely through analytics, right, 357 00:19:31,160 --> 00:19:33,800 Speaker 2: is that they realize that, well, okay, if we get 358 00:19:33,800 --> 00:19:36,000 Speaker 2: the ball down clost to the basket and somebody puts 359 00:19:36,000 --> 00:19:38,359 Speaker 2: it in, what's the expected value of that shot? 360 00:19:38,560 --> 00:19:39,240 Speaker 1: Two points? 361 00:19:39,280 --> 00:19:41,720 Speaker 2: It's two points. And let's say if it's close to 362 00:19:41,760 --> 00:19:43,600 Speaker 2: the basket, like I'm going to miss a couple, but 363 00:19:43,680 --> 00:19:46,360 Speaker 2: like I basically never missed, so like maybe ninety five 364 00:19:46,400 --> 00:19:49,399 Speaker 2: percent of those I make, So that's one point nine 365 00:19:49,640 --> 00:19:53,240 Speaker 2: expected points, right, Okay, But now if I shoot a 366 00:19:53,280 --> 00:19:56,479 Speaker 2: three pointer, it's for three points, so I only have 367 00:19:56,560 --> 00:19:59,160 Speaker 2: to make you know, like I f ix seventy percent, 368 00:19:59,440 --> 00:20:01,360 Speaker 2: that's two point one expected points. 369 00:20:01,560 --> 00:20:02,400 Speaker 1: Uh huh. 370 00:20:02,480 --> 00:20:06,040 Speaker 2: And these guys can make if they're open, nobody's there. 371 00:20:06,160 --> 00:20:09,000 Speaker 2: You know, they're making eighty percent of us. So a 372 00:20:09,040 --> 00:20:13,520 Speaker 2: three pointer becomes a much more attractive shot if the 373 00:20:13,600 --> 00:20:17,879 Speaker 2: probability of you're making that shot is sufficiently And so 374 00:20:17,960 --> 00:20:21,280 Speaker 2: that has absolutely changed the way that basketball's play. Teams 375 00:20:21,280 --> 00:20:24,680 Speaker 2: are now leaning more on three pointers absolutely, and that's 376 00:20:24,760 --> 00:20:27,359 Speaker 2: kind of like another one of those hidden moves that 377 00:20:27,520 --> 00:20:30,480 Speaker 2: was in the data, but nobody really believed because you know, 378 00:20:30,520 --> 00:20:34,680 Speaker 2: we want to see Michael Jordan's dunk the ball. Absolutely, 379 00:20:35,000 --> 00:20:38,200 Speaker 2: I think the shooting percentages have changed and that has 380 00:20:38,320 --> 00:20:41,320 Speaker 2: led to the evolution of the game that we see today. 381 00:20:41,480 --> 00:20:43,600 Speaker 1: But I wonder if it's sort of like an arms 382 00:20:43,720 --> 00:20:47,240 Speaker 1: race there too, because now if everybody is aiming for 383 00:20:47,280 --> 00:20:50,879 Speaker 1: three pointers, then everyone's going to adapt their defense to 384 00:20:50,960 --> 00:20:52,680 Speaker 1: block those three pointers for sure. 385 00:20:52,800 --> 00:20:55,000 Speaker 2: I mean, this is a big part of like in baseball, 386 00:20:55,119 --> 00:20:57,960 Speaker 2: how analytics is you know, ruining. 387 00:20:57,600 --> 00:21:00,600 Speaker 1: Baseball because like, if I go to baseball, I want 388 00:21:00,600 --> 00:21:03,320 Speaker 1: to see people hitting the ball and making dramatic plays, 389 00:21:03,400 --> 00:21:06,200 Speaker 1: not like, oh, you got walked. That's not as exciting. 390 00:21:05,920 --> 00:21:08,719 Speaker 2: Exactly, And that is more or less exactly. What has 391 00:21:08,760 --> 00:21:11,440 Speaker 2: happened over the last fifteen or twenty years is that 392 00:21:11,560 --> 00:21:14,119 Speaker 2: people like me and people who are doing jobs similar 393 00:21:14,160 --> 00:21:16,720 Speaker 2: to me sort of figured out, well, if we want 394 00:21:16,720 --> 00:21:19,480 Speaker 2: to win more games, like we need to draw more walks, 395 00:21:19,560 --> 00:21:21,800 Speaker 2: then we're not going to steal so many bases because 396 00:21:21,840 --> 00:21:24,520 Speaker 2: that turns out to be pretty risky. But it led 397 00:21:24,560 --> 00:21:27,520 Speaker 2: to a style of play that a lot of people, 398 00:21:27,560 --> 00:21:30,640 Speaker 2: including myself, find less interesting to. 399 00:21:30,600 --> 00:21:34,480 Speaker 1: Watch because it's prioritizing the long term goals that the game, 400 00:21:34,840 --> 00:21:36,879 Speaker 1: not the moment to moment excitement. 