It has been a while since last I posted :D
I have done several massive crowd simulation tests with Thinking Particles these days.
You can watch my R&D video here.(About over 200 Agents in this Video)
TP has a potential to simulate and control over on massive crowd agents , in most of case it works very well and easy to control but if the situation becomes complex , for instance if you want a group of agents(here let me call they Blues ) to react with behaviors of other group agents(Reds) , that might be your nightmare , you should build all your own behavior rules into TP that will become very very pain in the ass and consume you a lot of times.
Due to I haven't a experience of using Massive which software was designed to force on simulate huge amount of crowd with distinct complex behavior on echo agent , I can't compare between Massive and TP . Does Massive can make things more easily on this case?
Here is a another video.
Try to figure out the difference between different SystemScale Value , and to see how it affect on result of simulation and rendering.