Evolving Trading Strategies
Posted: August 19, 2007
Statistical models used to trade in financial markets breakdown and need to be retooled as markets invariably behave differently over time. Who's trading, what their long term and short term forecasts are, what patterns they trust, and how much money they each have all influence the ultimate ebb and flow of the market. Profitable strategies that detect patterns in market data and enter/exit trades accordingly always exist, but they change as the market changes. These detect-and-trade strategies are usually made up of the same basic building blocks but specific strategies stop being profitable over time. Genetic programming can be used to build detect-and-trade strategies that live, reproduce, and die every hour of every day based on their profitability in the current market conditions.
RallyFlag Trading
Of all the ideas here on petergroves.com, I've decided to start my first company based on this premise. By the end of 2007, I hope to be incorporated and trading algorithmically under the name RallyFlag Trading using the genetic programming software I'm developing. In no particular order, here are my reasons:
- My background is in developing predictive machine learning software, so I am fairly comfortable taking this on from a technical standpoint.
- Since there are no customers, I can make money as soon as my software works.
- The startup costs are next to nothing. A year's living expenses and around $10,000 for accountants, brokers, and equipment should be sufficient.
- Small startup costs means I can fund it myself, which means I don't have to spend six months trying to convince an investor I can make it work. It also means "failing" is cheap.
- I do still plan to start new companies as a career, but I've never done it before and expect a steep learning curve. While a proprietary trading firm may protect me from the whims of finicky customers and investors, I'll still learn how to deal with the lawyers, accountants, employees, office space, etc. I still think that's enough black holes to deal with for one startup.
Copyright 2007 Peter Groves. This text may be reproduced only in it's entirety in any medium without royalty provided this copyright notice is included.