It might be user error, but I’m having trouble figuring out what I’m doing wrong. The sim isn’t suggesting epic gems for any of my Holy Priest builds, also the neck seems like a low recommendation. I noticed that on my last saved build, I didn’t have an epic gem either (for Holy Priest). I can’t imagine the secondary stat gems would be better than a primary stat gem.
e4cdb735a59c4e639e2ba42454bc6e7e
Firstly:
That said, nothing appears to be locked to any item or spec.; I had as thorough a look to see if there were any ‘hidden locks’ within gems, etc, so it may well be what @yellowfive affectionately refers to as ’ an edge case’ or simply working as intended…
Sorry that my limited working knowledge can’t offer much by way of answers…
So in this case, using one of the epic gems would actually score just slightly better. When I say slight though, I meant very slight… like 0.25%.
I honestly do not understand the epic primary gems in this expansion… they are sometimes not worth using at all, and when they are worth using… they are just barely worth using. You could safely ignore them and not notice any real difference in-game.
Given the massive number of gear+enchant+gem combinations, it is not feasible for the optimizer to try every possible combination, so sometimes it will go down a path that finds a result that is a fraction of a percent lower than the theoretical maximum. These differences should be quite small though – small enough that one could not determine which result actually performs better in-game.
Thanks @yellowfive, that’s good to know that the sim might not check every combo, and I might have to manually pick something that is slightly better. Is there a way to queue for “full” sim, might even be a premium feature 
I understand the differences are small but I’m a nerd.
@eighjan Thanks for the reply, I did play with having Holy first, and there was no difference. I was Holy all last season, and it didn’t recommend an Epic gem. So I thought that I would throw out the question. I also got some weird sims where it wouldn’t recommend a gem at all, just leave it empty, which was odd.
Here is a sim with holy first: 7785c70fe3a34affa3f1e7d98f314423
The optimizer should almost never leave a socket empty – there are some combinations of settings that could do that though, e.g. setting gems to None or in rare cases with the gem+enchant threshold set to a higher value.
Once we finish polishing up 11.2, we are looking at ideas for some new features. A version of the optimizer that inspects many more combinations but runs slower could be an option. The current, faster optimizer already gets the best solution in a lot of cases, so we would have to run some tests to see how often a slower optimization actually makes an improvement, and see how much it would cost to run.
For example, if we find that it only gets a 0.25% better solution 25% of the time, but costs a lot in extra CPU time… I’m sure somebody would pay for that, but I wouldn’t feel great about charging for it. To me, that is not a good value proposition.
I appreciate you looking out for the end user. I’m sure you also need to weigh it against the cost of taxing your backend systems.
If the user were to turn off cost protections for gems and enchants, they would probably be looking to run the prolonged sim. You might be able to reserve it for that use case.