401 00:21:37,000 --> 00:21:42,320 Speaker 2: Absolutely, it's prioritizing winning the game, not making the game entertaining. 402 00:21:43,160 --> 00:21:45,080 Speaker 2: I mean, this is what people are talking about with 403 00:21:45,160 --> 00:21:48,000 Speaker 2: basketball and how it's the ruining basketball. They're talking about, 404 00:21:48,040 --> 00:21:49,600 Speaker 2: you know, instituting a four point. 405 00:21:49,359 --> 00:21:51,320 Speaker 1: Shot like from the half court kind of. 406 00:21:51,359 --> 00:21:56,280 Speaker 3: Yeah, something like that, Like basketball is going to become 407 00:21:56,359 --> 00:22:00,439 Speaker 3: people just standing around the middle of the court, basket 408 00:22:00,520 --> 00:22:01,879 Speaker 3: from the middle of the court. 409 00:22:02,600 --> 00:22:05,640 Speaker 1: Oh, I see, and that's not basketball. That's if you're 410 00:22:05,640 --> 00:22:08,880 Speaker 1: a fan, you'd be like, well, it's not basketball as 411 00:22:08,880 --> 00:22:17,919 Speaker 1: we know it. Could you say anything kind of about 412 00:22:17,920 --> 00:22:19,840 Speaker 1: how it has spread to other sports? 413 00:22:20,080 --> 00:22:23,520 Speaker 2: Yeah. One of my co authors on the paper that 414 00:22:23,560 --> 00:22:26,000 Speaker 2: we wrote a couple of years ago, Michael Lopez, was 415 00:22:26,080 --> 00:22:29,000 Speaker 2: hired by the National Football League a few years ago 416 00:22:29,080 --> 00:22:32,720 Speaker 2: to become their director of sports Analytics for the league. 417 00:22:33,040 --> 00:22:35,280 Speaker 2: And so I think you had baseball kind of like 418 00:22:35,359 --> 00:22:38,600 Speaker 2: leading the way. I think basketball came in, you know, 419 00:22:38,960 --> 00:22:41,800 Speaker 2: sort of after that, and I think the NFL husband 420 00:22:42,000 --> 00:22:42,680 Speaker 2: after that. 421 00:22:43,080 --> 00:22:44,879 Speaker 1: Okay, do you have any examples of it? 422 00:22:45,080 --> 00:22:47,320 Speaker 2: Yeah, Well, the thing that has attracted the most attention 423 00:22:47,560 --> 00:22:50,119 Speaker 2: is when to punt and when to go for it 424 00:22:50,160 --> 00:22:53,720 Speaker 2: on fourth down. So you know, in American football, yet 425 00:22:53,760 --> 00:22:57,200 Speaker 2: four downs to advance the ball ten yards, and if 426 00:22:57,200 --> 00:22:59,240 Speaker 2: you are able to do that, then you get another 427 00:23:00,800 --> 00:23:03,280 Speaker 2: to try again. But if you don't, it's the other 428 00:23:03,320 --> 00:23:07,280 Speaker 2: team's ball. And so what most football teams will do 429 00:23:07,520 --> 00:23:11,880 Speaker 2: is if it's fourth down and the ball is very close, 430 00:23:12,240 --> 00:23:13,960 Speaker 2: then like maybe they're going to try to get that 431 00:23:14,040 --> 00:23:16,639 Speaker 2: extra yard or two and keep going. But if not, 432 00:23:17,040 --> 00:23:19,920 Speaker 2: if it's like fourth down and eight yards to go, 433 00:23:20,359 --> 00:23:23,440 Speaker 2: then in their minds they're like, well, chances are we're 434 00:23:23,440 --> 00:23:25,800 Speaker 2: not going to make it. So if we give the 435 00:23:25,800 --> 00:23:28,679 Speaker 2: other team the ball here, they're going to be in 436 