Hello and welcome to the NVIDIA AI Podcast.
I'm your host, Noah Kravitz. Look, this podcast is great But if you really want to get brain deep in AI, the GPU Technology Conference is the place to do it.
At GTC, you can choose from over 600 sessions on dozens of hot topics.
You can get hands-on training and you can connect directly with experts.
Google, Stanford, Amazon, Apple, MIT, Pixar and hundreds more of the kinds of places whose experts you want to connect with.
It all starts March 17th in Silicon Valley, and right now you can save 20% on registration.
Just go to gputechconf.com and use the code GMPODCAST when you register.
That's gputechconf, C-O-N-F, .com. And the code is GMPodcast.
We'll see you there. In May of 2018, the state of New Jersey scored a victory in the Supreme Court that paved the way for legalized sports betting in the US on a state-by-state basis.
As of right now, March 2019, sports betting is legal in eight states.
But ESPN counts another 30, plus the District of Columbia, as moving towards legalization.
That said, scores of people regularly play what are known as predictive gaming games.
NC football to office pools based on picking the winners of things like the annual March Madness College Basketball Tournament that's right around the corner as we record this.
Today's guest co-founded a company, Swish Analytics, that leverages AI and data to help bettors and predictive gamers find an edge when they're picking winners.
But Swish actually does most of their business in the B2B space, which we'll talk about in a minute.
Corey Beaumont is co-founder and head of engineering at Swish Analytics, and he's joining the podcast from his company's headquarters in San Francisco.
Corey, welcome, and thanks for coming on the show.
Hey, thanks for having me. So obviously, I want to hear, the audience wants to hear about Swish, how you got started.
And the work you've been doing for a few years now, using machine learning and AI to crunch big data, and help make predictions on not just the scores of games, but all the things that bettors bet on.
But you and I, we talked offline. We're both big Hoops fans.
It's March right now. We're in the Bay Area.
Sorry, Warriors haters out there, but it is what it is.
And you guys were just part of a pretty big announcement involving another local team, the Sacramento Kings of the NBA.
So maybe let's lead with that time for basketball.
It is a good time for basketball for sure.
Yeah. It wasn't last night for the Warriors as we record this, but they're tired.
They'll be fine. Yeah. Maybe you can give us kind of the elevator pitch, the 30-second description of what Swish does, and we'll come back to that.
But then tell us about what the Kings are doing, why it's important to the future of legalized sports betting in California and the NBA. and what your role is in that.
And then from there, we can kind of get into more of what Swish does.
Sure. So Swish Analytics is a predictive sports company where we use machine learning algorithms to try to decide how a player is going to perform or a team is going to perform and, um, kind of to what end in what time period are they going to do that over the course of a game of a quarter? over the season, and so on.
We take these predictions and... Turn them into several different things for a lot of different companies around the sports arena, for marketing agencies, teams.
And externally to, to the U S a lot of sports books and other more sports betting type of companies like that.
Right. That's pretty much what we do in a nutshell.
We took kind of an interesting path where myself and my two co-founders, Joe and Bobby, we were working in fintech And we're using machine learning to kind of predict if a random guy in Ohio was going to be a fraudster or not.
All based off of public data and all that good stuff and kind of what they... give us, which may or may not be reliable.
I know you picked it out of the air, but to be clear, no shots at Ohio.
I lived in Ohio as a kid, so we're all good there.
No coastal elitism. Let's avoid that. Nothing against Ohio.
Yeah. But yeah, so we would take that data and try to predict what that, you know, what that user was going to do and how that user was going to turn out.
Right. We kind of figured, all right, well, if we can do this for some random guy. we could probably figure out what Kevin Durant's going to do tonight on the basketball court or, you know, what Tom Brady is going to do in the Superbowl.
Yeah. So we kind of took our learnings from that Lo and behold, we were actually pretty good at it.
So we decided to kind of take our first step into the sports analytics arena.
And we put up a website. Originally, it was called Bet Buckets.
