According to this study, in summary, momentum investing works on paper but does not work in practice. https://www.bloomberg.com/news/vide...h-says-about-momentum-investing-returns-video
Ask Mark Minervini on twitter. He did not graduate high school and seems to have done well with Stan Weinstein stage analysis.
The video was not really informative. The paper by the dude explained the ideas better. Looking at figure 2 and figure 3 will be sufficient to understand it without reading the full piece. The crux of his argument is that the popular factors like value, momentum, size, illiquidity, low beta, profitability may not have any structural excess return and are a product of changing valuations (P/B ratios). He identified valuation of gross profitability, momentum and illiquidity as stretched, while value is cheap. Momentum index may not in fact be replicable because of rapid stock turnover and low beta lacks statistical significance between valuation and subsequent 5y return. Roughly same conclusions are drawn for smart betas like equal weight portfolio, fundamental index, risk efficiency, low vol, maximum diversification, quality. I am slightly dissatisfied that in conclusion the author didn't give a binary opinion if he found an excess return or not but only cautiously warned that 10-15 years of data as many product promoters use for analysis are not enough and that value could be cheap. I am somewhat sceptical of all this factor methodology in the first place. If the questioned was framed not in P/B terms but if banks and natural resources were a better investment opportunity than tech and healthcare, would the answer be as easy? Maybe we should read Damodaran and do his cross-sectional regressions of P/E versus expected EPS growth instead of bothering with those time series factors that give us an answer we already know (yes, banks and basic resources are considered cheap). Is it appropriate to compare P/B ratios in 1967 and 2015? Surely there could have been some secular changes due to changing sector weights, accounting rules, composition of balance sheets. Is okay to regress 5y returns on valuation level instead of valuation change when dealing with time series?
I also take what Rob Arnott says with grain of salt. He is a value guy and his firm pushes fundamental index etf's. He is also founder of the "Research affiliates". There are two types of momentum, Cross sectional and absolute. What he is referring to cross-sectional for US equities. True, cross sectional is prone momentum crashes during worst of the time but there is premium. First momentum paper came out on 1993 ("Returns to buying winners and selling losers, implications of stock market efficiency" by Jagadeesh and Titman), cross section momentum premium is still positive. His definition of mutual funds with momentum not exhibiting no premium is also should be taken with grain of salt because of they are long only (Momentum factor is defined by UmD, Up minus Down). Momentum does exist in all asset classes (Stocks, bonds, currencies and commodities). For implementing it is better to use all asset classes, combine that with both absolute and cross-sectional momentum. Most of the trend following managed futures uses time-series (absolute) momentum http://pages.stern.nyu.edu/~lpederse/papers/ValMomEverywhere.pdf
It's probably true that such a general rule isn't going to be far from evens , after all nothing could be easier than following momentum , buy when goes up sell when goes down , there may even be something in it but not much otherwise we'd all be rich.
What, we're not all rich? ET is filled with traders who are rich, getting richer, and have algo's running 24/7 minting money.