I can't speak for anyone else, but I run fully automated trading for a rather small retail account (<500k USD account with T-Reg margin). Rather small because it is small enough to exploit certain edges unavailable to players with big pockets. Having said that - account size is certainly an edge. More details available in my journal. One single month IB statement was posted here. Speaking only for myself - there is certainly little motivation to prove anything when it comes to sharing here, so I can see why those rare profitable traders everyone is looking for might not be sharing any statements. For the snake oil salesmen, on the other hand, there is no excuse. As selling something and not backing it up with a solid evidence why it is better than competitors is just playing on human weaknesses and emotions. Talking about over-fitting and automation - time and live executions will show if that's the case. To mitigate overfitting risks, in general, it helps to stick with systems that have a lot of trades and fast feedback loop. My systems made >1500 trades this year and backtests included tens of thousands. PS. I am personally a bit skeptical about A.I. for picking trades, but not qualified to say it won't work. Val
AI is just a meaningless buzz word at the moment, AI isn't real and likely never will be, so stay skeptical. How complex is your logic ??
Pretty simple and very complex depends on how to look at it. I'm guessing you're talking about trading system rules and degrees of freedom, then - each individual system is relatively simple and uses only 1-2 rules for a setup. That is on top of other reasonable generic liquidity/volatility filters and more or less common entry logic. There are 5 systems targeting different anomalies. There are rules on how they interact with each other but they are designed to be not very correlated and each work better in some specific type of market. Basically this is a portfolio of trading systems. Execution/trades management are of a medium complexity. All trading routines are automated. I just check daily reports my software sends via Slack and have some weekly / monthly research to make sure everything is within model's parameters. And I don't touch AI for anything trading related. PS. Didn't mean to hijack this thread. Happy to answer any questions in the journal. Val
As a programmer I understand your logic, nice way of doing it linking methods wise, using multiple stocks I guess so rather than 1 index, give an edge a better over all performance. I could code my entries but exits are the issue especially with a tight SL, all too often its take 6pts or 10seconds later it'll hit SL, no way I can see to code that. Unless I just take 6pts every time, but then miss out on 30pt moves, tricky!
I'm not smart enough to compete with big boys who trade indexes. Re your exit - sounds like a simple bracket order would do. You send it with your entry order, active immediately. IB also supports delayed activation. You can specify time when order need to become active. Might be handy. But I not entirely sure I understand your problem with automating exit SL. How often its' hit is part of the system design.
I'm glad you asked I started by downloading free data from different sources in the web. Mostly hourly / daily timeframes and backtested against it. Having done that, I connected the bot to the API of my broker and the results were far away from the good results I had in the backtest. Strange. I started to deep dive and figured out, that my broker had quite different data on the same day as the data from other sources. I should mention here, that my bots trade CFD - not really the futures on CBOT, EUREX and so on. So market prices and especially prices outside the cash market are for every CFD broker different. I contacted the support of my broker and asked for tick data - which I received. Then I did the whole backtesting procedure once again and YES - I also had to adapt my code to this new circumstances - especially I had to modify the exit strategy rules. Because very often you have spikes (up / down) outside cash marked which would kill your stops. Sometimes I had the feeling that my broker ran his own AI bot to hunt the stops of their clients ;-) Item. Done that, I ran the bots on the DEMO account for several month. I found some bugs, fixed them, put them again on the DEMO account and so on - until I had nothing more to fix. After a certain time I decided that the REAL test is to run the bots on a LIVE CFD account. So I funded the live account with some hundred euros - just enough to run bots trading the low margin instruments - yes, and thats the current state.
SL fine, it's profit taking, guess I'd have to move to a always in system but then chopped to death at times so tricky.