Testing and Trading Strategies - which Technologie?

W

Wippi1817

Hi,
I would like to write an Algo Trading System which includes the following "requirements":

--> should 100% trade automatically
--> should handle Forex and Crypto
--> will include necessary tests of strategy incl. Walk Forward Testing

Which Programming language would you suggest? I guess Python or?
Which libs do you recommend the math and strategy calculations?

Thx
 
IMO QuantConnect is a decent introductory platform (Python or C#). Once you used that a while you may figure out things that you need and it cannot offer. This platform has no optimization though (which means WFO isn't necessary) since it's built for hypothesis driven strategy development, similar to Quantopian. I would advise against immediately building your own infrastructure without experience simply because it's such a major time hole and the prerequisite (and harder part) is finding a strategy with a real edge. QC allows you to immediately start looking for that edge rather than spend time on auxiliary stuff that might never pay off.
 
Wippi1817 said:
But why should it work without optimization? And Walkforward and strategy testing? What i have seen thats only backtesting or?
More...

Yeah, QC is only back testing (and live trading or selling signals/alpha). Depends on what you're doing. Optimization is slightly 'magic' anyway and is a large source of curve fitting. On the other hand, a lot of traders do use it in practice. Anyway, if WFO is absolutely critical to what you're doing then QC is not the right platform, though you could still implement than in their Lean engine probably if you're running it on your own machines (and it wouldn't surprise me if someone already has done so).
 
What type of algo do you plan?

Is it do once and end type of project or something you plan to work on for longer?
 
Snuskpelle said:
Yeah, QC is only back testing (and live trading or selling signals/alpha). Depends on what you're doing. Optimization is slightly 'magic' anyway and is a large source of curve fitting. On the other hand, a lot of traders do use it in practice. Anyway, if WFO is absolutely critical to what you're doing then QC is not the right platform, though you could still implement than in their Lean engine probably if you're running it on your own machines (and it wouldn't surprise me if someone already has done so).
More...

I would like to test strategies I think that is the most important thing. I would like to do entry exit tests, monkey etc. in order to see my if there is an edge or not. In addition the Walk Forward testing would be a major point. All these elements I could not see in QC - it is just another plattform where you can try out strategies or and live trade them.
 
931 said:
What type of algo do you plan?

Is it do once and end type of project or something you plan to work on for longer?
More...
it should be a longer project and not something for 2 months - my alogs would be not very complex but should contain a lot of tests.
 
Wippi1817 said:
I would like to test strategies I think that is the most important thing. I would like to do entry exit tests, monkey etc. in order to see my if there is an edge or not. In addition the Walk Forward testing would be a major point. All these elements I could not see in QC - it is just another plattform where you can try out strategies or and live trade them.
More...

I mean, you can code up an entry, then backtest it with random exits. You can also code a monkey test/random entries and exits. In fact, I wrote an algo that internally compared some fixed strategies against multiple random strategies and only traded them while they did significantly better, though that requires internally running your own backtest logic... The problem however is you can't programatically start backtests on their platform, which limits the the number of use cases quite severely as you note.
 
You must log in or register to reply here.
Back
Top