AMR Under value Convergence of Fates (Fury)?

I have seen similar posts here where the response is “depends on fight length”, but I can’t find anywhere else that devalues Convergence as much. I know your sims are different than simulationcraft, but with their sims some combination of CoF + an on use trinket always sims higher than 2 on use trinkets. Perusing warcraftlogs CoF is the overwhelming default choice, regardless of fight or difficulty and I have to believe some of those people have 2 BiS on use trinkets.

The fight length is the overwhelming factor in the ranking of this trinket.
If you’re comparing the AMR and SimC results make sure the fight lengths are the same.

To me it’s not a trinket I’ll use until I know the kill length the group will hit, then it it falls in the correct duration I’d use it if I was trying to rank highly. (Of course when I try that someone screws up and the fight length changes!)

One of the reasons you’ll see it on so many logs is because people like to blindly follow a list and not think about things. So you’ll end up with everyone using the same trinkets if they get them, that doesn’t mean that other trinkets wouldn’t get them a better result though, this isn’t a binary situation where CoF is a 1 and everything else is a 0.

Also good players will perform well with just about any gear, given it’s spec and ilvl appropriate.

“One of the reasons you’ll see it on so many logs is because people like to blindly follow a list and not think about things.”

I would possibly believe that if there was some real disparity in the logs, that’s just not the case. I can’t find a single person in the top 10 heroic parses not using it, and almost no one on any individual fight top 100 rankings. For the first 3 Mythic fights where at least 100 parses are available there are 2 guys out of all 3 fights not using CoF. I have a hard time believing 99% of the people killing mythic bosses are blindly following lists, those people tend to theory craft more than most.

In my experience not all the top players theory craft, it’s been 2-3 a group who do lots and poke the others enough that they’ll do some research.
Given the path of least resistance is to find a BiS List they’ll use that.

Very rarely did they run sims themselves based on their characters. This was a few years ago as I’m a lot more casual now, however I’d be really surprised if that part has changed. I’ve just joined a new, more serious, guild and one of the officers asked all the raiders to go and look at raidbots. So it doesn’t seem to have changed.

Either way it doesn’t matter, you’re free to use whatever you’d like. The results from AMR don’t force you to use them.
Zoopercat spends a lot of time comparing log results to the sim results. Did you read her blog post about comparing them?
There would have been many, many hours put into that. You could just read the warrior bit but that might leave you with questions which the first part answers.

Have you looked through the rotation and the trinket data to see if there’s a problem?
If there is I’m sure the team here would like to know, they actively look for bugs but there are a lot of things on their plates.

The other day on the Discord server there was a great conversation between one of the mage theory crafters and @Swol. I think it was one of the people who maintain the mage part of SimC. He was doing a comparison of the mage changes between the two sims to see if there were any errors in either.

I commented at the time that it was great to see.
It would be nice if all of the class/spec TC’s had the same attitude but that isn’t how its worked out.

You can always double check our results by running simulations using CoF vs. not. A quick look shows that it is indeed working – I get more Battle Cry casts with it than without, and quick napkin math shows that it is triggering at the correct rate.

For example, in the test I ran, it is getting ~2.5 extra Battle Cry uses in a 5-minute fight. It triggers roughly 5 times per minute, so that would total 125 seconds of CD reduction on a 50-sec cooldown… but we’ll be generous and call it a 45-sec CD on average due to odyn’s champion, so 125/45 = 2.78. Due to how things time and line up and randomness, sometimes you’ll be above or below the “expected” extra uses.

CoF is also one of those items where, if you get lucky, it can be really good. So it is not surprising that the top rankings use it. Just because all the top rankings use an item though… doesn’t mean that “on average” some other item might actually be better. By “on average” I mean that you have a higher probability of doing more DPS on any given fight – that is how we recommend gear. We do not recommend gear to give you a low probability to do unusually high damage. Most people would prefer to do well more often, than to sometimes get lucky and do abnormally high damage.

“Either way it doesn’t matter, you’re free to use whatever you’d like. The results from AMR don’t force you to use them.”

I am aware, I use as much info as I can, but I start to question it when it’s so drastically different from reality. That’s why I posted this comment.

"Have you looked through the rotation and the trinket data to see if there’s a problem?"
I have not, I have looked mainly at logs, and the fact that hundreds are using it and 1 or 2 aren’t is too much for me to believe it’s the blind following the blind. Might be a bit too early to tell but I can’t find anywhere other than AMR that recommends using something over CoF in ToS as a Fury warrior if you have it.

At AMR we take a data-driven approach to theorycraft that is particularly good at finding combinations of gear that are unexpectedly good.

We have special logic in our rotations to get max value from using two on use trinkets, which is missing from simc.

I believe that you could make CoF rank highly on average by adjusting the fight length. When looking at log data, you are seeing results heavily affected by confirmation bias. Top parsers think CoF is best, so they use it. People look at top parsers to pick trinkets, they use it. And so on. This is why looking at only log data can often miss viable, or even optimal, gear combinations.

So, I would say that the fact AMR finds different results than what is popular shouldn’t be alarming. We double check for bugs when this happens, but it isn’t a sign of a problem necessarily.