I didn't mean to like anything. I was just reading through the thread earlier, must have accidentally clicked it.
). Initially my interest was the use of ML predictions from another very traditional and what I saw as a non scientific approach, training response from prescribed loads set by coaches in endurance sports.... long story but very interesting things coming in that field, like many others with ML applications. Some coaches have already taken the dive and the outcome has been evolutionary rather than revolutionary so far but very early days yet. That of course led to doing different tuts from all over the net on many different subjects and of course a good look at how it was being used in the trading world. What is available out there on the net from what I could find was truly woeful. Stuff like feeding in EOD OHLC data into scikit learn algos and hoping to predict next day being up or down. Just sh1te!!
Dude that looks like cherry picked nut job site from the fringes of the net. You cannot talk about the 'scientific processes' then quote that. But anyway..... ML and the cloud of doom coming to the keyboard warriors.....
I have spent the last year and a half learning Python and ML applications towards trading (its been cool as an old dog learning new tricks and I really nerded out
The deeper I went looking for applications the more and more disappointed I became with ML likelihood at being a revolution for the single guy on a keyboard. Once I had experience with what was possible with simplistic applications of ML it became clear that it was no different from any other angles at cracking the market. You are going to need may hours of dedicated learning, a very thorough background in market fundamentals other wise you'll will waste a heap of time trying to make simplistic and unrealistic applications/ideas work and mostly, as always in the market, a unique approach.
I suspect the pessimism from system traders about the future is the result of them now playing in a field that has had so little barrier to entry for so long. The EOD systematic trader looking for an edge with low risk over a few days by running decision tree algos is now 40 or so years old as a field. At the start you were a first mover with access to a lot of low hanging fruit. Try and imagine how many backtest have been performed on EOD and 1 min bar data since the start of the PC era!!! Talk about a carcass being picked over.
Python is a very cool language to learn and play around with but its no 'insider knowledge' tech. I have had a basic look at what a few Hedgies are doing in Melb with ML too and from what I was able to guess its more about risk and metrics than uncovering a new edge but when you are trading many many mil maybe that where the edge is?? Its another tool added to what has come before it. I'll continue to investigate what I can do with it. I have now moved far enough away from the beginners stage that I can apply my own unique approaches to collecting, assembling and searching data. From my discretionary trading results that is where the profits are.
What is disappointing in ML & the trading field is how little there is to find in true applications to trading. Useful ones. Unfortunately this thread adds to that.
Hello and welcome to Aussie Stock Forums!
To gain full access you must register. Registration is free and takes only a few seconds to complete.
Already a member? Log in here.
