Raise your hand if you’ve ever wanted to customize Mr. Robot’s Best in Bags gear suggestion, even just a little bit?

/raises hand
Mr. Robot understands that us humans can be particular about certain things, or prefer something because we like it better (a concept he is still trying to learn). He also knows that the way we play our characters, or the situation we find ourselves in, could mean an alternate gearing strategy ends up more helpful than the original one he found. You get it – you’ve been there.
The latest feature lets you customize the mathematically best solution with things that you might like better. Combining math and feelings isn’t easy, but admittedly, it’s quite useful! Mr. Robot starts with your Best in Bags solution and then offers you alternatives, quantified so you can see the trade-off.
This lets us think of gear not in the terms of one ultimate set, but instead as a cloud of top-performing sets. The idea is to give you the information you need to confidently choose your gear setup.
How To Use It
After running Best in Bags to get your optimized gear, you will now see a third tab after the results tab.

Click on that to see a list of alternate setups that you could use. Each alternate favors a particular Azerite trait, stat, or both.
Click on one, and Re-Optimize will light up. Press that to see your new result with those preferences in place. If you have a preference chosen, it will show above your gear table.
When you get a bunch of new gear and do Best in Bags again, your preference will be remembered. Mr. Robot will continue to use your preference unless you tell him otherwise or the solution is no longer good enough (your score is too low compared to the original Best in Bags solution).
To clear your preference, click the delete icon next to your preference (as seen in above screenshot). Or go back to the Customize tab, clear your preference, and pick a new one if you’d like.
What does tier 1, 2 and 3 mean?
Each alternate setup is assigned a tier.
Tier 1: Nearly equal in value to the best possible set of gear in your bags. You probably won’t be able to notice the difference in-game, so confidently use whichever tier 1 setup you like the best.
Tier 2: You might be able to see/feel a little loss in-game, but it won’t be enough to seriously impact your performance. You could still rank high and raid at the mythic level without compromising your team in any way. Other hard-to-measure factors could still make one of these setups the best for your personal situation.
Tier 3: You will definitely feel a tad weaker with these setups in most cases. You would probably only want to pick one of these if you either really like playing it, or there is something very specific about your team or a boss fight that makes it work better for you.
Why do we use tiers instead of showing you a list with specific scores like… 0.6% worse than the best setup, 1.1% worse than the best setup, etc.? To prevent undo bias against a particular setup. For example, say that one alternate setup is 0.1% worse than the best in bags solution, and another is 0.15% worse. If we put those in order and show those numbers, people are going to tend to pick the highest one. That’s just human nature. But those are so close to each other… nobody can definitively say which would perform better in-game. So we band them together in tiers to eliminate that bias and encourage people to try different setups. The idea is to help you feel confident picking anything within a tier because they are all so close to each other.
Azerite and gem/enchant thresholds
When you have chosen an alternate solution that prefers a specific azerite trait and/or stat, we disable the Best in Bags azerite threshold and gem/enchant threshold options. We do this because gems and enchants are one of the easiest ways to get more of a particular stat, and preferring a particular stat is implicitly saying “get me more of this stat regardless of score.” Same idea with azerite traits: by setting a preference you are saying “put this azerite trait on as many items as you can, regardless of score.”
We’ll keep an eye on this if it ends up not working out that well — it would be pretty complicated to come up with some kind of “threshold” that works in conjunction with these new preferences at the same time, but not impossible.
How It Works
Here are some details about how we choose the alternatives to show you, and how it picks the specific gear that it does for each alternative.
How do we pick setups?
Say that you have 1 billion possible combinations of gear in your bags (pretty typical for someone who raids and does Mythic+ a decent amount). You might assume that we suggest alternate setups like this:
- Score all 1 billion combinations of gear
- Put them in order from best to worst
- Show you the top 30 or so setups
We do not do it like this.
Why not? Because the top 30 setups will usually be very small variations of each other. For example, the #2 setup would probably be the #1 setup with your wrist slot swapped out for an item that is the same item level with the same stats but in slightly different amounts. That’s not very interesting, and furthermore, those two setups would be so close in value that calling one of them #1 and the other #2 is kind of silly. You could use either one and never tell the difference in-game, both from a performance perspective (meters) and a gameplay perspective (how it feels to you).
Our goal is to show you about 30 significantly different ways to gear — setups that favor different azerite powers to the extent that your gear allows, heavily favor particular stats to the extent that your gear allows, or a combination of the two. This gives you way more interesting choices than 30 slight variations of one set of gear.
How do we pick the gear in each setup?
To understand this, first we need to understand how Best in Bags picks the main setup. The approach is simple in concept (but quite difficult to execute):
- Generate a ton of simulation data ahead of time (we sim everything for you, so you don’t have to spend time doing it yourself)
- Do a bunch of fancy statistics to turn the simulation data into a thing that can quickly and accurately score any combination of talents, stats, azerite, trinkets, etc.
- Run an optimization algorithm on your gear to find the one highest-scoring setup from the billions or trillions available to you
So now we want to find alternatives to the highest-scoring setup. Each of the alternatives represents a “preference” in the optimizer for a specific azerite trait, a specific stat, or both. This gets tricky because the optimizer can no longer rely on “highest score” alone — we have introduced some fairly complex constraints.
Without getting bogged down in the computer programmer details of it: we find the highest-scoring set of gear that favors the preferred azerite trait and/or stat without being dumb. By “being dumb” I mean that we won’t put on an azerite item that is 60 item levels lower just to pick up an azerite trait. We use some heuristics to keep it reasonable. Same idea with stats — if your only ring with your favored stat is significantly lower item level than your best one, we won’t use it.
There is no single “right” answer for these rules of thumb — you are making a trade off between score and how much of the trait/stat you want to get. “Score” is a pretty objective measure. “How much I like Critical Strike” though… can’t really be measured objectively. That said, we think that we’re using a pretty reasonable threshold. If you see any cases where you feel it should be going more or less aggressively for your preferences, please tell us!
There Can Be More Than One
Besides being super convenient, a major goal with this customization feature is to dispel the myth that there is only one way to gear in World of Warcraft. This is a topic that warrants a longer discussion, but here’s the short version: for most specs, you could heavily favor completely different stats, or use completely different azerite powers, and still perform at the highest level — both on paper and in the real game. This has been proven both in-game (via combat logs, etc.) and out of the game (via simulators, calculations, etc.).
It is challenging to present multiple choices to people in a clear way, and in a way that quantifies just how much damage/healing/toughness you might be trading off to get your preferences. But it is a challenge that we think is important to tackle. The game is just more fun if you have more choices, and can confidently make those choices.