anchak….
I don’t have time for hypotheticals but I did several backtests in different universes. The backtester I used includes dividends, slippage, splits, name changes and lists stocks that it can’t find at sell. As for stocks missing at sell time you can assume you have a 100% loss on every one and it doesn’t change the conclusion because there are close to the same number in the high price selections and the low price selections. Another interesting factor is holding the higher priced stocks also results in a higher sharp.

First of all, I’ll acknowledge that price as a factor doesn’t work at all in the S&P. But it works both in the US and global stocks and especially smaller caps.

I’m not sure you can get to the paper but “Machine learning goes global: Cross-sectional return predictability in international stock markets” The abstract “We examine return predictability with machine learning in 46 stock markets around the world. We calculate 148 firm characteristics and use them to feed a repertoire of different models. The algorithms extract predictability mainly from simple yet popular factor types—such as mo mentum, reversal, value, and size. All individual models generate substantial economic gains; however, combining them proves particularly effective. Despite the overall robustness, the ma chine learning performance depends heavily on firm size and availability of recent information. Furthermore, it varies internationally along two critical dimensions: the number of listed firms in the market and the average idiosyncratic risk limiting arbitrage.”

Across all countries using 11 different Machine Learning models the found price was 11th most significant factor in predicting future price.

Might not seem logical and I don’t understand why except perhaps failing companies usually slowly decline in price before they disappear and very successful companies don’t have to lower their price to attract investors. But I’m just telling you what works using machine learning on historical data tells me and many others.

By the way how can ML commit the price to memory of an individual stock? ML training data has been stripped of all stock ID’s and Ticker information before training.