Is slippage minimization even a tractable problem besides applying loose heuristics derived from empirical insights, e.g., identifying reliable early-signals of narrowing spreads and increased liquidity for a given exchange?

A few years ago I learned Python and RL by coding a high frequency trading agent for cryptocurrencies. In backtesting it was phenomenal. While executing though, trading fees, slippage and other factors negated all the advantages. It was quite disappointing, but I’m happy that it got me started on coding and AI.

It's funny how all these trading experts post nothing but great successes in relation with the latest hyped tech like machine learning. The reason for this is simply "It is difficult to get a man to understand something, when his salary depends on his not understanding it."

Some of them are honest enough to realize they're crooks, basically milking their employer for salary which keeps going as long as they shoehorn and manipulate those "backtests" into somehow showing great profits.

I seen these guys. None of these strategies work in real life, they're basically random occurrences on a very particular set of data. Zero robustness that is, you change anything in the data: the underlying index or stocks, the time period (ex: 2015-2020 instead of 2020-2024), the sampling period (ex: 15 minutes instead of 5 minutes), or just (and I was laughing my ass off seeing this), literally one single fucking sample off: start 5 minutes later in the backtest and everything blows to pieces.

And the ignorant public keeps drinking their kool aid because they want, they have to belive in Santa.