It was just a kind of... very lightweight predictive analytics site that if you wanted to kind of get an edge on some fantasy stuff or what have you, you can go on there and get some information and use it pretty easily, that name of bet buckets quickly became an issue. as we tried to advertise and get our name out there and stuff like that.
So we kind of, did a little name swap and upgraded the website.
And that's how we became Swish Analytics actually.
Cool. Million things I want to ask you about from that.
But you mentioned doing a lot of work outside of the U.S.
Legalized betting is a little more prevalent outside of the States.
In the States, there's this kind of state-by-state move towards legalized sports betting, which, you know...
It stands to be an enormous industry and not just the betting, but lots of opportunities around the betting.
The NBA, I mean, we're biased. We're both Hoops fans.
But the NBA has kind of long been known as forward thinking amongst the major sports leagues.
And so people are talking a lot about the NBA being the league in the U.S. that could really kind of do the most out of the gate possibly with sports betting.
So what's going on? The Kings Arena, they have this predictive gaming lounge.
And you guys have an app called BetRush, I believe, that was just announced to be part of it.
What's going on up there and how are you involved?
Yeah. Yeah. So the NBA as a league, we do see them as very progressive and kind of the leadership that Silver's providing them to touch in these different areas that are kind of more taboo for some of the other leagues.
As it relates to the Kings specifically, we ended up getting in contact with them and they're looking for really how we can help them build... what a 21st century or new age type of sports book might look like.
I think when people think of the sports book, They think of kind of like a dark area in the back of a casino with a bunch of TVs and cigarette smoke, right?
Yeah, at best. Right. If you're lucky. So what the Kings have really done, they're investing in kind of an Apple store type feel, very modern and contemporary feel of the sports book and how it might work.
And we are the technology backing that. And so we created a product called Betrush, which the way you will be doing a demo for several games and hopefully we get to expand that into the full season and into other teams to really change the idea of how sports betting and sports books are seen, if that makes sense.
No, it does. It's fascinating. legalization and regulation is one thing, but then, you know, it's the whole experience and the image and absolutely.
So let's shift gears a little bit and talk about how you guys leverage machine learning and AI.
You know, you spoke at the beginning about your background in FinTech and as you were speaking, it made perfect sense, right?
If you can... If you can try to model what an individual is going to do, that individual could be a potential fraudster, could be a sports superstar and how they're going to perform on the court.
So without getting into your secret sauce, obviously, how does it work?
What kind of factors do you look at? What kind of, I don't want to say results, but you know, what's the end result in terms of what a what a user might see.
And then also maybe you can talk a little bit about the B2B work you're doing, which is kind of, I think, more the core of your business model.
Sure. So the interesting thing about sports is, and as it relates to machine learning, is obviously you need data, but within sports, There's only so much data available and there's only so much diversification in that data.
If you compare it to website tracking or marketing or something like that, someone may be able to develop something that is a unique product that follows the mouse clicker on the screen and they get to use that data to model against.
We don't have the ability to do that. We only have the data that is available from the leagues.
And so we have to work pretty hard internally to kind of develop these data sets and look at these different features and what matters and what doesn't matter.
But some of the things, as we know, if you follow basketball closely, this year and the previous years have really been changing in terms of of number of points being scored and number of threes taken and the pace of the game.
Yeah. Everything's much, much faster, more points scored.
Yeah. Like you said, the past two years in particular.
Yeah. I think the most three pointers ever shot in basketball were the last three years and it's continues to grow.
Right. So it's really interesting from that point of view, how The data is changing and we need our machine learning algorithms to also change with that.
But we want to make sure that you're not over overfitting or something like that, a model to a particular change or something that we're seeing in this year versus just kind of the nature of basketball, if that makes sense.
It does. Of the sports that you guys work with and look at, what's the easiest to predict?
Cool. Or what's the hardest? So something that is hard to predict, which we don't do college sports yet, but we're definitely evaluating them.
Mm-hmm. The data there is interesting because you have guys who play one and done.
