>>13279782TLDR; I think it does worth it.
It is definitely fun to play with. Some stuff is much easier to be done there than in other languages, e.g., playing around with symbolic equations, plotting stuff, etc., specially exploratory work.
I work with mathematical modelling and use it sometimes when I am trying to understand a problem: integrating and deriving equations, plotting feasible solution space, etc.
Cons: anything that is more "algorithmical" rather than available function is annoying as fuck and slow. Building a for loop with a bunch of conditions is disgusting. This can be solved by integrating mathematica with python. I heard they have a lib/bridge that allows you to use Mat functions in Py.
I am an engineer, so I dislike a bit the "lack" of programming and excess of fluency you need to get with the language to get things done. But that's their selling point, "you don't have to be a programmer". That said, the language structure is very unorthodox and you are gonna get confused and frustrated at first. On the other hand, it contains the most beautiful and complete documentation I've ever seen. If you hit F1 for help, you will see extensive examples of how to use certain functions and tweak the parameters to create whatever variation you want.
I bought it and, to me, even if I use only once every 3 months, it shows itself worth it every single time.
> Pic related. A countour plot of a problem's solution space and a parametric subproblem (animated). On the right, the help window with an example of usage of the function. 
