Library for implementing the "Plateau Finding", a technique for trading systems optimization.
plateau finding is an optimization technique that, in the parameter space of a trading system, looks for "plain areas" in that sapce (a surface for the case of 3 dimensions) in order to pick the highest of the points in the area and consider it as a local optimum, and use it in the trading system with the benefit of out of sample stability in the performance metric defined. All of this in contrast with "hill climbing" type of optimization algorithms, that look for a "high peak" (global optimum) and present high sensibility to parameter value changes and reflect that in a greater out of sample deviation of values found in sample. In a trading context, is better to have several "neighbooring" configurations that produces "good and stable" results all of them, than, having one configuration with "the best" result but with neighbooring configurations producing "poor" results.
Instructions for running this project.
-
Install dependencies
$ pip install -r requirements
-
Select parameters
In the data.py you can select the input parameters for the examples.
Msc Juan Francisco Muñoz Elguezabal - franciscome@iteso.mx