00:23:28,720 --> 00:23:31,800 Speaker 2: a position to maybe score against us. So what we're 437 00:23:31,840 --> 00:23:33,240 Speaker 2: going to do instead is we're going to kick the 438 00:23:33,280 --> 00:23:35,720 Speaker 2: ball as far as we can down the field, and 439 00:23:35,840 --> 00:23:38,440 Speaker 2: then even though we will have given the other team 440 00:23:38,520 --> 00:23:40,879 Speaker 2: the ball, we will have sent them all the way back, 441 00:23:41,000 --> 00:23:43,240 Speaker 2: you know, as far as we can. That's called a punt. 442 00:23:43,440 --> 00:23:46,760 Speaker 2: And then so the question becomes when should you go 443 00:23:46,800 --> 00:23:49,000 Speaker 2: for it on fourth down and when should you not? 444 00:23:49,920 --> 00:23:53,480 Speaker 2: And you know, so statistical analysts started looking at this, 445 00:23:53,760 --> 00:23:58,520 Speaker 2: and what they found was that generally teams were overly conservative, 446 00:23:58,880 --> 00:24:01,560 Speaker 2: that is, they did not go for it on fourth 447 00:24:01,600 --> 00:24:06,160 Speaker 2: down as often as the statistical analysis would suggest was optimal. 448 00:24:06,320 --> 00:24:08,560 Speaker 1: I see they said it chickened out, they're trying to 449 00:24:08,560 --> 00:24:08,960 Speaker 1: go for it. 450 00:24:09,000 --> 00:24:14,280 Speaker 2: Well, yes, so that's one of them prominent statistical analysis 451 00:24:14,359 --> 00:24:15,600 Speaker 2: contributions to football. 452 00:24:15,800 --> 00:24:18,400 Speaker 1: I see, general teams aren't going for it more, people 453 00:24:18,440 --> 00:24:19,720 Speaker 1: are putting it less. 454 00:24:19,880 --> 00:24:23,240 Speaker 2: Yes, there's been a movement definitely towards going for it more. 455 00:24:23,280 --> 00:24:26,080 Speaker 2: And now a lot of times they'll even have like 456 00:24:26,359 --> 00:24:29,800 Speaker 2: a little on screen graphic that's telling you, you know, 457 00:24:30,119 --> 00:24:32,239 Speaker 2: whether the team suppos to go for it or not. 458 00:24:32,560 --> 00:24:34,240 Speaker 1: Whoa, it's part of the broadcast. 459 00:24:34,320 --> 00:24:35,440 Speaker 2: Yeah, it's part of the broadcast. 460 00:24:35,600 --> 00:24:38,720 Speaker 1: Like there's an analytic team saying, oh, you know, every 461 00:24:38,720 --> 00:24:42,000 Speaker 1: time it's fourth and down between these two teams, they. 462 00:24:41,840 --> 00:24:43,280 Speaker 2: Should go for it exactly. 463 00:24:43,480 --> 00:24:46,000 Speaker 1: And in this case, it's sort of something fans could 464 00:24:46,040 --> 00:24:48,520 Speaker 1: agree on, right, Like fans want to see more people 465 00:24:48,640 --> 00:24:49,239 Speaker 1: going for it. 466 00:24:49,359 --> 00:24:51,920 Speaker 2: Yeah, that's a good point going forward on fourth down. 467 00:24:51,960 --> 00:24:55,360 Speaker 2: It's exciting, probably more exciting than punting on fourth down. 468 00:24:55,880 --> 00:24:58,720 Speaker 2: Don't remember exactly when this was, but sometime in the 469 00:24:58,840 --> 00:25:01,240 Speaker 2: Belichick era, but it was a big game between the 470 00:25:01,240 --> 00:25:04,679 Speaker 2: Patriots and the Colts and the Patriots had a fourth 471 00:25:04,720 --> 00:25:07,280 Speaker 2: and whatever, and they went for it and they didn't 472 00:25:07,280 --> 00:25:09,600 Speaker 2: get it, and Peyton Manning's team got the ball back 473 00:25:09,640 --> 00:25:13,600 Speaker 2: and then they went and scored. Oh you know, everybody 474 00:25:13,760 --> 00:25:18,399 Speaker 2: was talking about how stupid the Patriots were. And so 475 00:25:18,440 --> 00:25:20,840 Speaker 2: it was another one of these cases where a team 476 