Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I am currently doing Andrew Ngs Coursera on machine learning to update my comp engineering skills and potentially find some trading ideas. Only half way through but a lot of the intelligence seems to be a misconception... Machine learning uses Algos and historical training data to make a best fit prediction... usually for much slower moving relationships (ie house prices). It currently struggles when taking in a large number of parameters with little data (think 5 minutes of order flow in the eminis what the parameters would be) to give meaningful predictions. It will probably be quite a while before they can figure out the intraday trading game which is a much more dynamic changing dataset.
Having followed the poker AI vs Human game. My intuition is that there was a lot of human input to the AI between sessions (they were able to update their Algos each night). That is where the battle will be in trading for purely discretionary traders. Discretionary traders tuning machine learning Algos intraday with large computing resources will most certainly be hard to compete with based on the poker match outcome. So... Why not learn a bit about it (ML) and employ some AI in your trading plan eh?
If you are not profitable as a trader, it is very unlikely that adding AI/ML will make you profitable.
If you are already profitable and you know what is your edge, then it is not that difficult to come
to an even more profitable system. The reason is that you will know what features to feed into
the system.
Naturally would make it easier for experienced traders to input features for their system. If you aren't profitable then that is not to say ML will help... Could be a million reasons why not profitable... But learning to be the best at using the plough (farming analogy again) in order to be profitable might not be good enough with the rapid growth of programming / trading automation, and still not produce any observable results either. Always need to maintain a broader view of what is at play in the markets (ML, discretionary etc.. market makers ). Thus reasoning for getting upto date on ML, learn to program a little etc.. or at least learn to use ADL which will no doubt include function blocks for the gradient descent / regression algos used in ML soon enough.
Definitely makes sense to learn to trade the order flow / price action or whatever starting out, but as the game changes, you have to learn to use modern tools as well E.g learn order flow or value spreads... whatever works for you.... And then figure out a way to employ automation in trading to compete.
until the point .. if ever.. general AI exists, will always be muppets behind the machine / coding part whole make mistakes... input incorrect bias etc.. so retail traders have a chance perhaps to be that unique consistently profitable trader.
I had a similiar discussion in my previous life as a pilot. Could AI replace pilots? Planes can take off and land and navigate without pilots but they can not handle the nuances of which clouds you can fly through and wich ones that you can't along with a multitude of little decisions that humans must make. Trading is very similiar. A well trained human will always be able to benifit from small nuances in the market that as of now computers can not detect. Thats why in my opinion most automated systems fail. They simply can not read a change in market conditions.
At the end of the day, AI trading is all about pattern recognition. They decide what to do based on past experience with the same of similar set of variables. Whatever AI trading does will be reflected in price like all the other market participants.
I predict that markets will become more predictable as the players (AI) will be more rational.
So in 2014 88% of transactions in the eminis had an automated order on at least one side of the trade. This could be just an autospreader but interesting to ponder how much ML volume contributes. Definitely firms publicly doing well in this area so worth staying in the loop.