> They were all found by using phrases like "predict stock market" or "predict forex" or "predict bitcoin" and terms related to those.
Yeah, searching for any finance papers with "predict" or "machine learning" is literally the lowest quality tier you can get. These papers are often written by grad students who can pump an easy paper out by "applying" some already known ML algorithm to financial markets. Of course it's not gonna work. It also kills me when I see ML models who need stationarity assumptions applied to non-stationary time series data. Yeah, good luck with that.
THAT being said, there is lots of high quality research which has been replicated over and over, showing that alpha does exist in the market (and which funds have made billions off of). I would like to see the OP try to replicate some of these instead. To give some simple examples:
1. Try searching for papers with the keywords "and the cross section of expected returns". For example, the momentum factor which can be tested and replicated with only linear regression. > There is substantial evidence that indicates that stocks that perform the best (worst) over a three- to 12-month period tend to continue to perform well (poorly) over the subsequent three to 12 months. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=299107
2. Statistical arbitrage strategies which were known to work well until the mid 2000s. Also been replicated many times, furthermore, you can see the gradual decline in profitability pointing to the theory that "alpha decay" in this case is real. https://www.math.nyu.edu/faculty/avellane/AvellanedaLeeStatA...
3. High frequency strategies. No way OP or any retail trader can replicate this, but firms make billions of dollars per year consistently doing this.
In conclusion, to make a claim that there is no alpha in the market seems highly suspect, and perhaps just needs a more nuanced view of how trading firms make their profits.