Encoding trader 'horse-sense': ideas and experiments using historical foreign exchange data
Selene Makarios (Stanford)
"Trade with the trend."
"Cut losses short."
"Let gains run."
An important paradigm of classical A.I. is the formal encoding of common-sense information and heuristic procedures. The three trading dictums bulleted above, which we call ``horse-sense'' trading principles, are found explicated in many lay-audience treatments of trading, and elsewhere. We present experimental evidence that suitable encoding of the three principles can result in a parameterized trading model in which historically-inferred parameter values yield algorithmically-driven returns exceeding those of the analogous strategy with randomized parameters, that is, a strategy that ignores the historical data. We also present a related discriminator of historical forex price-charts versus price-charts generated by Wiener/Martingale processes. For the populations of historical and randomly generated price-charts tested, the discriminator rejects the null hypothesis of a common mean for the two different groups, at a 99% confidence interval. These results could be construed as evidence against the Random Walk Hypothesis. Finally, we describe an interesting relation observed between the performance of the heuristic-based system on our price-series, and the amount of information (actually the algorithmic entropy) in the series, as heuristically determined.