Jan

26

The key thing about markets is that as soon as the algorithm "solves" the problem and big players start trading off that, the trading itself changes the nature of the problem. It's not just that markets have a much larger "game space" than chess or Go, but that every move in the game changes that game space.

Julian Rowberry writes: 

Machine Learning is just optimizing a solution to a problem, but with a lot of data. The solution still needs finite data and to be solvable. There's just too much data in markets to plug some data into an algorithm that optimises what you're feeding it to predict where it's going.

Useful for stuff with limited data like; where to route orders to which exchanges and when, setting a postal route, or self driving cars efficiency and safety. Perhaps the best way make money out of it in markets is to look at which companies are using it smartly with limited data sets and avoid those who are trying to use it for things it can't do, or using it as PR, and avoid them. There's something you could test.

Larry Williams writes: 

The take away from my efforts in this was there is too much randomness in the data for anything to be learned.
 


Comments

Name

Email

Website

Speak your mind

2 Comments so far

  1. Anonymous on January 29, 2019 11:22 am

    This is the game that moves as you play

  2. Jim Davis on February 6, 2019 1:19 pm

    The best edges are structural edges.

    Sadly, most , if not all of those are now very expensive and not available to retail.

    SOES , years ago , a good example

Archives

Resources & Links

Search