00:25:21,080 --> 00:25:25,639 Speaker 2: tried to pursue the analytically optimal strategy and a backfired 477 00:25:25,680 --> 00:25:28,520 Speaker 2: on them, and the sort of media and fan backlash 478 00:25:28,680 --> 00:25:31,920 Speaker 2: sort of overwhelmed all the times that maybe they did 479 00:25:31,960 --> 00:25:34,400 Speaker 2: go for it in other situations and made. 480 00:25:34,200 --> 00:25:38,040 Speaker 1: It it's like, you can't win with sports fans, can you, Yeah, right, 481 00:25:38,119 --> 00:25:41,639 Speaker 1: right right? And you said it spread into even things 482 00:25:41,680 --> 00:25:44,120 Speaker 1: like esports and chess. Yeah. 483 00:25:44,320 --> 00:25:47,720 Speaker 2: The concept of rating systems in chess very much goes 484 00:25:47,800 --> 00:25:51,560 Speaker 2: back many years. So things like ELO ratings were specifically 485 00:25:51,640 --> 00:25:52,840 Speaker 2: designed for chess. 486 00:25:53,040 --> 00:25:57,080 Speaker 1: Uh the quickest side here. An ELO rating, named after 487 00:25:57,119 --> 00:26:00,520 Speaker 1: the physicists and chess player arped Elo who invented it, 488 00:26:00,560 --> 00:26:03,960 Speaker 1: basically tells you how good you are at chess. A 489 00:26:04,040 --> 00:26:07,280 Speaker 1: lot of people play chess online these days, especially young people, 490 00:26:07,520 --> 00:26:09,840 Speaker 1: and so you might have had your kids or your 491 00:26:09,920 --> 00:26:13,760 Speaker 1: younger cousins talk about their ELO rating. For example, the 492 00:26:13,800 --> 00:26:16,639 Speaker 1: top chess players in the world have an ELO ranking 493 00:26:16,760 --> 00:26:20,040 Speaker 1: of about twenty eight hundred, whereas the beginner would start 494 00:26:20,080 --> 00:26:23,600 Speaker 1: it zero. But here's the thing. Your ELO rating is 495 00:26:23,640 --> 00:26:26,879 Speaker 1: not just where you stand relative to other players. The 496 00:26:27,000 --> 00:26:31,080 Speaker 1: formula for it actually tells you the probability of who's 497 00:26:31,119 --> 00:26:34,199 Speaker 1: going to win between two players. So you can use 498 00:26:34,240 --> 00:26:37,560 Speaker 1: it for say, figuring out exactly how much to bed 499 00:26:37,720 --> 00:26:40,800 Speaker 1: on a chess championship match, or how much money to 500 00:26:40,840 --> 00:26:44,119 Speaker 1: pay a chess player to be their sponsor. And yes, 501 00:26:44,240 --> 00:26:47,720 Speaker 1: these days chess players have sponsorship deals. But the same 502 00:26:47,760 --> 00:26:51,200 Speaker 1: idea is also used in other sports like tennis. 503 00:26:52,640 --> 00:26:55,600 Speaker 2: Like if you think about tennis versus chess from an 504 00:26:55,640 --> 00:26:59,240 Speaker 2: analytical player rating system perspective, like the same thing, right, 505 00:26:59,320 --> 00:27:02,160 Speaker 2: it's I see this person plays this person. They each 506 00:27:02,200 --> 00:27:04,760 Speaker 2: have a rating before the match, and somebody wins the match, 507 00:27:04,800 --> 00:27:06,439 Speaker 2: and then they each have a rating after the match. 508 00:27:06,840 --> 00:27:10,000 Speaker 2: What are the ramifications of the particular modeling choices that 509 00:27:10,040 --> 00:27:12,520 Speaker 2: we make when we use those rating systems? 510 00:27:12,880 --> 00:27:15,359 Speaker 1: I see? And esports? What do you know about the 511 00:27:15,440 --> 00:27:19,080 Speaker 1: using esport? Another quickest site here in case you didn't know. 512 00:27:19,480 --> 00:27:24,080 Speaker 1: Esports or electronic sports are a thing that's where people 513 00:27:24,119 --> 00:27:28,679 Speaker 1: compete on video games. League of Legends, Valoriant, counter Strike. 