What are good results for daytrading
- Thread starter C++Programmer
- Start date
I have seen more than one "successful" trading system fail when put to the test. Just a month ago I was introduced to a system that was "earning" 3% a day! Sounds real great, eh? Well, when the rubber hit the pavement, it ended up losing about half a percent a day, on average. I think many new traders under appreciate the difference real world trading will have on their systems. Also, the "hidden optimizations" that dozu888 mentioned are very real, and very hard to completely eliminate from any system doing back testing.
That being said, I disagree with compaque's remarks (nothing personal, eh
). I think there are some successful systems are out there. I just don't think they are being sold. Like you say, if they work, why would they need the extra income?
Good luck to you, C++. Let us all know how the actual trades go!
Cheers.
That being said, I disagree with compaque's remarks (nothing personal, eh
). I think there are some successful systems are out there. I just don't think they are being sold. Like you say, if they work, why would they need the extra income?Good luck to you, C++. Let us all know how the actual trades go!
Cheers.
Which language do you guys think is the easiest for linking?? My guys have "opinions" based on their experience, not necessarily based on which is most error free. Any ideas??
I suppose with the supersoes in place, backtesting stocks on the nasdaq maybe gives something close to reality. on NYSE you can forget about it.
Thats why I think most intraday systems are developed on the electronic markets like the emini.
C++, try your system on ES and NQ, how do they do?
On the stock side, I do think scanning/pattern recognition has the upper hand and out of thousands of stocks, you gotta be able to find that perfect set up..... no experience in mechanical stock systems myself though.
Thats why I think most intraday systems are developed on the electronic markets like the emini.
C++, try your system on ES and NQ, how do they do?
On the stock side, I do think scanning/pattern recognition has the upper hand and out of thousands of stocks, you gotta be able to find that perfect set up..... no experience in mechanical stock systems myself though.
There are many applications that allow you to grab the data: www.mytrack.com, www.taltrade.com, to name a few. Getting the data is not really a big deal. Handling the orders is the bigger deal. This is particularly true in changing orders. You can cancel and put in a new one, but then you have to wait for the cancel confirm or maybe put the new one in without the confirm or maybe your execution software API has a replace function.
I use RediPlus and they have an API for quotes and orders and a DDE link for quotes. Ideally, you could use the FIX protocol www.fixprotocol.org. That way, you wouldn't have to make your app dependent on your execution software. But then you have to get a decent FIX engine and come up to speed on the whole FIX protocol. I do my programming in Excel VBA for the simple or Q&D stuff, sometimes full VB if I don't want the Excel overhead and don't need the grid. Or for the max speed, I use VC++ (which is also good for multiple markets at once as VB/VBA is not good at handling dynamically created controls like I need).
Your first choice is your execution software.
cashonly,
This is all real good info, thanks for the links etc. - you sound pretty serious about this, are you working on something? I am interested in trying to create some software that can do tape reading via neural nets or the like, I understand Ray Kurzweil has done some work in this area with his FatKat system (www.fatkat.com), does anybody know how well this turned out?
This is all real good info, thanks for the links etc. - you sound pretty serious about this, are you working on something? I am interested in trying to create some software that can do tape reading via neural nets or the like, I understand Ray Kurzweil has done some work in this area with his FatKat system (www.fatkat.com), does anybody know how well this turned out?
I'm working on more stuff than I can shake a stick at - mostly automated trading interfaces, but not to the extent you are. Can you quantify tape reading? It seems more like an art than a quantifiable science. It's like "ok, you do this, then this, and if this happens, do this, except when this happens" and "well how about if this happens"... "well, that depends...". That's one of those things where the programming is easy, but getting the rules is the tough part.
a system based on tape reading is too dependant on the specialist system. NYSE open book can change things, eventually i think everything will move to electronic.
so you run the risk of doing all the work for total waste. better stick with the pure electronic markets like the emini's, NAZ aint bad though.
so you run the risk of doing all the work for total waste. better stick with the pure electronic markets like the emini's, NAZ aint bad though.