I use python a lot and R when I need a library. revolution analytics version of R is faster but not free. As far as matlab, I am all ears in wondering what's so great about it; aside from a new gui and lapack it's the same as 20+ years ago
1. In Matlab, things that are supposed to work actually work. Especially as versions are upgraded. See my earlier post as to why R blows goat dick when it comes to this aspect alone. 2. Support. You get what you pay for. When there is an issue, it tends to be resolved quickly. Money talks. 3. Oh gee, let me see, Excellent debugger, decent documentation, JIT acceleration (Matlab is quite fast these days and appears to be getting faster with each new version), Code generation, Parallel Processing Toolbox (multi-core, multi CPU, and GPU), Distributed Computing Server (DCS) ties in nicely with the PPT, etc, etc.
I do not doubt that you have had problems with R but I have upgraded through the last 5 releases with essentially no problems. (I never had many problems before that either). I am using Windows. Every time, I do a clean install and re-download and install all additional packages that I am using from scratch (about 15 in regular use). This takes only a few minutes. I did try other approaches in the past, such as copying libraries from the old installation, and these are problem-prone. I have quite a bit of code and have had no real problems keeping it running.......
Revolution wll do all of the above for a price that's a fraction of a full-blown Matlab+packages and you get institutional-grade support. R-studio will do most of it and I love the R-studio server. I am not even talking about a vast variety of markup/doc tools, graphing libraries and commercial-grade financial projects that are out there for free. I have switched from Matlab to S+ many years ago when I moved over from fixed income to equities. My reason for switching was the fact that Matlabs derivative libraries were not up to snuff, statistics package was incomplete and the user community was shrinking. No regrets so far. If I am going to switch again, it's going to be to some new language like Julia that supposedly is a prefect math+rapdev language.
We still run some R code that has as yet not been ported to Matlab. It will when we get the chance. It's just that for us we made the change to Matlab and "haven't looked back". As a result, our research process is more efficient and the whole setup allows us to concentrate on the important things, like extracting money from the markets. If in the future we determine some other tool/dev-research environment to be more suitable, for us, we will effect the change post haste. In the end, it's whatever rocks your boat.
Out of curiosity, what sort of stuff do you guys do? Vol PMs/funds that I worked-at/dealt-with use Excel, R, Stata and Matlab in the order of preference. FI guys seem to like Matlab. I don't even think the boat matters, it's whatever works for the team or whatever the bank/fund uses. Long time ago I was friends with a professional bicyclist. He once said "An amateur usually has a preference, he likes Shimano or Campy. A pro does not have a choice, he rides whatever they give him."
Mid to high frequency trading (30 seconds to a few days but rarely more than a day). Mainly equities and futures. The directive to use Matlab for research comes from the head honcho and is not negotiable. That aside, I think it is an excellent choice for what we do. This is one of the most productive research groups I have worked in, and the research toolset plays a big role.
So you too "ride what the team gave you" I do agree that Matlab makes most sense given what you guys do.
Yes, we are pretty lucky to be allowed to use Python for some of our work, as the firm's founders have a thing for not liking anything open source.