After an unproductive search for effective AI factors, I decided to try those in my Dividend NG DM model. The universe is based on the S&P500.
Wow, what a surprise!
The predictor (linear) used a 1-month look ahead and the training period ran from 2004-06-29 to 2018-06-29.
Placing the two holdings with the highest returns (NVDA and AVGO) on the Restrict Buy List yielded;
Well done, p123!
My model making for microcaps also seems promising in the similar universe and the same period/number of positions:
After excluding the most profitable DVN, PCAR, XOM, CVX, MRO
So it is actually not a good idea to exit your positions in the past few years even if you invest in large stocks.
Edit: I chose a wrong period. If I chose the correct period, the situation would be like that
Differences are mainly in 19-20
