With the explosion of data and vast computing capabilities, machine learning and artificial intelligence are all the buzz these days. However, a serious consequence of this is the ease of producing a wonderful looking backtest that fails to deliver when put into practice. In statistical parlance, this is called overfitting, or a situation where researchers have merely captured noise in data. This is believed to be largely responsible for the claim that “most claimed research findings in financial economics are likely false.”
An outstanding paper written by Spring Valley Asset Management gives context to this claim by demonstrating that seemingly inconsequential changes to an investment model can introduce substantial risks when put into practice. It then explores a novel approach to mitigate these risks. Whether you design, or allocate to, systematic investment strategies, the concepts described in this paper are of immense importance. I hope that you’ll find this paper as educational as I did.