Hello there, I recently developed an application using the IG rest API webservices (of IG Markets) which makes it possible to automate trading strategies. In my application it is also possible to save all streams that are received, so all rate-changes (ask and bid) are saved. This way it is possible to test out a strategy and see with real data if it works. Is someone here willing to share a winning trading strategy? I can test it for you then with real data. Regards, Maurijn van der Does
If curve fitting / parameter optimization is the last thing you do to your trading engine / plan, I believe it can be OK (but need to see results/confirmation first). Especially when aware of the problem, using it to learn more and it doesn't make too wild pnl gyrations. However, if someone only do this in order to search for a raw edge, without any ideas to back it up, it's probably going to take a long time to find anything worthwhile, or you might miss the real edge = your understanding of some dynamics in the markets that can be exploited. I could never trade someone else's plan. Hopeless proposition personally. Need to "grok" everything from bottom up myself, and then it's easier to further develop understanding. For me, this takes years as don't know anyone else with this particular interest, but not really able to listen to canned solutions either, as they never work universally anyways. Also, where is the bottom anyways? Same with systems I haven't touched for a long time. Can't further develop them or change them, as I wouldn't know the impact on established positions and future performance. Good contributions here.
Why not build a modular system? You start with 1 module. When you observe that in certain situations you should trade another way, you split the original module in two module. Each module will trade it own specific patterns. The advantage is that you can optimize your system by building a number of modules that perform well in its own specific situations. A second advantage is that if you want to make changes, only the module you change will have to be checked as the other modules are not touched and will perform as before.
Probably a good idea. I'm currently at the stage of crawling before I can stand up, or maybe standing up but can't walk properly yet. So for now focusing on one whole system at the time, and may run older systems just for kicks, or turn them off since they don't perform too well. But time will tell as they are long-term systems, so can't be impatient with them. I do like the idea of having one system and integrating different perspectives into it, but if the rules are too contradictory, I can see that may not be an option and modularizing also has its benefits.
Some reversal strategies work well on sideways but make chaos on trending times and trend followers may work good on trending and bad on sideways. If you have found some indication in past data for potential upcoming condition it is nice idea to switch multiple algos. Algos could be balanced for taking similar risks also.
I didn't add a trend/consolidation filter yet, didn't expect the results to be as good without filter. Uses 16 tick TP, 16 tick SL. If you look at the data, the counter signal cuts the loss quicker than 16 tick SL trigger. Even though 46% profitable, the ticks accrued on profits is greater.