Posted 09 August 2014 - 09:43 AM
Qualifying statement:
Ok, I am not in competitive play and I don't plan to be, I am in this to have fun and just play with a team who just wants to have fun. I am the most average player you will find as I drop with both high and low ELO players. Most of you never see me and never notice. I don't need to be noticed, I like it that way.
Topic statement:
I can understand why they cannot say who they talked to and why those people cannot announce who they are. I think these changes were reviewed and probably reviewed with several people. My feeling is that there was more to what they discussed than what you saw. PGI has shown precedent with how they treat NERFs and other mechanics modifications. They always start with a set point say of 100. They see how this balances, knowing that 100 is way too high a setting. After some data gathering they will go ahead and change the setting to say 25, knowing full well that this is way too low a setting (this is by design which I will get back to later). They collect more data at this point. They will them adjust the setting to 75, a large swing back higher, but not fully back to where it was. Now they have once again a data set to compare back to the previous settings. This time they can actually get better idea of how low to put the setting back to such that they are not over compensating. Say 50 this time (remember this is just hypothetical with defined values I made up to illustrate a point). Now 50 gives you good, but not great balance, but at this point good is enough to keep up long term play and data collection for later finer adjustments.
Now, why this method? Well simply put, its a basic Design of Experiment methodology, specifically a Surface Response. Your looking for broad trends and directions to go. A good Surface Response design must include data from both ends of the extremes and a theoretical mid-point that you are trying estimate. This is very common in Six Sigma strategies and is used in many industries, including game design. Further refinement comes in a factorial design later, I am not going into this as I doubt they even know the subtly and nuances of using this type of data analysis tool in a virtual environment.
Does this seem a bit heavy handed? Sure it does.
Does it seem far fetched? No, its a very common statistical data analysis tool.
Is this something that should have been done in Beta or a PTS? Undoubtedly, full open servers are not the place to do this kind of testing, unfortunately though you need a very large data sampling to make it work right.
PGI needs to be more open and honest in their dealings with the players. Define what your doing, tell us your intentions. We know for a fact your LRM modifications in the past followed this method, as well as your projectile speed and AC fire range.
Its very very critical for you to tell us whats going on. You could have saved a ton of complaints and issues from the general population.