2.1.2 Varying confidence thresholds wrote:In the original confidence design, rewards, half life time and stage thresholds were all fixed and thus independent of time. As a result it was possible to stage down in late game which resulted in weak guns being used against a lot of (higher stage) defense buildables. The match would end in a tie eventually. This was fixed by giving stage thresholds a half life time, too. This works but having two independent nonlinear functions seems like unncessary complexity to me.
Ishq wrote:I think the main reason complex feels to opaque and complex is because we are too focused on providing exact numbers. For instance, in reality, the amount of confidence you have and your confidence threshold are both meaningless individually. They only mean something when put together, so why even display them separately? Display confidence as a percentage to the next stage or a progressbar, thus hiding the fact that confidence is actually dictated by two varying nonlinear functions. It doesn't matter how hard or complex our underlying systems are as long as we can wrap a pretty interface over it. The player doesn't need to know that you will get exactly 25.3455 confidence for building a buildable in a certain location, he really only needs to know he will get more building there as opposed to building somewhere else. The simple fact that confidence awards are directly proportional to the risk is intuitive enough to understand. I am for adding more complexity and more special cases to the confidence system to make a finely tuned beast, but hide that behind generalities. For instance, beyond perhaps dretch, granger, and the naked human, no one has reward values for each class memorized. They don't need to know either, they just need to know that they will get more credits for killing a tyrant than a dragoon, and more for a dragoon than a marauder.