514 00:27:29,080 --> 00:27:33,000 Speaker 1: These are hugely popular competitions, with millions of viewers and 515 00:27:33,440 --> 00:27:35,720 Speaker 1: billions of dollars in prize money. 516 00:27:36,640 --> 00:27:39,760 Speaker 2: I think what's really interesting about esports is that in 517 00:27:39,880 --> 00:27:42,199 Speaker 2: all the things that we've talked about so far, the 518 00:27:42,240 --> 00:27:45,480 Speaker 2: game exists and then we have decided to collect data 519 00:27:45,520 --> 00:27:49,160 Speaker 2: about it. But in esports, that data is like inherently 520 00:27:49,320 --> 00:27:52,240 Speaker 2: part of the game itself because the games are happening 521 00:27:52,280 --> 00:27:56,200 Speaker 2: on the computer. Ah, so the computer is already keeping 522 00:27:56,359 --> 00:27:59,080 Speaker 2: track of all the things. You know, who's killing whom 523 00:27:59,119 --> 00:28:02,760 Speaker 2: and who's this portion of the board and how many WHOA. 524 00:28:02,800 --> 00:28:07,359 Speaker 1: It's built into the game where the game itself is 525 00:28:07,480 --> 00:28:09,600 Speaker 1: tracking every possible statistic. 526 00:28:09,920 --> 00:28:12,600 Speaker 2: That's how it works. I'm not much of any sports 527 00:28:12,720 --> 00:28:15,280 Speaker 2: participant in myself, but the papers that I read on 528 00:28:15,320 --> 00:28:18,040 Speaker 2: this subject a little better couple. Their approach is sort 529 00:28:18,080 --> 00:28:20,399 Speaker 2: of similar to what we've talked about with bisball or 530 00:28:20,440 --> 00:28:22,080 Speaker 2: any of these parts. It's sort of like what are 531 00:28:22,080 --> 00:28:25,520 Speaker 2: the best strategies for winning the game? And we've got 532 00:28:25,640 --> 00:28:29,400 Speaker 2: now millions and millions of games that people are playing 533 00:28:29,640 --> 00:28:32,800 Speaker 2: all the time, and we can analyze that data to 534 00:28:32,880 --> 00:28:35,320 Speaker 2: figure out, like, what are the strategies that pay off 535 00:28:35,359 --> 00:28:38,440 Speaker 2: for people. You know, which of these strategies tends to 536 00:28:38,720 --> 00:28:41,240 Speaker 2: produce the most kills or produce the most wills? 537 00:28:41,320 --> 00:28:44,320 Speaker 1: And you know, gosh, I don't know about you, but 538 00:28:44,560 --> 00:28:46,840 Speaker 1: I can't wait for the sequel to the movie Moneyball 539 00:28:47,200 --> 00:28:49,760 Speaker 1: where it's just Brad Pitt sitting on his sofa playing 540 00:28:49,840 --> 00:28:54,840 Speaker 1: Call of Duty all day. Okay, So that's the use 541 00:28:54,880 --> 00:28:58,080 Speaker 1: of math in sports. As we mentioned, it's the kind 542 00:28:58,120 --> 00:29:00,920 Speaker 1: of thing that you need people with PhDs in statistical 543 00:29:00,960 --> 00:29:05,040 Speaker 1: science to really compete in sports. Analytics uses a wide 544 00:29:05,120 --> 00:29:10,640 Speaker 1: range of mathematical models, including regression models, Bayesian inference, facial 545 00:29:10,720 --> 00:29:14,760 Speaker 1: statistics in more and more of these days. AI. But 546 00:29:14,880 --> 00:29:18,600 Speaker 1: here's the big plot twist or upset to use sports lingo. 