Protege CEO Bobby Samuels on the data nobody has unlocked yet

Bobby Samuels (0:00)

One of the things that we think about is not just like the letter of the law, but the spirit of the law of like, yes, there is data that we can probably use. As a company, our general take is the creators of IP should be compensated accordingly. It's easy to build certain things and whatnot, but I think that companies should continue to focus on what they're uniquely good at and not try to spin up all these tools. If there's like highly sensitive data that you want to be able to look at before you license, or can only be shared in very sensitive situations, you basically can create a synthetic data set that just allows somebody to sort of understand the general nuances and relationships of the data.

Auren Hoffman (0:43)

Hello, fellow data nerds. My guest today is Bobby Samuels. Bobby is the co-founder and CEO of Protege, a platform that licenses real world data from industries like healthcare and media to companies building AI. Protege has raised $65 million from Andreessen Horowitz, CRV, Flex Capital, where I am.

Before this, Bobby led partnerships at LiveRamp and also ran Privacy Hub at Datavant. Bobby, welcome to Summation.

Bobby Samuels (1:12)

Thank you. It's great to be here and always good to see you, Auren.

Auren Hoffman (1:16)

Okay. Now, going back, I just want to start with this, okay?

I actually went back in my notes. I was lucky enough to interview you for a job there at LiveRamp when you were a senior in college. This is 2013-2014 timeframe. I think that was your senior year. I literally just checked my interview notes a couple of days ago. And it said, and I don't exactly remember what it said, but it said, here's an interesting conversation on the economics of horse racing. Do you have a recollection of this conversation that we had about horse racing?

Bobby Samuels (1:49)

So I was thinking about when we first met before this. And for whatever reason, and kudos to you, I think I remember more interview questions from our interview than like any other ever, pretty much. Okay, so there are three questions that came up. One was that, which so basically.

I almost like I was very like into horse racing. I actually wrote my college essay on like betting on horse racing. I because it's like, you know, it's it's very data driven. And I was like a data, you know, I grew up, I like load baseball stats and like that permeated all this betting on horse horses and stuff.

Auren Hoffman (2:31)

Yes, yes.

Bobby Samuels (2:32)

I mean, I like my my dad because I was not technically legal. So I guess like, you know, right.

Auren Hoffman (2:38)

I was placing the bets on your behalf.

Bobby Samuels (2:40)

Yes.

But then like is I thought I'm sort of writing a thesis and then decided ultimately, not to like there was a lot around basically economic theory as related to betting on horses. And it's actually like there's a lot of interesting parallels now where getting data for horse racing is actually like incredibly difficult and expensive because there's like one company that essentially has a monopoly on it. Oh, really? Yeah. Called the Daily Racing Forum. It's like very. So one of the things I had to actually like the probably the hardest thing for the for the users was going to be getting the data. Well, there were interesting things like.

Auren Hoffman (3:17)

Susquehanna like bets on horse racing too somehow, right? Like even like companies like that.

Bobby Samuels (3:21)

I think they just like pay a ton for data. Yeah. So I mean, there are a lot of like actually equivalents to things I think about now.

So one was how difficult it was to get the data. The second was like there are biases. So there was a thing that I think has been like more or less arbitraged away called like the favorite long shot bias, which was favorites were chronically undervalued, long shots were chronically overvalued.

Auren Hoffman (3:45)

That's true, like on Polymarket still, like they always have, they show these biases on Polymarket all the time.

Bobby Samuels (3:50)

Yeah, because it's like, oh, it's a hundred to one or a thousand to one, like might as well. And then it ends up being like, actually, it should have been a thousand to one. But because everyone did that, it's nine hundred to one or something.

And then the the third is much like a FICO score Nielsen rating. There's actually like this statistic that has developed called the Bayer speed rating, which is essentially this guy named Andrew Bayer. He also he went to Harvard. He dropped out because he wanted to do horse racing stuff. Created this like composite stat that is like used by some group is like the number that distills all these different variables is like the agreed upon speed rating for a specific race. So there's actually like all of these pretty interesting parallels. A lot of the data businesses.