Ed Thorp on the limits of models
Update: Link to the audio interview
Ed Thorp is a legend. “Fortune’s Formula” details his obsession on finding an edge in games and the stock market. He is also prominently mentioned in “The Quants“, is the author of “Beat the Dealer” and founded Princeton Newport Partners, one of the most successful hedge funds ever without a single down year. This is an excerpt from a recent NPR interview:
GROSS: So Ed Thorp, were you invested in the market a lot either personally or through your own hedge fund in 2008 during the stock market collapse?
Mr. THORP: Yes. At 2008, unfortunately, I didnt have any place to hide. I did not have a hedge fund then, so I wasnt managing more than a very small part of my own money. The small part, I managed it fine. I made about 18 percent that year, but most of the money I had was in other things – other people’s hedge funds and so on. I was trying to retire and live the good life.
GROSS: So you lost a lot of money with the money you had invested in other people’s hedge funds?
Mr. THORP: A significant amount. We’ve mostly recovered but not quite all of it.
GROSS: So what are some of the ways that you feel the mathematical models that you helped create didnt foresee the housing bubble and therefore the housing collapse?
Mr. THORP: Well, the problem is that the people using these methods and models didn’t use them correctly. They thought they applied to things they didn’t apply to. It would be as though I were talking about gambling odds and somebody were to ask me, what’s the chance of a plague of locusts? Or you know, what’s the chance that seas will run red with blood? I was actually asked that once in a murder trial and my answer was: Those are one time events. They’re out of the collection of assumptions you make when you make probability estimates. And so probability estimates have nothing to say about events like this. You can’t stipulate the probability that one of these things will happen. And these huge tail events – that is, way out on the tails of a probability distribution -come for reasons that are kind of outside the assumptions and the experience of the models.
GROSS: Ed Thorp, as far as you can tell now, how are quants being used on Wall Street? Are these mathematical models relied on as heavily now after the stock market collapsed as they were before?
Mr. THORP: My impression is: pretty much. There is a giant industry now. Many thousands of people who otherwise would’ve gone into engineering and science have gone over to Wall Street to work on these things because the pay is better and it’s fun. And they’re still there. They have a vested interest in staying there. And I think people – a lot of people think that we’re just going to go back to business as usual in this country, that this is all going to blow over and we’re not going to have any significant increase in regulation. We’re not going to have any significant listing of off-the-book derivatives on exchanges like the commodity futures exchanges and that we’ll set ourselves up for another big fall. That’s what I’m afraid will happen.
GROSS: So how’s that affecting how much you want to have invested in the market?
Mr. THORP: Well, its tough. The question is, you know, where do you go? There’s – we only have one world we live in. If we had a market on Mars you might think about going there. But I think its going to be very difficult. And I personally find it hard to guess exactly when some bad thing will happen and how long it will take and what will trigger it. Just like in this last crash, you know something bad is going to happen, you just don’t know when or how.