Tuesday, June 2, 2009

Cold War

In the last post it was proposed how to model a computer trading system that would be capable of analyzing the behavior of all participants in the stock market, calibrating its own predictions, and making trades accordingly. Now we'll see the reasons why this method introduces yet more exotic behavior into the stock market, making it difficult to realize the profit that would otherwise be expected.

The first reason is simply the presence of intermittent trades with effectively random timing from the perspective of the software. The Leeroy Jenkins effect is most powerful against weak patterns. Stronger, more self-sustaining patterns will require a larger Leeroy event to upset.

Secondly, there are news events in the future that are effectively random. The initial impact will be beyond the design of such a system. Only after it receives new data that can measure the reaction amongst fundamental investors can the system recalibrate itself to meet the changing conditions.

Less importantly, the model written in the last post makes some assumptions. One is that traders continue using the same tools over and over again and that the popularity of indicators changes at a slow pace relative to trading regimes.

A very dangerous assumptions is that no other computer systems with unknown behavior and unknown influence on the stock price. In reality there are many such systems, and by getting a step ahead, all that happens is a new type of feedback in introduced. Since none of the computers know precisely what the others model is or what the trading implementation is, effectively all computer trading systems are tied up in a constant poker game. Dead reckoning and playing the cards right (from a programming perspective) might still be, on the whole, a reliable strategy, but the notion of predicting, much less controlling the stock market at the most granular levels to create wide-reaching patterns is far fetched.

Essentially, the picture of the stock market is shaping up to be a bit like a system of springs, weights, and dampers, with some of the springs trying to consciously understand the system, but because the knowledge of the system and interaction with it changes the spring, and thereby the system, this consciousness is always just out of reach. In short it's impossible to predict the chaotic dynamics of the market when characterization thereof influences participation, changing the system at its most fundamental level, trades and traders.