| 12:25:05 AM - Mon, Jan 31st 2011 |
|
Hey, everyone.
Something that I've noted about theorycrafting in other games (such as WoW) is that almost all of the simulation is Monte Carlo simulation. That is, it is a 'spreadsheet' simulation designed to find top DPS.
However, one of the major assumptions/limitations of Monte Carlo simulation is that time is not that important. Particularly, it is difficult to see system dynamics evolve with a Monte Carlo simulation. For this type of simulation, discrete-event simulation is much better.
Basically, Monte Carlo simulation can give a theoretical maximum DPS for a given character, but the assumptions of the model can't be violated. One of the major assumptions (at least in WoW) is that the character cannot move. After a maximum is achieved, theorycrafters generally do 'postprocessing' on the information to abstract out the effect of movement. However, even with this abstraction, it assumes movement is movement. That is, the movement isn't purposeful.
What I would like to propose is to use discrete-event simulation (specifically, agent-based discrete-event simulation) to create a better model for theorycrafting in Rift. Taking the Monte Carlo simulation 'rotations' as a starting point, it should be possible to create an agent-based simulation that can put an entire party or raid together, and run a specific encounter to see the impact of that encounter (and the composition of the group) on raid DPS.
As a graduate student in Industrial Engineering, I have access to several simulation packages. Arena and Simio are both discrete-event, but Anylogic is also agent-based. I would like to offer my services to the Rift community to create agent-based simulations to determine and test spec-, group-, and raid-based DPS.
|
