Easing into GloNet

simulation

#1

Rather than waiting for all the hot fixes to finish, I suggest letting the GloNet go wild with a reduced simulation to begin trialling the concept. Reductions could include:

Error Targets
The error target could be reduced to 1% (from 0.1%).

Talents
Rather than considering all 2187 combinations, look at just the two or three rows which tend to have the biggest impact on strategies. This will result in more accurate results than assuming a single set of talent choices for all players, but less accurate than considering all possible combinations.

Legendaries
Rather than considering every possible combination, consider only a handful of legendaries that are commonly chosen by high-end mythic raiders (based on a review of combat logs). This will give some people more accurate results than assuming a single set of legendaries; but others will have to wait if they are using legendaries that were ignored.

Set Bonuses
Consider only the previous raid tier, with the possible combinations being none, 2p or 4p set bonus. This will give accurate results for the majority of players (at least until the new raid starts), but will be inaccurate for players with less common sets like the class order hall or previous raid tiers.


Once the hot fixes become less frequent, you could then adopt a middle-ground approach by slowly adding in more talent rows, legendaries and set bonuses to the equation (while incrementally improving the error target).


#2

We already plan to prune down to less combinations – even with the global network, it would take forever to do every combination. Also, many of those combinations, with a little bit of smarts, can be shown to not warrant their own gearing strategies because they are “significantly similar” to other combinations.

“Significantly similar” we define by: they tend to have very similar characteristics insofar as what stats/other gear they prefer.