Log in

Can machine learning techniques be used to predict market direction? The million models test.


This is an interesting article by a high frequency professional about the machine learning techniques is general. This articles deserves all the possible consideration. Is he wrong? Or are we trying to do again and again something that has proven to be wrong, I mean to use Neural Nets, Support Vector Machines, Random forest,  Nearest Neighbor Classification

“Insanity: doing the same thing over and over again and expecting different results.” A. Einstein

 Or for us  this is just the way we gain experience ;).

On the other hand I think that this is completely normal, you can't just plot the data and tell to the neural net. Predict me the future! I have also tried this and that is the first think I thaugt when learning about it. But it just does not work like that. I think this guy is completely right but on other things he completely miss the point (or intentioally misses the point on the net you have good enough reasons to doubt the intentions of every author you read including me). 

Yet it is possible to find pockets of predictability and ther you could learn the characteristics with your machine learning techniques and then exploit the results. However this is truly the different thing and that is not only science but art too. This is really where the magick lies.

The Elliotware approach is a handy and simple approach to this problem.