I think there are two cases when running some automated trading software: 1) HFT or relatively HFT algorithms (ex: scan a lot of instruments, trigger on transient condition), case when there's no other way but to rely on fully automated execution, since humans can't monitor enough instruments, can't trigger fast enough, get bored watching monitors for days only to react within 10 seconds etc. 2) Low frequency statistical algorithms, usually running on higher amount of capital per trade. Book example: the wheel, where you infrequently do a few trades of high value. in case #2 Is it too much of a hassle to have the software just initiate the trades, but then manually make sure they get executed properly? Like when trading $100k, probably doesn't hurt to watch the execution, since you only need to do this once a month. Otherwise seems to me it's quite tricky to build an engine that can be trusted to completely automate the execution of several high volume trades. So much shit can go wrong: market moves away and you only get partial executions, connection gets lost, application crashes just at the sensible moment etc.
You just said it, lot of people on this forum are only daydreaming to build fully automated system, unless you are an a very large instituation with unlimited resources.
Build strategy first. Execute manually to see that you are printing money. Once printing, rest is easier. Hire coders to help you, you can even hire babysitters for your program
Totally disagrees. I have since 2014 built a totally automatic trading system, which did about 1.8 million trades last year. That is far from a high frequency system, but not a particular high amount of value per trade either. The main advantage is that it keeps the "psychology" out of trading. It has during these 7 years given me a lot better result than if I had executed the trading myself. At least I believe so.....
I totally agree with ElTrado. I did the same. Started in 2009 by automating the signal/model first. 2 years later I automated the order/risk management part. It's running just fine since then. It just needs one manual intervention per week, to restart the IB Gateway on Monday mornings.
See this thread/journal and decide for yourself (small capital, but it works): I wrote a bot and gave it 25k to trade
agree with el trado. I was part of a 2 man team about 7-8 years ago trading ~1M shares a day. Architecture and software design is key. also, the people watching monitors all day are not traders in "autotrading"
I agree with this. You need to be monitoring all the time if you are trading full size. Things can get out of hands quickly.
Same here: developed automated trading some 6 years ago. Since then running without issues, besides from a weekly restart. Additional benefit for me: I live in SE Asia and would need to wake up in the middle of the night to trade near the EOD in the US. With automation I can sleep properly. Thus automation is beneficial for my health.
Not really. Just add boundaries, and have the system shut down or close all positions if thing goes outside of these. I leave my system on when I leave for vacation, without any possibility to control it. It has never been a problem.