So they come for their freshman year, they play and then they go to the professionals.
Right. So there's not much data there. Right.
That is an issue. Every, every sports a little bit different.
So like football. you don't have that many games.
And so you have a whole lot less data. You have a whole lot less data on that basis.
And also you... If you think about sports data, you really only accumulate data when you have the ball Right.
Everyone else off the ball that is affecting the play, they don't really have too much data unless you're just doing like X, Y coordinate tracking and stuff like that.
Right, right. So... You have to figure out how to quantify that into something useful.
Baseball's great. You know, it's a batter versus pitcher matchup, and you know exactly what's going to happen.
The umpire is a little bit of a question mark at times, but...
We saw that actually in spring training, I believe, this year.
The MLB is trying out some robotic umpires to standardize some of the strike zone stuff. that you see in baseball.
So interesting things. Cool. And so tell us a little bit about your B2B work.
I know you guys have an API and you make that available to clients.
Right. Although we love our consumer side of the site and interacting with the individuals who use our data to do things, we see the real advantage is allowing bigger companies to do a lot of different things with our data.
So we've seen everything from companies using our data to price tickets.
Obviously, companies in Europe and in Asia and all over the place are using our data for for betting and odds making and things of that sort.
We've seen opportunities like BetRush, for example, where teams aren't necessarily using it for the management of the team, but for a product within a facility.
To go back a second, when you said to help price tickets, dig into that a little bit, what does that mean?
Yeah, so if you've ever purchased tickets on like StubHub where the prices are going up and down, Some stadiums from kind of a if you were to buy a ticket from a box office or online or something like that, based on the game, that same seat will cost more, obviously.
Right. Obviously for like playoffs and things like that, but even from game to game.
Yeah, the Warriors, I think they call that dynamic pricing.
Maybe? Yes. Yeah. Exactly. Exactly. And so what they might do is use our data to figure out what games are going to be closer than the others.
What games are... predicted to kind of be a blowout.
Right. If a game is going to be a blowout or it might be boring.
Should that organization do a t-shirt giveaway or a bobblehead day so they can get the fans in the door?
Right. So we see things like that and really the data that they use initially to figure that out is Swish.
Very cool. We're talking with Corey Beaumont.
He's the co-founder and head of engineering at Swish Analytics in San Francisco.
And Swish is in the business of... using AI, using machine learning to help predict not just the final score, but all manner of things related to sports, predictive gaming, sports betting.
Corey, let's go back in time a little bit and talk about you and your background.
You're the head of engineering at SwishNow.
You mentioned you worked in fintech before.
As a kid, were you into data, were you into sports, were you into money?
How'd you get started? Sure, sure. As a kid, I grew up in and was born in San Francisco, and then my parents ended up moving to Texas and my my bond back to California aside from my family was really the sports teams.
Sure. So I was watching every game and just, you know, even when you'd be watching. you'd be watching a game and it'd be cutting in from commercial and they show you kind of a Bay area.
And I was like, Oh, I missed that. Right.
Right. You know, and I played soccer my whole life.
And so at the time I was playing sports and I was also a pretty big gamer online.
I was playing a lot of Counter-Strike. And it's interesting to now be on the NVIDIA podcast when I was I was playing games on a Sony VAIO or something that had a small NVIDIA chip in it way back in the day.
So this is pretty cool. Yeah, yeah. But my dad, my parents are engineers and whatnot.
And so I picked up hacking in some Counter-Strike stuff and picked up some technical projects pretty early on.
You know, by the time I was exiting high school, I was pretty comfortable with, with writing some code and some programs and stuff.
And so as I went to college, I figured I can study computer science and business because I figured If anything, if you major in business, you know how to manage money, right?
Right. Which I didn't know at the time. One would hope anyway.
Right. In theory. So I went to school to do that, and I actually had to drop my computer science major in order to get my master's in business in one year, which... allowed me to go to London for a year and study.
And so this is where... really the sports and the betting and the money came together for me because sports betting is legal in the UK.
