Warning, this will be severely off topic.
Hi Ishkur. Well. First of all I want you to know that I don’t think I’m really qualified to answer that question with any level of confidence. I don’t KNOW what AMR does. I only assume. If you want a good answer, ask someone who actually knows the inner workings of AMR. So -disclaimer-. There.
I assume AMR does what most simulators are forced to do: It runs a bot that does a preset number of things on a preset number of conditions. Maybe they actually simulate a bossfight, with all the mechanics and player reactions involved, with rDD and mDD bots doing their thing, maybe getting hit by stuff, standing and running, and the healBOT script healing them. I don’t know this.
What I know is, in order for any kind of measurement to take place, you have to run a simulation.
Anyone coming up with that simulation is going to have to face a ton of choices looking at all the variables that go into a Raidencounter, and the options are always twofold:
You look at things like: (Raidsize, Raid rDPS, Raid mDPS, mechanics that force movement, how long the fight takes, prepots, margin of error for every simulated player involved, how often to people have to cancel their casts, how often does melee have to switch targets, how much stuff is there that is randomly distributed damage, anything else that might play into it)
Looking at all those possibilities the choice is: Safe resources and make the variable a fixed value, maybe some kind of average pulled from real life data. Or simulate it with something they feel is realistic, maybe also based on averages pulled from real life logfiles, running it few hundret times, and average the outcome of those simulations.
To understand where this started you have to realize that most simulators started with simple DPS rotations, you could maybe even set up a margin of error for human delay, and maybe the simulator even had a knob to tell it special stuff, like “hey, you know I have AOE situations in this fight”, stuff like that. People knew they would never actually reach the numbers that the simulator told then were possible, but it was still worthwhile, because all they wanted to know was how to MAX the amount of dps during the time that they would be able to dps like a bot. I know I have to do other stuff aswell, but if I find out what numbers give me the most flat out stupid DPS, I still get to know what values are best for me, as long as I’m exactly simulating the rotation that I am actually using “most of the time” at least.
But healing is different. You can’t just min/max healing per second over a 10 minute fight and be done with it. You may have the perfect rotation for that, and last exactly 10 minutes, but as soon as you get a 5 minute fight and there is someone who knows how to get more HPS out of his almost double manapool (5 to 10 minutes -5 minutes of mana regen) that person is going to outperform you by a lot. Sure, it won’t be as efficient, but after 5 minutes with your 10 minute rotation you will sit there with 45% mana left over, and he or she will have dumped all that mana into additional healing that would not have lasted 10 seconds beyond that. And this is just one of the more blatant ways to demonstrate a difference. How big is the raid? The rogue simulating his DPS rotation doesn’t care. But you care. A lot! How much unavoidable damage is there per second? Is your 10 minute average HPS even enough to heal that, or do you HAVE to have burst healing as well in order to even survive the encounter? The simulator would get super complex, super fast. And simulating a different bossfight would probably mean at least rewriting and retuning the script for the whole thing, every single time…
So…. what DOES AMR actually simulate? I have no idea. You can set up Strategies and run simulations on your own system to try things out, but as far as I know there is actually no way to view or edit the “boss fights” that are being simulated… I mean those here:
Long story short: I think AMR is GREAT. I love it. It is a tool for measurement where no tool for measurement exists, and as such, even an inaccurate measurement is better then no measurement at all. You just have to be careful that you don’t religiously believe in everything it tells you, especially if you don’t know the basis of how it came to believe that.
If it gave me something like: “Here, those are THE best stats to heal through a 10 minute enounter with no movement, using all your mana and achieving maximum HPS on average throughout the fight” I’m like “great, so know I just need to adjust in my head for 3 phases of burst damage and about 1 minute total of 3-5 seconds movement, and the fact that the fight actually only lasts for 6:30 with the lineup we run”
3 Phases of burst would then mean “3 events where Mastery is valuable”. Since 3-5 seconds of movement means that even moderate incoming damage will stack up on everyone, assuming you are not all playing resto druid, Mastery gets additional value, even if it’s just single digit percentile. Now the fight is 6:30 you burnt, let’s say, 15% extra mana healing bursty stuff in a way that was mana inefficient, that leaves you with 3:30 minutes minus 15% Mana left over. As you can see, you are totally going to do feelycrafting at this point, and the only thing usefull to do is to trust your gut and check out if your assumptions are actually correct.
You may end up wanting less crit, or if you are unable to adjust that, you may want to heal more mana inefficient at least, to actually make use of the crit instead of having it translate into left over mana at the end of the fight. In both cases though, mastery might become more valuable.
The thing that comes up a lot is that: “Mastery (in our case) is for progression” what that means is: Mastery is for healing people who take damage, who are on the edge of not being able to complete the encounter. Lots of low HP = great for mastery. Fights last longer, so you can’t just burst heal 3-4 events and be done with it, you actually start thinking about efficiency a lot more, healers are going oom trying to deal with mechanics and maybe even compensate for mistakes.
The way you play determines what stats are worth to you. And the way you play will very likely change over time. That is why I think AMR statweights are bs. It is not cookiecutter one size fits all, those times are long gone, not even DPS classes have that anymore. They too have to reevaluate their rotations and statweights with every change, be it in gear or be it in playstyle+encounter length (due to increased dps). I think AMR should do a better job of informing people about what they actually simulate, and what that means. But I still love the tool, I am actually subscribed to it even though I barely have money for that kind of stuff.
EDIT: One more thing I want to say. If you really want a good answer, you will have to talk to the people that actually know how AMR simulates things, or how Simcraft, or any other tool, simulates things. Everyone else, like me, is just speculating. I have zero inside knowledge about that, for all I know they could be rolling dice, picking a random warcraft log and averaging those numbers just as well as running thousands of simulations and statistical calculations every minute with a tool that was originally developed for cancer research and uses deep learning techniques, evolving algorythms, AI and what not because the programmer is a bored super genious intern at google who graduated in computer science at the age of 12. I really have no idea.