I think we all know EQDpm is not a tool for parsing, so it isn't measuring anything in realtime. But you act like just because it isn't, it's outputs aren't based on real, quantified data from the game, e.g. haste, weapon damage and delay. You can map eqdpm's numbers to numbers from the game and then talk about its predictive precision and accuracy. Yeah, it isn't the dictionary definition of accuracy and precision, but it is a pretty minor cognitive leap I'm asking you to take here. The terms accuracy and precision can describe a simulation OR measurements taken by instrument. My only error here is that I don't actually have real numbers to justify what I'm saying, but that was admitted from the beginning.
There is no replacement for laboratory (in game) measurements, but their downside is that any mistakes you make are less likely to show up as systematic bias. In other words, it might be even harder to draw conclusions from in game data because it will be wrong in unreproducible ways. At least with a simulation, most confounding variables will cause skew of equal magnitude for high and low values.
We are dealing with a messy system here and I didn't want to make it seem like I know the answer. But barring the possibility of p99 being wildly wrong with its combat mechanics, we should expect similar performance from epic fist/whatever and epic hasted tstaff based on our limited, flawed data.
Also, I don't need any additional support for my idea of an asymptotic approach to the delay floor, other than that it would be totally unreasonable from a game design perspective to allow a 9/20 weapon to perform like a 9/16 weapon. It destroys the risk/reward structure the game is built around. Again, it is possible the code is actually hugely unfair about this, but it is not a hallmark of a successful game to stop rewarding effort by removing proper scaling as limits are approached. Your 9/20 should go to 9/13 and a 9/16 should go to 9/11 with 81% haste.
Not sure why it is my job to provide data while you can just bring anecdotes. EQDpm is still useful for ranking items--it has a well-tested internal precision and uses real item data as input values--though it fails to accurately model P1999 dps.
I love your idea of doing some real parsing. I hope you can find a way to control the environment enough to get meaningful numbers. I don't wanna fight here, because you are mostly right. A parse is pure data, all I have here is a simulacrum. But it isn't pure garbage.
|