And so... As an avid soccer player, I was able to really look at the game and see, kind of compare what I thought was going to happen with what those sports books were doing or with what my, my close group over there thought.
Right. And quickly found out I was really bad at that.
So in being really bad at it, right, I got more interested in trying to figure out why and how does it really work well.
As time went on, I had to come back to the States, at which point I finally made that move back to San Francisco and started working for... that FinTech company where I met my two co-founders.
And, you know, you get to talking and all of a sudden, everyone is a sports better and you're like, Hey, well, we're also kind of technical too.
Like maybe we can try to do something with this.
And, uh, the rest is a fairy tale. Yeah.
What year was that? How long ago? About 2014, 2013, I guess.
So about four or five years ago. Yeah, yeah.
When you guys started Swish, did you have the model in mind to...
Start gathering data and then crunching it and making predictions off of it, analyzing it, and then licensing it out?
Or did you imagine yourselves more of a consumer-facing thing?
That is an interesting question that I was actually just talking to one of my partners about.
To look back and to say that we got to hear... we wouldn't have guessed this is the path that we would have taken to get Swish to where it is right now.
Right. And, you know, we were a part of the LA Dodgers accelerator.
Oh, interesting. Which was... Being a Giants fan was really... I had to really question what I was doing, but... I'm not a big baseball guy, but I know what that means anyway, so yeah.
Yeah, I mean... Honestly, it was one of the greatest opportunities that we to get through Swish and we met some really great people who are still with the company and have helped us.
You had Tommy Lasorda teaching you recursive algorithms?
You know, funny enough, my second day there, I think I met Magic Johnson.
Oh, no kidding. Okay. Yeah, and he's just legendary across anything.
Sure, yeah. So yeah, that's kind of how we got to where we are.
Very cool. So then where you are now, spring, March, 2019, how big is the company?
And, you know, what are you that you can talk about anyway?
What are you getting into next besides the bet rush stuff with the Kings and obviously all the work you're already doing?
What does the future look like for you guys?
And then maybe for the industry of... I mean, sports analytics is a big word, but I'll let you kind of take that to your purview.
What do you think the future holds going forward?
Sure, sure. So for us specifically, we really see an advance in a need, I guess, for in-play betting. and in-play predictions and predictions in general.
And baseball is really looking into a lot of this stuff.
So how do you keep that fan engaged, whether they're at home, or in the stadium?
And how do you kind of add on to the experience of being at the game or what have you? while being at home is really comfortable, right?
So that's not just betting on who's going to win or covering the spread, but getting into...
How many points in a quarter? Or is this guy going to strike out at his next at-bat?
That kind of stuff. Exactly. Exactly. And that's got to be... I mean, that's a huge, endless opportunity, it would seem like.
It is. It is. I mean, you can really... the ability to place a bet on Steph Curry, three pointers, but also Russell Westbrook turnovers or, you know, Chris Paul.
For example. For example. It's a really powerful thing, especially in that quarter or on this next possession or in the sense of baseball, what's going to happen at this at-bat?
What's the likelihood that this is going to be you know, a double or a triple or a foul tip or what have you.
The optionality in what we call markets, like kind of the player, the game, the stat type, and the outcome.
And how do we do that in a live atmosphere as the game is going on, especially in a In games like basketball or in hockey where there are no stoppages, really.
It's not like football where there's a play or... baseball where there's a pitch and then there's a little rest in between built in.
Yeah. Yeah. For, for us from a technical point of view to, um, kind of leverage a lot of the stuff that NVIDIA is pioneering to, um, Make sure that we're speedy and efficient in that form or fashion.
And are you talking about speed? inefficient in giving your guidance on predictions or about in empowering the actual predictive gaming system itself?
I'd say empowering in making sure that when this data comes in, how do we make sure that it's applied to the model?
And we get an output in seconds, if not fractions of a second.
And then that gets shot out in a real-time state to... someone's phone in the stadium, right?
Right. That's kind of where we're headed, I think.
