Who can forget Long Term Capital Management? The Connecticut based hedge fund was the talk of the financial world of 1997 when two of it’s directors won the Nobel Prize for Economics. It was also the talk of 1998 when the fund lost $4.6 billion in four months. The fund closed in 2000. More on them later.
Research surfaced last week from a start up based in Utah that could predict elections in the United States. The company’s algorithm analyzed sentiment across Twitter and other social networks to predict the outcomes of 90 primary races across the country in the last month. The races were across Senate, Congress and Gubernatorial races. In 87 percent of cases, including those that traditional polling got wrong, the algorithm’s predicted winner was correct.
The start up is called PollicIt and is based out of Utah State University. It is run by Professor John D. Johnson and five students from the John M. Huntsman School of Business.
The results harken back to last Summer when a hedge fund emerged from nowhere that planned to use Twitter to predict market sentiment and deliver returns to its investors. The fund lasted just one month but not for lack of performance.
The fund was called Derwent Capital Markets. It also claimed 87% (87.6% to be exact) accuracy in its predictions but these were not being made of the election. They were being made about the Dow Jones Industrial Average. In its first, last and only month the fund gained 1.85% beating both the market and most hedge funds in last year’s dreary Summer.
The firm closed when it realized it could get out of the hedge fund game and into a broader business on the advice of one of their biggest investors. The company’s web page is now a blank asking visitors to follow them on Twitter to await their launch of some kind of market sentiment service.
All of this has happened before. We return to Long Term Capital Management. When they started trading in February 1994 they had attracted over $1 billion in assets. The firm employed revolutionary mathematical models to fixed income arbitrage and eventually spread into merger arbitrage and finally the equity markets.
They performed extraordinarily well for some time and with two Nobel Prize winners on the board nothing could go wrong. But things did go wrong and they went badly wrong. In 1998 Russia was on the brink of absolute collapse and the world markets responded in a frenzy. LTCM began to shed money and ended up having to be bailed out by the Federal Reserve Bank of New York.
True there are differences. The new trend in Twitter predictions is based on current sentiment rather than historical data. There is a lesson to be learned however. Web Bot is a company that releases predictions of what will happen in the future using complex analyses of big data using their own secret algorithm. The company has predicted that there will be a major catastrophic event at the end of 2012.
Their prediction is often given as a proof of a coming Mayan Apocalypse. The data however is self reinforcing. If people talk about a catastrophic event the algorithm adds this into its predictions. The algorithm’s results are taken as further proof and so on.
Market sentiment is a very finicky measurement. Most people’s real feelings cannot be compounded to 140 characters or less. Twitter may be accurate 87% of the time in stable conditions and that is fine for elections. When it comes to the market the big shock that will shake the predictions is the one that will wipe you out.
Use your head not your Twitter account to predict the market’s movements. Better yet listen to the best of the best and see what the Gurus are saying.