If all of those softwares are capable of finding the analytical solution to your problem, simply pick the one in which the answer is in the most convinient form for you.
Mathematica has a huge capabilities to express things in variuos different forms.
Those are symbolic softwares (not numeric) so you cannot simply say that the answer provided by one of them is better/more accurate...
As Luc-Andre Gregoire noted Matlab Sym is a subset of Maple. The Matlab Sym interface between Matlab and Maple is somewhat awkward so if you are doing purely symbolic computations you probably will be better off using Maple or Mathematica. In some instances, where you mix symbolic and numerical computations it may be convenient to use Matlab Sym. For example, the voltage stability toolbox described in the following paper is built using Matlab Sym:
S. Ayasun, C. O. Nwankpa, and H. G. Kwatny, “Voltage Stability Toolbox (VST) for Power System Education and Research,” IEEE Transactions on Education, Vol. 49, November, 2006, pp 432-442.
Using Maple for symbolic computation is good, but it uses lot of memory for evaluation. In addition for robust symbolic computations Mathematica is good and is efficient in using memory.
clearly if you are interested in symbolic computing, Maple and Mathematica offer a much broader set of tools and a better interface than Matlab Sym. Of those two, the 'best' is very subjective. From a purely technical point of view they both offer excellent tools and they both continue to push the boundaries of symbolic computing. However, they are quite different in usage style so you should probably test them and see what you are comfortable with. I personally use Mathematica because I like the free more intuitive (to me) usage. But others might like more structure that is found in Maple.
Maple, mathematica and Matlab sym all are powerful simulator, but for analytical solution in my opinion Mathematica out performed our rest in many cases specially in ODE solver