Monday, November 10, 2008

Still Struggling to Understand the Wisdom of Crowds

Recently TechCrunch offered a not-so-positive review of a new service called Piqqem that seeks to harness the wisdom of crowds:
Can Piqqem Use The Crowd To Pick Stocks? Don’t Bet On It.
Judging by the history of posts on this blog, this would be the point at which I decry Piqqem for failing to understand the factors that contribute to the wisdom of crowds. But, in fact, this time the service gets it: it asks users to predict stock futures, and they can't see other peoples' predictions until theirs are already in. Thus the service protects against the herding instinct that makes you (and me) more likely to favor a stock if we see that other people expect it to go up in price (otherwise known as the cumulative advantage).

Instead, this time it's TechCrunch that gets it wrong:

"If nothing else, Piqqem is certainly a good place to get ideas for stocks to invest in. But does it really have any chance of ever beating the market? Like any social investing site, its picks are only as good as the people who contribute to it. But beyond that, there is fatal flaw to this approach.

When it comes to stocks, the best prediction market out there is the stock market itself. It is the biggest prediction market out there, with millions of people predicting the future price of stocks every time they buy or sell shares. All of those predictions are aggregated together in the form of the price. To think that a few thousand, or even a few hundred thousand, people on Piqqem can do better is naive. And in fact, if you look at the prediction lines on Piqqem they already closely hue the actual stock price."

Where to start? The errors in this passage are almost too numerous to count:

  1. "The picks are only as good as the people who contribute to it." Actually not true at all -- and that's why it's the wisdom of a crowd, not the wisdom of a crowd of experts. If you really want to know how this works, read the book, but here's the short version: experts tend to make incorrect predictions because they're deeply entrenched in accepted ways of looking at a problem. That's why an aggregate of amateurs can make predictions that are actually better; as individuals they know nothing, but when you average their individual errors, you end up with an accurate consensus.
  2. The author (Erick Schoenfeld) conflates prediction markets with the wisdom of crowds. The book does talk about prediction markets, but it's not the same thing, for one primary reason: investors in a market are strongly influenced by what they see around them (which is where "irrational exuberance" and stock market crashes come from).
  3. Schoenfeld cites the fact that Piqqem's predictions are holding close to actual market performance as a criticism of the service -- but isn't that exactly what Piqqem is trying to do? If Piqqem's predictions are perfectly accurate, a historical graph of its predictions will exactly match the market's performance; it will simply predict that performance a little in advance.

Bottom line: Schoenfeld clearly has not read (or, possibly, understood) the book, and doesn't really get the idea behind Piqqem. Take his criticism with a grain of salt.