I’m using AskMrRobots best in Bags already for a few months to simulate my resto shaman and until 7.3 it worked fine. I always simulate Gearing Strategy for Raid Healing Krossos with Default Rotation and save the machine learning version. Hope this is still correct?
Before 7.3 when I chose best in bags, the result was never seriously lower than the current gear. My typical worst-cast deviation for simulations is something like 30k on 800k HPS, which is below 4%. Considering this margin, the best in bags solution was always worth considering.
However, since 7.3 this changed. I ran multiple simulations for gearing strategy since the last patch hit. Last try was yesterday. Since 7.3 it might happen that I drop from 850 HPS down to 750 HPS when simulating the suggested outfit. None of the suggested outfits was ever better than my current gear, when simulated. Yesterday was the first simulation after the netherlight crucible was available and I even made sure that I selected the first tier on all simulations. However, nothing really changed, still the same mess.
The problem seems to be with the computation of the gearing strategy. When making single runs I still see the impact even of small changes like changing the enchants from hast to crit. It seems all reasonable.
However, the gear that is suggested to me even looks weird on the first look. I have full ToS heroic set bonus and it wants to replace it with 2 items of the previous tier. This means replacing +50% healing rain with +15% on healing wave. Even at the same item level I would expect +50% healing rain to be more effective as a group healer. Not to mention exchanging 915 ilvl with 890.
Am I doing something wrong? Even if I do something wrong with the gearing strategy, I would expect that I make the same mistake for single runs.