547 00:29:19,000 --> 00:29:22,160 Speaker 1: It turns out all these models are still not the 548 00:29:22,200 --> 00:29:26,400 Speaker 1: most accurate way of predicting who's going to win a game. 549 00:29:27,120 --> 00:29:31,200 Speaker 1: There is another indicator which at its core has almost 550 00:29:31,240 --> 00:29:34,680 Speaker 1: nothing to do with math. So when we come back, 551 00:29:34,840 --> 00:29:37,600 Speaker 1: we'll talk about what this method is, and I think 552 00:29:37,640 --> 00:29:39,960 Speaker 1: you're going to be shocked to find out what it is. 553 00:29:40,520 --> 00:29:42,040 Speaker 1: Oh and also we're going to find out if the 554 00:29:42,040 --> 00:29:44,959 Speaker 1: Dodgers won the game I went to after all, So 555 00:29:45,080 --> 00:30:05,840 Speaker 1: stay with us till the last inning. We'll be right back. Hey, 556 00:30:05,880 --> 00:30:08,680 Speaker 1: welcome back. I'm here watching the LA Dodgers play the 557 00:30:08,720 --> 00:30:12,080 Speaker 1: Arizona Diamondbacks and it's the top of the last inning 558 00:30:12,400 --> 00:30:17,080 Speaker 1: and the score is Dodgers five, Arizona Diamondbacks four. Dodgers 559 00:30:17,080 --> 00:30:24,760 Speaker 1: are up by one point. But we're talking about sports analytics, 560 00:30:24,880 --> 00:30:27,080 Speaker 1: or the idea of using math and science to help 561 00:30:27,120 --> 00:30:29,840 Speaker 1: teams win at sports, and so far we're talked about 562 00:30:29,880 --> 00:30:32,240 Speaker 1: where this idea came from and how it spread to 563 00:30:32,280 --> 00:30:34,680 Speaker 1: almost every sport. Now, I don't know who's going to win. 564 00:30:34,960 --> 00:30:37,479 Speaker 1: Anything can happen in the next few minutes. But according 565 00:30:37,520 --> 00:30:41,200 Speaker 1: to doctor Balmer, there is a way to almost perfectly predict, 566 00:30:41,400 --> 00:30:44,280 Speaker 1: at least from a statistical point of view, exactly who 567 00:30:44,400 --> 00:30:47,400 Speaker 1: has a better chance at winning. And it has almost 568 00:30:47,440 --> 00:30:57,920 Speaker 1: nothing to do with mathematical models. The fourth idea, what's 569 00:30:57,960 --> 00:30:58,640 Speaker 1: the fourth idea? 570 00:30:58,880 --> 00:31:02,040 Speaker 2: Oh yeah, betting my At a certain point, you get 571 00:31:02,040 --> 00:31:05,120 Speaker 2: down to the fact that the best estimate that we 572 00:31:05,440 --> 00:31:09,640 Speaker 2: as human society have of estimating who's going to win 573 00:31:09,680 --> 00:31:12,000 Speaker 2: a particular game are the betting market oughts. 574 00:31:12,440 --> 00:31:14,640 Speaker 1: Is that right, that's the best estimate? 575 00:31:14,880 --> 00:31:17,880 Speaker 2: Yes, this is a result that has been corroborated time 576 00:31:17,920 --> 00:31:18,600 Speaker 2: and time again. 577 00:31:19,120 --> 00:31:20,960 Speaker 1: But who sets those betting odds? 578 00:31:21,040 --> 00:31:24,520 Speaker 2: People like me that work for sports gambling outfits. So 579 00:31:24,560 --> 00:31:27,600 Speaker 2: they set the lines, but then based on the money 580 00:31:27,640 --> 00:31:31,320 Speaker 2: that people bet, the lines can change, and so by 581 00:31:31,360 --> 00:31:34,400 Speaker 2: the time the game actually starts, you have a sort 582 00:31:34,400 --> 00:31:38,040 Speaker 2: of reflection of humanity's collective wisdom about who's going to 583 00:31:38,120 --> 00:31:38,680 Speaker 2: win this game. 584 00:31:39,200 --> 00:31:43,280 Speaker 1: That blew me away. You're relying on people's intuition because 585 00:31:43,440 --> 00:31:46,600 Speaker 1: the average sports better doesn't have a math, they don't 586 00:31:46,640 --> 00:31:48,760 Speaker 1: have access to the data, They just have a feeling. 587 00:31:49,080 --> 00:31:51,160 Speaker 2: Well, it's this notion of wisdom of the crowds. 588 00:31:51,240 --> 00:31:51,760 Speaker 1: Uh huh. 589 00:31:52,000 --> 00:31:54,960 Speaker 2: A lot of statistics is just like the average of 590 00:31:55,000 --> 00:31:58,440 Speaker 2: a lot of things is better than any one person's 591 00:31:58,480 --> 00:32:01,560 Speaker 2: one guess, you know, uh huh. And this is just that. 592 00:32:02,000 --> 00:32:04,880 Speaker 2: But with money at stake, which is where people tend to, 593 00:32:05,760 --> 00:32:09,520 Speaker 2: you know, use all of their collective wherewithal to make 594 00:32:09,520 --> 00:32:10,360 Speaker 2: the best estimate. 