Geoscientist vs. the prediction markets
Alan Jewell November 15th, 2006
Chris Masse posted earlier about my blog and then graciously (and concisely) asked me to be a guest author on Midas Oracle. I believe his words were:
- Who the hell are you?
- What the hell is CASTrader?
Lo and behold, here I am, honored to be in famous company, and at Chris’s suggestion, I’ll answer those questions in my first post. My name is Alan Jewell, and I’m also an engineer, computer programmer and a lone-wolf investor in Colorado, USA. I’ve been writing software over a span of 20 years to use time series data to predict the location of oil, a dam’s response to earthquakes, and the water flow from a mine adit, among other things. I’ve also examined relatively obscure techniques like Shannon’s Demon and the Kelly Criterion to make money in a favorite prediction market, the stock market.
Enough about me (or “moi” as Chris would say), what is CASTrader? CASTrader is a software system under development to trade the prediction markets (financial for now) by quantitative means. It is designed to use a bunch (1000 now, possibly millions later) of diverse software agents (traders) to trade the market collectively. Version 1.0 showed some success in finding tradeable market price patterns in the venerable Dow Jones Average but is not ready to go “live” with real money. Hopefully, a new version will be ready January, 2007.
The next version will feature a near frictionless (except for the bid-ask spread) internal, alway on Continuous Double Auction (CDA) dark market. This internal prediction market will be connected to the real market by arbitrage agents that profit from trading the price discrepancies between the two markets. With any luck, this “net trade” will pump profits from the markets rather than the other way around. The bad traders will die when they run out of money, and the good ones will survive and eventually “mate” via genetic algorithms, producing better and better traders over time - all automatically. Right now, though, I’m focused on coding the market and market makers.
I am a believer in Andrew Lo’s Adaptive Market Hypothesis (AMH) which allows for excess returns to traders and arbitrageurs in return for their services of making a prediction market more efficient and likely more accurate and liquid. I think the AMH is a good way of describing how any prediction market will evolve over time. I also believe market price fluctuations aren’t as random they appear if one works at predicting them.
The long-term plan is to build internal prediction markets for virtually anything (economic data, elections, etc.) that impacts financial markets to allow the agents to hedge. If arbitrage opportunities exist with other prediction markets, I would eventually explore connecting CASTrader to them as well.
For those interested, there is more in a recent post roundup, and I’ll certainly add my relevant thoughts here from time to time from a prediction market developer’s and trader’s standpoint. Thanks to Chris for inviting me to be a part of this.
- All Guest Authors's Posts , Analysis (Meta) , Mechanism Designs , Software
- Comments(0)
