Astrological Machine Learning for Astronomical Profits

ph1l said:
Thanks. I'd name the hedge fund something like Fractal Recurrent Assets Underlying Dynamic Systems:)

What you have done with your FasterQuant software is most impressive, and I think its only a matter of time before some institutional investors use it.
More...
Man...that name is awesome and thanks for the kind words about FasterQuant!
 
Even though the training data spanned more than 55 years, I'm not too surprised a genetic program strategy with so many inputs could show good results after training.
upload_2019-7-23_9-10-37.png


So with six independent models trained on the same data, I ran a test on an out-of-sample period of 1238 input instances for 1957-12-26 through 1962-11-23. Independent means the random number generator used for the training has a different sequence. When 3 or more of the top 7 rules for any of the models had a signal, the simulated number of trades was 266 with 125 wins (a win is greater than 1.6560 percent gain) and had a mean gain of 100 * exp(1.2726 / 100) - 100 == 1.2807 percent.

For this test period, the mean gain for trades taken on all days was 100 * exp(0.519945877769271 / 100) - 100 == 0.5213 percent. The test results were significantly better than the mean result for the entire test period. So maybe this type of system is almost usable.
 
ph1l said:
Even though the training data spanned more than 55 years, I'm not too surprised a genetic program strategy with so many inputs could show good results after training.
View attachment 206167

So with six independent models trained on the same data, I ran a test on an out-of-sample period of 1238 input instances for 1957-12-26 through 1962-11-23. Independent means the random number generator used for the training has a different sequence. When 3 or more of the top 7 rules for any of the models had a signal, the simulated number of trades was 266 with 125 wins (a win is greater than 1.6560 percent gain) and had a mean gain of 100 * exp(1.2726 / 100) - 100 == 1.2807 percent.

For this test period, the mean gain for trades taken on all days was 100 * exp(0.519945877769271 / 100) - 100 == 0.5213 percent. The test results were significantly better than the mean result for the entire test period. So maybe this type of system is almost usable.
More...


With the monkey meme, you might enjoy the concept behind -

https://libraryofbabel.info/
 
ph1l said:
I guess that way would save a lot of monkey chow and typewriter ribbons.

But if the library was ever completed, this might happen.:)
More...

Not only have you confused many readers here with the concept of "typewriter ribbons", but then you introduce the concept of how God could not kill the universe because of physics?

You are a crazy man.

 
You must log in or register to reply here.
Back
Top