Stockbot v1.0 activated – focused on inter-day trading. Simple regression based trading against S+P with a heuristic that brackets using historical standard deviation from the equity’s 200 day moving average. Updating indices every hour. v1.1 will have a new bot focused on intra-day trading.
I’ve created a simulation written in Java focused on trading stocks. Each Stockbot in the simulation will have a specific trading strategy and heuristic by which they trade. The first is a simple regression based trading algorithm. The Stockbots will log their performance and I will iterate on the heuristics they use. Ultimately, I think providing an “analyst” recommendation of the 1 year target for a stock + using the current price as compared to the 200 day moving average will provide the best results.
For intra-day trading, my basic thought process is to log the previous day’s prices over each minute of the day. Use the regression formulas in play for the inter-day trading, but identify the previous day’s mean and standard deviation and make trading decisions based on where the current price falls (within 2 SD).
Ultimately, I’d like to implement a rules engine based implementation (using Drools perhaps), but haven’t been able to wrap my head around the existential first-order logic axioms that would be needed (yet).
Other metrics that could be added as heuristics would include: price-to-book (as a historical benchmark), 200 day high/low, and eps.
The current use case target is each Stockbot will notify (report) trades it believes will be successful. Based on their historical performance, I will determine if I will listen…