I'd just start looking at examples of something like a simple moving average crossover system in multiple languages.
-Matlab
-Python
- R
- c#
- Easy Language
- AFL ( Amibroker Formula Language)
And compare. Maybe think of ideas you want to backtest/model and see what work is out there.
I myself have been using Amibroker and it meets my needs so far. Benefits of proprietary languages are to me is that I'm using the same software as many other users and it's easier for myself to find examples of a problem I'm trying to solve. The basic backtester is almost dummy proof and is very easy to understand with the documentation. Many great built in functions and simple options make things easier too. It also allows with the custom backtester and the multiple backtest modes for advanced ideas to be done.
Issues that I've had is that it's not easy to learn past very simple ideas.(no language does though). Programming takes a lot of precision I guess you could say.
For instance for myself one of my first ideas was to try to backtest entering a pullback/consolidation in a trend with some sort of momentum in the previous swing. Not an easy concept for myself anyways to program since computers only understand math basically. Theirs just way too many if,and, nots to capture what you expect to and to exclude what you think is obvious.
I myself have grown an interest in perhaps learning some Python as it has a lot of interesting material out there and seems to be gaining a lot of ground in the financial world. One issue I can see from someone that hasn't really dug into it is that it is open source, it has a large amount of libraries and material is really scattered all over the place. The other issue is that say I'm having issues with Amibroker and can't understand what's wrong I can go to the Yahoo group or e-mail support and they know the software inside out and can quickly go, oh this function was used incorrectly, try this instead, where as with Python, it's free and open source and nobody really has to be accountable for any bugs or issues and could take some effort to resolve. But there are huge communities so perhaps it is easy.
Here's a video I watched recently with Ernest Chan with his comparison/opinion on Matlab,R and Python. It is one mans opinion and he does seem very leaned toward Matlab which I wouldn't consider easy from example codes I've seen but I do believe it's popular in the academic/university world so some people already have some knowledge which makes it a time saver.
Oh thought I'd add this think as I fooled around with it one night. Get an idea of Python in a simple way(not related to financial time series analysis). May be others as I did try one with C so perhaps theirs a learn c# or R learn website with the interactive editor built in.
http://www.learnpython.org