Nate Silver
speaker
1,377 appearances
5 recordings
5 series
first heard Feb 2025
last heard 29 Jun
Nate Silver’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 3 in all, peaking in Jun 2026 with 2.
Appearances
So Pocota, the anagram is very, I'm not going to bore your listeners for even 15 seconds. You'd be shocked with their tolerance for boredom.
So you might as well say it now. Picture empirical comparison. It's deliberately nerdy. An optimization test algorithm, I think.
This obscure player who was always like a thorn in the side of the tigers.
Oh.
I have a friend who's a neuroscientist. When he's like, yeah, our brain's just kind of just predicting and inferring. We never actually know what reality is. We're just making like an expedient internal map simulation.
It's to forecast how baseball players will do. The innovation is that most predictions will just say, okay, this player, Shohei Otani, will hit .302 with 42 home runs and 108 RBIs next year. Whereas this gave like a range of options, so it was probabilistic. And that's kind of one of my big things is we don't know the future. Baseball players get injured or stop using steroids or whatever else.
They're hungover, they get divorced. They're human beings, and there's luck. But no, the ideas have like a range of outcomes. And so you're dealing with the uncertainty in the outlook. But yeah, people use it for fantasy baseball. Major league teams used it. You did this in 2000. Was there already fantasy baseball at that time? Yeah. I mean, fantasy baseball dates back to the 80s.
It's kind of in the Moneyball tradition.
Well, I do want you to tell me, what year did that happen? Moneyball was 2002 or 2003, somewhere in that range.
About the same time. So coincidentally, the company was called Baseball Perspectives, and this kind of blows up, in part because of Moneyball and the Red Sox, who are being very stat-friendly, win the World Series in 2004. Yeah, it becomes a whole thing.
People have this bad habit of pretending that, oh, the model just flaws out of a coconut tree. It'll quote former presidential candidate Kamala Harris. And no, you have to make a lot of decisions. In sports, at least, the data is high quality. We record everything that happens on a major league playing field. But you're making decisions at every turn.
A lot of it is about where you add more complexity and where you're pruning because models break. So you want like a complex simplicity.
Algorithm is the term that would technically be used for our political forecast. Literally, you input the polls and press the go button and it takes five minutes and runs 50,000 simulations and spits out a bunch of data. But that makes it seem like it has a mind of its own. It's hard because on the one hand, you don't want to dictate what the data says, right?
On the other hand, if you already have the answer you want, that's not really objective science either. So it's kind of this iterative process between I want to systematize, so I have to follow rules and not just be totally ad hoc. If I feel like it's handling this player badly or this election badly for a politics model, and I change the rules, what's that mean for the other elections?
What's that mean for the whole system overall? It's not like there's just one right answer. What are hedge funds and banks doing? Or what are the sports bettors doing? They're usually using multiple models. You want a model that's robust. So basically, if one thing breaks and it will still spit out a reasonable answer, that's where some of the art comes in as opposed to the science.
It's as a hobbyist. You're trying to win your fancy baseball league. You're trying to solve a problem that you're determined to solve. You're trying to maybe win money, gambling, for example. You need to be very hands-on and have skin in the game because academics, to cliche a little bit, tend not to have good street smarts for modeling.
A lot of the skill is like being able to look at a data set and say, the value that number put out, something's wrong there. And there must be either a bug in the input. We put the numbers in wrongly or the code failed somehow.
No, I still do it for the most part, all my own coding. I'm a control freak in that way.
You know, if you can get 56% of your points, that's right, then you're like a world class.
And that comes from poker, too. Poker players can discern like a 52% probability from 48%, right? Most people think 100-0 and 50-50, right? There are a lot more gradations between 50-50 and 100-0.
Showing 1181–1200 of 1,377 · page 60 of 69
← Previous
Next →