Posted by: zyxo | November 16, 2008

Piqqem : Prediction market for prediction errors


NASDAQ in Times Square, New York City.Image via WikipediaA “prediction – prediction market“. That is what piqqem is. Does it makes sense to create a prediction market to predict the predictions of another prediction market? After all, that is what stock markets are : prediction markets.
(see also the discussion by Erick Shonfeld on techcrunch)

I think it is an interesting idea.

A basic technique often used in data mining is : increase the prediction accuracy by combining several predictions, obtained by different methods. This can be done either by averaging independent predictions or by using for example a second method to predict the residual error of the first method.

Piqqem uses this second technique. The difference between the current stock price and the future price is the error on the prediction of the stock market. It is this error that piqqem wants to predict, using the wisdom of the crowds.
So it makes sense : predicting the errors of the first method by a different method.

Of course if i) this works and ii) this gets widespread, the info of the piqqems of this world will influence the stock prices reducing their added info to nil.

Reblog this post [with Zemanta]

Responses

  1. thanks for blogging about us, Monsieur Mixotricha/Zyxo. here at Piqqem, we’re also very interested in complexity and stuff. and you may be interested in reading a critique of Schonefeld’s Techcrunch piece by Will Hankinson, our own illustrious engineer:

    http://simianlogic3d.com/blog/2008/11/10/who-says-the-crowd-has-to-be-right/

    and another thoughtful critique by a fellow named Mark Diller with zero conflict of interest:

    http://communitymusings.blogspot.com/2008/11/still-struggling-to-understand-wisdom.html

    either way, thanks again, and we’d Love for you to join the crowd at Piqqem.com!

  2. Hi Zyxo,

    Thanks for pointing out piqqem, it is truly an interesting idea…during my ventures of predicting the financial markets with combined data mining, information extraction and text mining techniques i have found that there are patterns on investor reactions that keep repeating over and over again…but the true challenge is in quantifying this reaction

    Regards,

    Themos


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: