written by john lewis

iP.robo project

I’m a huge fan of iPredict, a New Zealand based prediction market, and have been dabbling for years now. Generally contracts are available on political or economic events as well as current events.

An over-simplified description of a contract is that it pays out $1 if that event occurs, or $0 if it doesn’t. This means the market finds a price that corresponds to the market prediction of that event occurring – if the price is $0.72 then it means the market is predicting a 72% chance of that specific contract happening.

For fun I wanted to see how an automated trading engine would perform if given a small set of rules to trade by. I developed iP.robo in CakePHP and it started trading on iPredict in February 2011.

Initially, iP.robo made some awful trades thanks to some shoddy logic on my behalf. This dragged the ROI in -10% territory but has since been optimised and iP.robo is starting to perform better. My benchmark for iP.robo is to see if it can beat a standard savings account ROI of about 4% p.a.

Currently ROI is: 12.15%

Update Feb 2014: Ongoing issues with the API mean the bot is offline most of the time.

Posted in: