I once spent a fair chunk of time trying to utilise the ML algorithm k-nearest neighbour in assisting me to try and make a more “refined” approximation about the probable magnitude of the volatility one period forward (T+1) on FX pairs. All this is, is a fancy practice in which data in the past (e.g. close price changes) is used to try and associate a data point “now” in an attempt to build relationships and assumptions about what might happen next.
I’ve still yet to draw any really hard empirical conclusions on whether it’s helpful or not, but what I can say for sure is that although it may urge you to delve into machine learning algorithms (not sure if it is or isn’t!), you still have to work damn hard in your research process because it is nothing more than a tool, not a short-cut by any stretch of the imagination.
There will of course be hedge funds who utilise these kind of tools for the task at hand. Conversely there will be hedge funds who do not see the purpose nor fit.
Trading is not so much about the tool used but rather the empirical evidence in the underlying research that defines a robust foundation toward profitable and confident trading.
As a side note: here is a fantastic video about algorithmic and machine learning trading: Algorithmic Trading Strategies with MATLAB (really for the technical minded with a background in programming and statistics).
There’s a lot I could talk on this topic but it stretches beyond the realm of just a few paragraphs. In the coming weeks and months I will aim to delve into the aspects of how one can define a winning strategy the steps you can put into place to take your understanding of profitable trading to a whole new level.