I'm new to trading, so please correct me when I'm wrong, but I'm wondering why traders seem to make such big secrets of their strategies? Other programmer's forums are full of source code snippets with which people explain and exchange ideas, but I see almost no code here, except from what I've posted myself. When someone posts "What do you think of this strategy", I would expect a script of that strategy, but instead there's only some performance figures. As far as I understand, the usual fear is that the strategy might expire when published, or even, heaven forbid, that someone else might make some money using your idea. But given the trade volume today, posting strategy ideas is not likely to cause expiration. And the disadvantage of this secrecy seems to be that all work alone and don't exchange ideas, and mostly don't accomplish much. I was told that few traders today are able to get a living from automated trading, although that seems not be too difficult. We could at least in the last months live without problem just from profit generated by strategy tests. Maybe all would benefit from a more open approach - or what do you think?
At least for me, I don't use any standard platforms for back testing. I use R for quick analysis and python for back testing, a custom ATS for simulated and live trading with IB. For data, I have my own database that I built containing some purchased data as well as recent updates from IB. My python back tester is a huge set of classes, and I modify just the core routines which are in heavy object oriented syntax so it wouldn't be very useful to post. I think what's important more than implementation for me is the kernel of truth in the algorithm, the actual underlying physics that makes it happen. If I get that from ET, I will immediately code it into my framework and test.
cause the people who actually do trade and trade profitably -they paid to get there.by trial and error. why would i share something with you? something that i spend on 1-3-5 years of development and tens of thousands of dollars in losses during this process? i might agree that working in team on something from scratch can and should be faster,but from my experience team work on trading didn't work very well. at least for me.
I understand that money plays a role, but programmers - at least in the area that I know, game programming - probably make a lot more money than private day traders, and still work in teams and are not so bound to secrecy. And at least the costly trial and error period that you mention would probably be eliminated if you had learned of working, profitable algorithms and test methods right from the start.
Because an edge discovered would be traded until the curve flattens. Unlike game programming, traders are competing against each other. Teamwork makes sense but you'd need to know you're getting as much back as you're giving, there is no way to stop the other party leaving with your work. Intellectual property rules can't be enforced efficiently in trading, notice how many lawsuits the HF people are in at any given time.
exactly. plus-the liquidity can be a big issue. once again-that was exactly my experience . once you show to other "team workers" that your idea does have some potential-they typically grab and run away. the only way to work in team , in this field is literally sit next to each other,live near by and know and trust to each other. now go head and try to find good trusted folks,who would like to contribute few years of their free time on something that might or might not be successful.
that hugely depends on your strategy or system. most of intraday short term strategies,specially in stocks CANNOT BE REPLICATED by any sim.
Why don't you test a few of those codes posted at other websites and come back with the performance report?
good point! good place to start: http://www.wealth-lab.com/Charting/Strategies/Default.aspx?id=s9qQFHF6huG/ZXcrZN/msg==
I would argue this is where there is a potential edge -- you are saying that certain strategies can be done by a human trader, but cannot be done by code at least using today's knowledge. If that's the case, I think there is room to profit, because every day new algorithms are born that are able to do things previously only possible by humans. On the other hand the reverse is also true -- there are strategies that can only be done by a computer and not by a human. HFT is the obvious example. But another area is risk / money management: If you have a profitable strategy in principle, but execution is difficult to perfect by hand, then it would make sense to code it.