If you're building your own strategies with The Boss, there are three very important factors that this new generation of would-be AI billionaires is going to miss:
1) The strategies are extremely simple, so it's likely that these complicated neural nets, LSTMs, and recurrent networks are going to over-fit to the past. They are too complicated. The BSX strategy only has 2 lines of code.
It's like when Edison asked a new hire to measure the volume of one of the light bulbs...the student made all kinds of mathematical equations. Edison walked up, dunked the light bulb under water, and poured it into a measuring cup.
That's why I pushed my team into using LGP (linear genetic programming), which I suspect is why these edges are so strong.
2) Often times they often use limit orders above current prices so you get a good deal when you sell. More like a market maker.
3) They all switch into an inverse ETF at times. This not only makes the strategy more profitable, it makes trading a basket of them much more resistant to sudden shocks -- because there's a degree of hedging going on.
Not to worry you, but there has been more hedging than usual going on in these strategies.
If you'd like this strategy and 44 more, then this is your last chance. Here's why:
No comments:
Post a Comment