InTrade: A Harbinger of Prediction Woes to Come?

April 7, 2013

Key word search is not to useful when there are trillions of content objects. Clustering trillions of objects is not economically feasible, so the sets are trimmed. Who’s to know? Predictive analytics sounds so darned promising because “real time processing” is cheap, plentiful, and trivial to boot.

What can go wrong with text processing, text analytics, social crowdsourcing data, and the other Lone Ranger silver bullets? How can predictive systems come back and bite a user, an investor, or an employee who loses her job?

I suppose that the article “InTrade Announces $700,000 Cash Shortfall And Risk Of Imminent Liquidation” describes an anomaly. Here’s the key point in my opinion:

..the company has posted the following message on its site, which says that it has discovered a $700,000 cash shortfall that must be rectified immediately in order to avoid liquidation.

InTrade, is or maybe was, a prediction market. The company says:

It’s a market that allows you to make predictions on the outcome of hundreds of real-world events. Stock exchanges find the price of stocks, and futures markets find the price of commodities. Prediction markets find the probability of something happening – a predefined, uncertain future event.

InTrade is more than voting. The company uses a range of methods to answer yes or no. Life should be so simple. The company even posted some Golden Rules to make the system almost foolproof; for example, “If you sell shares you profit if the market value of the shares goes down. Your profit is maximized if the market settled at $0.00.”

Eel bites can be painful due to “alien style jaws.” Investors in some predictive outfits may experience similar bites.

There are many meanings for the word “prediction.” I don’t want to get into a squabble that InTrade is one type of prediction and an outfit like Digital Reasoning or Agilex is another type of prediction. I want to capture several thoughts so I can include them in my text analytics lecture later this month, chance willing, of course:

First, predictions are slippery eels. I once offered predictions to my clients. Now I offer clients. I learned that regardless of methods predictions jump into a murky pool and get lost. Stick your hand in the pool and one can come up with nothing or an eel clamping on the extremity. Ouch.

Second, predictions and various methods and the companies built upon them can simply fail. Why not predict that? I think that getting hoisted by one’s petard is part of life.

Third, InTrade may be one example of what can happen when hyperbole outraces the capabilities of the numerical recipe crowd. Will other companies in the fancy math business suffer similar fates? I don’t know. I won’t predict.

If you are into fancy math, why not plug your retirement nest egg into one of the analytics outfits and let me know how that works out for you. Azure chip consultants, feel free to weigh in and explain to me and my two to three readers how such a clever idea could end up in a pickle of reality.

Stephen E Arnold, April 7, 2013


Comments are closed.

  • Archives

  • Recent Posts

  • Meta