Wednesday, October 28, 2009

Howe, Crowdsourcing

I was driving back from the grocery store listening to
CT public radio when I heard a pitch (pun intended) for a proposed crowd-managed baseball team. One of the guests on the show was Jeff Howe, author of Crowdsourcing: Why the Power of the Crowd Is Driving the Future of Business (Crown Business, 2008). So, naturally, I had to add this book to my series on crowds. For those of you who are sick to death of the theme, this will be my last post on crowds, at least for the foreseeable future. Of course, my foresight isn’t any better than anyone else’s.

Perhaps the best known instance of crowdsourcing was the Netflix Prize, awarded last month. For $1 million Netflix in effect hired groups of Ph.D.s for a dollar an hour to improve its movie recommendation system. The winning team of seven was made up of statisticians, machine-learning experts, and computer engineers from four countries.

Howe, a contributing editor at Wired magazine, takes the reader on a quick crowdsourcing journey from T-shirt company Threadless, iStockphoto, and open source software to P&G’s Connect and Develop initiative (responsible for Swiffer). He highlights the national bird counts coordinated by the Cornell Lab of Ornithology, the world of the citizen journalist, and of course Wikipedia. And just as Page cited Surowiecki, Howe cites Page. (You see, I could have sequenced my reading simply by date of publication instead of my more complicated heuristic that reached the same end.)

Today I’m going to focus on the trading crowdsourcing model at Marketocracy, where investors can open a virtual portfolio of $1 million and join the more than 85,000 virtual fund managers on the site. Marketocracy then monitors these portfolios in search of ideas for its own Masters 100 mutual fund, launched in 2001. Initially, the fund followed a simple model, weighting its positions to match those of the top hundred portfolios. The model was effective. By the end of its first year of operation it beat the S&P 500 by 14%. In both bear and bull markets the model outperformed. But then came 2004 when the market became choppy. The Masters 100 began to underperform its benchmark dramatically; as investors fled, its assets under management withered from $100 million to $50 million in just over a year.

The problem was not only a changing market but that old demon--herding. “As the top investors got to know one another, they started conferring on their positions.” (p. 175) In response the Marketocracy team introduced changes, among which was one that made it impossible for members to see each other’s trades. The team also decided that its pool of 100 was too small (and probably insufficiently diverse) and its algorithm too simplistic. For instance, it was overlooking the specialists—those who, though they didn’t perform brilliantly overall, had unique expertise.

This hybrid model flagged an oil-shipping company called Knightsbridge Tanker whose stock a sub-set of the Marketocracy traders (none in the top 100) was buying aggressively. Intrigued, the management at Marketocracy sent e-mails to these traders to find out why they were loading up on the stock. It turned out that the company had a lot of tankers about to be scrapped. How did the traders know this? Someone checked in Singapore where the tankers were registered. “The conventional wisdom is that when a tanker reaches the end of its life, it’s worth zero. But the price of steel started to go through the roof in the interim, and all that was about to be returned to investors as dividends. Marketocracy made a killing.” (p. 176)

Marketocracy took a good idea, found flaws in it, improved on it, and I’m sure continues to improve. But it’s not a poster child for crowdsourcing. It has only a two-star Morningstar rating and assets under management of less than $20 million (last year it lost a whopping 46%). So far this year it is beating the S&P 500 handily (28% vs. 20% as of October 23) but trailing the S&P MidCap 400 against which Morningstar benchmarks it; MID is up 30%.

No comments:

Post a Comment