595 00:32:10,720 --> 00:32:12,640 Speaker 1: I see, there's a lot of noise, Like you never 596 00:32:12,720 --> 00:32:17,080 Speaker 1: trust one sports betting guide to their hunch, but if 597 00:32:17,120 --> 00:32:20,720 Speaker 1: you have a million of them, then collectively they sort 598 00:32:20,720 --> 00:32:23,480 Speaker 1: of have sort of absorbed all this data in their 599 00:32:23,520 --> 00:32:28,360 Speaker 1: squishy brains and have somehow projected that into their model 600 00:32:28,480 --> 00:32:31,240 Speaker 1: that averages out and somehow that gives you an estimate 601 00:32:32,040 --> 00:32:34,400 Speaker 1: and you're saying that's better than anything we can come 602 00:32:34,480 --> 00:32:39,800 Speaker 1: up with. Yes, what does that mean that it's better? 603 00:32:40,120 --> 00:32:43,440 Speaker 1: Like over time, it's more accurate over time. If you 604 00:32:43,560 --> 00:32:46,120 Speaker 1: say that it's a forty seven percent chance that this 605 00:32:46,200 --> 00:32:48,280 Speaker 1: is going to happen, that it actually turns out to 606 00:32:48,280 --> 00:32:51,600 Speaker 1: be a forty seven like forty seven times out of 607 00:32:51,640 --> 00:32:54,760 Speaker 1: one hundred in the future, like they actually win, They'll 608 00:32:54,760 --> 00:32:55,160 Speaker 1: be right. 609 00:32:55,720 --> 00:32:57,840 Speaker 2: It's calibrated. It's well calibrated. 610 00:32:58,080 --> 00:33:00,000 Speaker 1: Oh the time. 611 00:33:00,080 --> 00:33:03,280 Speaker 2: It's like, if the betting odds imply that the Cardinals 612 00:33:03,280 --> 00:33:06,080 Speaker 2: have a forty seven percent chance of beating the Padres, 613 00:33:06,240 --> 00:33:09,040 Speaker 2: then if you were to take all the games in 614 00:33:09,120 --> 00:33:12,960 Speaker 2: which the odds were the same as that game, uh huh, 615 00:33:13,000 --> 00:33:14,840 Speaker 2: the team I guess it would be the underdog in 616 00:33:14,880 --> 00:33:18,080 Speaker 2: this case, they would win forty seven percent of those games. Wo. 617 00:33:19,440 --> 00:33:23,280 Speaker 2: That's wild. It is wild. But it's also like it 618 00:33:23,360 --> 00:33:25,520 Speaker 2: has to be that way, because if it wasn't that way, 619 00:33:25,720 --> 00:33:28,000 Speaker 2: then you'd have a whole bunch of people who would 620 00:33:28,000 --> 00:33:30,880 Speaker 2: figure that out and they would bet on the other team, 621 00:33:31,360 --> 00:33:35,240 Speaker 2: and that would move the line right that's incredible. 622 00:33:35,400 --> 00:33:37,320 Speaker 1: So you're almost saying that the wisdom of the crowd 623 00:33:37,360 --> 00:33:40,480 Speaker 1: is better than someone with a PhD in statistics. 624 00:33:40,560 --> 00:33:44,240 Speaker 2: Yeah for sure. Well, because there's a bunch of people 625 00:33:44,240 --> 00:33:46,720 Speaker 2: with PH's and statistics who are part of the crowd. 626 00:33:47,040 --> 00:33:48,320 Speaker 1: Oh, I see, that's part of it. 627 00:33:48,320 --> 00:33:51,080 Speaker 2: It's like you got all those people, and you've got 628 00:33:51,120 --> 00:33:52,520 Speaker 2: a whole bunch of other people. 629 00:33:52,440 --> 00:33:55,080 Speaker 1: Just fans. Fans who knows who won in the last 630 00:33:55,160 --> 00:33:56,000 Speaker 1: one hundred games? 631 00:33:56,040 --> 00:33:57,120 Speaker 2: Maybe exactly. 632 00:33:57,280 --> 00:34:00,560 Speaker 1: Wow, that's wild. Okay, we have only a few minutes. 