Subject: Re: A one-year backtest!
showed that a US stock market trading strategy trained on more than 10,000 parameters and just a year of data beat a simple buy-and-hold benchmark.


Hmmmm...
"beat a simple buy-and-hold benchmark"

It did so for how long out of sample?

I don't care if it's memorizing all the data in the universe for the year up to a given date, and has to be re-run every month, if it demonstrated that it could keep on predicting what will do well the next month. If it works for six years after the system was built, sign me up.

Jim