What can you do relying only on technical analysis or even the weirdest features you could image?

It ain't much, but it's a CAGR that comes with low turnover, low volatility and low beta

Only factors about price and/or volume

There is less bias

The beta and the volatility is low so we have a high sharpe ratio

The number of stocks is 50 and the liquidity screen of the universe is 50k so you can invest some money

Edit:

You can profit more by training and trading in the Europe:

You needn't trade in Turkey to earn that:

Edit2:

Even models trained on the U.S. microcap universe perform well on European stocks.

1 Like

When I discovered that patterns could be found even from 19th century level data or pure price data, I not only wanted to push the limits of modeling: is it possible to get patterns from the strangest features? Can someone predict future stock returns if they only know the country, sector, market capitalization, stock price and ticker of a stock, and other information that seems very unlikely to work with just that?

In fact, it seems that they can.

It also works in the bigger universe:

Is this because of the composition of the universe?

No. In fact, testing this question further proves the stock selection ability of the strategy.

If this isn't a data breach, then it does break my perception, as I previously thought that I couldn't get enough information from these features themselves to do actual trading

Edit1:

Edit2:

Edit3: Update