Using Bernoulli Models to Classify Trading Outcomes

Summary

The article represents trading histories as sequences of outcomes and applies probability models to describe them. It first reduces trades to an equity curve, arguing that different trade arrangements can be equivalent when they produce the same curve. A two-state model uses the Bernoulli distribution to describe sequences such as gains and losses. A generalized version counts outcomes across multiple discrete states and combines the number of possible sequences with each state's probability to estimate event probabilities.

The author suggests classifying trade samples into states and clustering them to evaluate a strategy, filter its results, or find profitable subsets. These are proposed applications rather than demonstrated trading results: the article provides probability formulas and illustrative diagrams, but no empirical performance evidence. Its approach assumes outcomes can be represented by a finite set of states and that the probabilities used to describe them are meaningful. The author notes that strategies with infinitely many states require separate treatment, and frames the method as a component for future work on adaptive trading systems.

Key ideas

• A trading history can be represented by its equity curve, which may correspond to many equivalent arrangements of individual trades.
• The Bernoulli model describes sequences built from two outcome states using their probabilities and the number of possible arrangements.
• A multinomial extension can represent trades classified into more than two discrete states.
• The method depends on selecting useful outcome states and estimating their probabilities.
• The author proposes clustering classified samples as a possible filter, but does not present evidence of improved performance.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.