633 00:34:00,680 --> 00:34:02,479 Speaker 1: I'm going to a Dodgers game this evening. 634 00:34:02,760 --> 00:34:03,400 Speaker 2: Oh, wonderful. 635 00:34:03,440 --> 00:34:05,400 Speaker 1: What should I look out for? What should I expect? 636 00:34:05,600 --> 00:34:08,399 Speaker 2: I'm so excited for you. I love Dodger Stadium. It's 637 00:34:08,440 --> 00:34:11,080 Speaker 2: a great place to watch Basaka and enjoy the weather 638 00:34:11,600 --> 00:34:13,480 Speaker 2: and just enjoy the twilight. 639 00:34:13,800 --> 00:34:16,640 Speaker 1: Well we have show hey uh badding tonight. 640 00:34:16,400 --> 00:34:18,719 Speaker 2: Well you'll watch the greatest baseball player of all time? 641 00:34:18,760 --> 00:34:18,960 Speaker 3: Then. 642 00:34:19,040 --> 00:34:21,520 Speaker 2: Wow, I mean, the Dodgers are the best team in baseball. 643 00:34:21,560 --> 00:34:24,040 Speaker 2: I don't think anyone really doubts that. You know, they 644 00:34:24,120 --> 00:34:27,600 Speaker 2: have used analytics, they have gone deeper, and they have 645 00:34:27,719 --> 00:34:30,920 Speaker 2: put the resources behind it. So it's like moneyball is 646 00:34:31,320 --> 00:34:33,520 Speaker 2: sort of how do you win more games with less money? 647 00:34:33,600 --> 00:34:36,080 Speaker 2: But the Dodgers are doing moneyball with money. 648 00:34:37,440 --> 00:34:42,080 Speaker 1: It's like money moneyball. Yeah, okay, looked at a CBS 649 00:34:42,120 --> 00:34:46,920 Speaker 1: sport says the Dodgers are favorite to win, okay, minus 650 00:34:47,040 --> 00:34:49,400 Speaker 1: two sixty six favorite on the money line. 651 00:34:49,600 --> 00:34:51,239 Speaker 2: So, and I don't know that I'm going to be 652 00:34:51,239 --> 00:34:52,759 Speaker 2: able to do this off the top of my head, 653 00:34:52,760 --> 00:34:56,680 Speaker 2: but like, you plug that number into a fairly simple formula, 654 00:34:57,160 --> 00:34:59,239 Speaker 2: and that's going to tell you that the Dodgers have 655 00:34:59,360 --> 00:35:01,920 Speaker 2: a sixty of winning this game. 656 00:35:02,360 --> 00:35:02,760 Speaker 1: Okay. 657 00:35:02,840 --> 00:35:06,439 Speaker 2: And so now that gives us a sports analyst the 658 00:35:06,480 --> 00:35:09,640 Speaker 2: most accurate prediction for what's going to happen in this game. 659 00:35:11,960 --> 00:35:13,359 Speaker 1: All right, we'll see how it pans out. 660 00:35:13,400 --> 00:35:17,680 Speaker 2: Then, yeah, exactly, all right. 661 00:35:17,560 --> 00:35:20,360 Speaker 1: Did the Dodgers win after all? Here's the audio of 662 00:35:20,360 --> 00:35:51,480 Speaker 1: the last few moments of the game when, Yeah, the 663 00:35:51,560 --> 00:35:55,360 Speaker 1: Dodgers won, just like the fans, the mathematicians, and the 664 00:35:55,360 --> 00:35:59,759 Speaker 1: betting markets predicted. Thanks for joining us, see you next 665 00:35:59,800 --> 00:36:08,280 Speaker 1: time you've been listening to science stuff. Production of iHeartRadio 666 00:36:09,040 --> 00:36:12,000 Speaker 1: written and produced by me Or hitch Ham, edited by 667 00:36:12,080 --> 00:36:15,960 Speaker 1: Rose Seguda, Executive producer Jerry Rowland, and audio engineer and 668 00:36:16,000 --> 00:36:19,080 Speaker 1: mixer Kasey Pegram and you can follow me on social 669 00:36:19,120 --> 00:36:22,120 Speaker 1: media to search for PhD comics and the name of 670 00:36:22,160 --> 00:36:25,080 Speaker 1: your favorite platform. Be sure to subscribe to Science Stuff 671 00:36:25,120 --> 00:36:28,239 Speaker 1: on the iHeartRadio app, Apple Podcasts, or wherever you get 672 00:36:28,239 --> 00:36:30,560 Speaker 1: your podcasts, and please tell your friends