Aside from, you know, we're always evaluating new sports and new opportunities from that point of view and new territories and things like that and new partnerships.
We're really focusing on the next evolution, which is the in-play aspect of sports betting and whatnot.
Yeah, so as it relates to kind of the greater economy of sports analytics and what's going on kind of here in the US, but also globally.
It's an interesting time to say the least where a lot of, now this is at the state level, states are coming up with different things and different rules and different costs and different regulations where some places you can do this and some places you can't do that.
So we as an organization are at the forefront of that trying to make sure things get done properly and things make sense for the greater economy.
But I think we're going to see a lot more variation in sports gaming and that doesn't mean just like sports betting, right?
But these offshoot fantasy games or like micro-betting, where you're betting cents and dollars instead of $100 or $10. hedge fund type of sports book related things where a hedge fund may go out and start placing some bets on some games based on some good predictive analytics that they have A lot of, and I said this a little bit earlier, but a lot of different things for a lot of different people.
And that's where we like to place ourselves in the business-to-business aspect.
If there's something that you need from a predictive sports analytics point of view, We got you.
We can do it. And I think we've proven that.
I've been biting my tongue really wanting to say like, oh, it sounds like a smart bet you guys are placing on yourselves.
But, but it does, I mean, it sounds like you're, you're super well positioned at a, an industry that, you know, is, is.
It seems like it's not just going to take off straight line, but more explode into all these different directions, or I'm sure already is.
But that's exciting stuff. So brass tacks, before I let you go, is there sort of a metric understood as what a good rate of success is in predicting?
I mean, I'm sure it's different across countries. different sports and different types of bets.
But say for picking who's going to win the NBA title, or picking the MVP or even picking a certain game is like 51% accuracy considered good?
Is there a number? Um, so it really depends if you're looking at it from a betting perspective, obviously you want to be above 50%.
But you also have to take into account the sportsbooks cost or VIG if you want to. call it that.
And which is usually two, 3%, maybe 4%. Okay.
So in order to be profitable, you really need to be above that 54% margin.
Got it. Okay. All right. So going forward, and I don't know if you guys have official predictions on this stuff.
If not, feel free to give your own take, but we'll stick to the NBA.
Are the Warriors going to three-peat? Having watched a lot of basketball...
I think the Warriors definitely do have a good shot at three-peating.
That was an easy one. Here's a harder one.
Is LeBron going to miss the playoffs for the first time in how long it's been?
You're really setting me up here. As many Warrior fans, we kind of despise the Lakers.
And having LeBron stealing one away from the Warriors...
I'm not the biggest LeBron fan as a basketball player, at least.
I can relate. I can relate. I have a lot of respect for him, but not necessarily a fan.
Yeah, however, they do have Luke Walton, which coached the Warriors for some time.
It's going to be close, to be honest. I think that they... can certainly make the playoffs.
And if anyone is going to get them, get that team to the playoffs, it's going to be LeBron.
Do I think that they will make the playoffs?
I think that they will. Do I want them to?
Not necessarily. But I think in the coming years... the Lakers are going to be a pretty good team.
All right, I'm going to let you off the hook with the predictions.
If folks want to learn more about what Swish is doing in all these different arenas, you guys have a website, you have a fun Twitter feed.
Where should people go to learn more? Yeah, you can reach out to us at swishanalytics.com.
You can reach us on Twitter as well, at Swish Analytics.
If you have any questions or anything like that, you can always shoot us an email at info at swishanalytics.com and We're happy to help.
Excellent. Corey Beaumont, thank you. It was fun talking to you.
And I'm going to admit when you brought up the hedge fund example, It made perfect sense, but my mind blew just a little bit and I just bit my tongue at my pool.
But it's a brave new world. It'll be really interesting to see. how this stuff develops in the coming years.
Definitely. It's kind of the Wild West, but it's getting there.
All right. Thanks for joining us. Take care.
I really appreciate it. Thanks. Thank you.
Thank you.
