Verily, I do nothing of the sort! What I do is use estimations of DPS to determine what the changes in DPS ought to be per stat point (relative to that gearset), then use that to maximize weights into a new gearset, and repeat over and over until I'm not getting new results. Then I settle on the best one I found thus far. So for true BiS, that especially means redoing melds to give all candidates for "best stat" a fair shot, whereas for the tome-only set, you're stuck with whatever ballpark it's in.
Weights are always relative to a gearset, so basically I don't look at thresholds and I don't come up with half-baked rules of thumb as to when one is better or whatever. Ever since the page system got added I have simply assumed you know what gear you will get in the end and should meld towards that.
Alas, it gets worse than that. Because the reward from stats has rounding stuff going on, you do not get rewarded for every single point of a stat you put in. That means maximizing weights does not guarantee you find the best result, although you can expect to get close; e.g. even when CH>DH by weight, it is often the case that +40 DH > +40 CH, and very very often +12 DH > +12 CH. There is no good way to tell that just by looking.
This is the problem I'm looking to solve. I think I've devised a computationally sane way around it; only building it and testing the program can prove it, though.


Reply With Quote


