Would you say that there are important things that you can do with Amibroker that you can't do with RealTest?
Certainly. The first area would be power and flexibility of chart customization. My charts are flexible enough to be useful for reviewing the trades from a backtest, but that is their main purpose, so they're not as infinitely customizable as AB's.
The second area would be perhaps some kind of esoteric indicator that truly requires complex code to specify (can't practically be specified using a series of array assignments). RT has all the standard indicators but lacks, for example, matrix processing.
That said, there are things RT does that AB can't do or can only do with difficulty. A very simple example is to simply go through all the trades in a back test in sequence and see them (one at a time) on charts, automatically sized to show the entry and exit bars clearly. The multiple strategy support which Val has highlighted is, of course, another.
Could you run a two parameter optimization with RealTest and have the results display in 3D?
I have a powerful optimizer with multiple optimization modes including genetic. I don't currently have 3D graphing of 2-parameter results, though I do have a heatmap graph which actually conveys no less information than the 3D graph (it just looks less fancy). It also has a unique ability with 3+ parameter optimizations to show any 2 parameters as the average of all other results with those parameters.
Some of the optimizations that I run for Amibroker can take several minutes and that's running up to 32 threads on the computer processor. Do you think RealTest would perform faster or slower?
RealTest is, in general, quite a bit faster than AmiBroker. This is especially true when using Norgate data with extra series such as original price or index constituency. The difference is mainly due to RT's data architecture. You define the subset of the database that you actually need, and run an import (extract that subset) once. After that, the extracted data remains in memory for very efficient (direct memory pointer) access.
I would be hesitant to switch from Amibroker but I would certainly be open to using something that is simpler, faster and works better.
I would be hesitant to suggest that anyone who has been through the full AB learning curve and developed their production strategies with it should switch (had this same conversation with Cesar Alvarez). That said, I'd be happy to have you try it and see if there are ways it could be useful to you, in parallel with AB. (One might be to simply have a way to double-check your AB backtest results.) And even if you don't find it useful, that would be useful feedback for me!