Common issues found in academic papers about the use of machine learning in trading include among others, data snooping, p-hacking, selection bias and above all in many cases lack of skin-in-the-game and understanding of the field. In this article we include a specific example of a recent paper. Access to article requires Premium Insights subscription.
I had a conversation with a finance professor a few years ago and he referred to the extreme pressure on academics in all fields to publish articles for tenure credits. Finance proved a good field for many academics coming from engineering, computer science, physics and math because they could apply the concepts from those fields with slight modifications and publish papers for tenure credits. The result has been an exponential rise in papers but in reality there is nothing there but just dealing with old problems with obfuscated math. There have been notable exceptions of course but the bulk of production is noise.
Let us look a specific very recent paper about a machine learning application to trading and go over the common errors found in most papers of this kind although the introduction with the math part always looks impressive.
This goes beyond an exercise in finding errors in a specific paper and involves compiling a list of issues that must be often tackled during strategy development.
Specific disclaimer: This report includes charts that may reference price target levels determined by technical and/or quantitative analysis. No updates to charts will be provided if market condition changes occur that affect the levels on the charts and/or any analysis based on them. All charts in this report are for informational purposes only. See the disclaimer for more information.
Disclaimer: No part of the analysis in this blog constitutes a trade recommendation. The past performance of any trading system or methodology is not necessarily indicative of future results. Read the full disclaimer here.