I've been practicing manual trading with NT Sim + real data feed. I will need to try live to get a better understanding of it. Thought the psychological aspect is out of the equation when algorithm are involve.
When you mention "type of market" does this mean "bond vs stocks vs futures" ("different group of instruments")?
I remember reading some article where people were saying they were regularly adjusting the algorithm to adapt better to the newest data. It seems there is a fine line between over fitting and forecasting a little bit ahead (I do not really mean forecasting, but I'm not sure of the right word), is it what you experience? I notice for instance ES in 2004 move a bit differently than let say ES in 2013 (different volatility and volume..) and also this is even more different than NQ, or CL. I'm wondering if normalizing can be enough to work with ES from 2003 to 2015 or if there is not way the same code can handle such a long periods?
For your recent algo, can you share as an example what instruments and time frame it can handle when you are back testing?