Generated pawn weights are incompatible with suggested gear in Best in Bag feature

Hi when i use “Best in Bag” feature, and click send to addon I get following string, where it clearly values Haste above all secondary stats (Haste>>Versatility>>Crit>Mastery).
( Pawn: v1: “Fseven - Discipline”: Class=Priest, Spec=Discipline, Avoidance=0.02, CritRating=0.85, HasteRating=1.09, Indestructible=0.01, Intellect=1.21, Leech=0.49, MasteryRating=0.81, MovementSpeed=0.01, Versatility=0.97 )

While the Best In bag feature gives me totally different priorities: Crit=Versatility>Mastery>Haste.

Are the pawn weights correct or the Ask Mr Robot suggestions? If the second, where can I find the correct stat weights for given gear setup?

This is my personal view on this:

Weights aren’t as simple as Pawn is showing it. Your weights are never fixed, it always depends on your current rating for all your stats. This is truly where Mr. Robot shines!

At the current gear/gems/enchants you have available, crit is what gives you the best bang for the buck. If you could add something more onto your exactly current gear, haste would be the best secondary stat! However, not at the cost of crit in your current set of gear.

Pawn can show you a ballpark value at best, Mr. Robot will truly let you know if it’s an upgrade or not by doing Best in Bags.

So the Pawn weights you get is correct, IF you don’t sacrifice any other stat.

1 Like

Yeah, ragowit has the right idea:

There are many, many good stat distributions for your character. They won’t all have more Haste, they won’t all have more Crit. Best in Bags uses a scoring function built from simulations to find the best scoring set of gear possible for you. This function is NOT linear. Stat weights ARE linear (which is why they are inferior to our scoring function).

The way we create the stat weights for use with pawn is that we examine all of the stat combinations at your item level which are better than your current Best in Bag gear. In your case, most of those have more haste than you do - and that is how we calculate the value. Basically… stat weights and pawn give you an extremely rough idea of whether an item might be an upgrade for you. You might get an item with haste on it… and BiB won’t even pick it.

I know it’s a bit confusing, but, that’s why we created the website and our combination optimizer! Especially now with personal loot - ranking an item with an addon in-game doesn’t matter so much. You get an item, export to AMR, run BiB. Then you know if you should use that item right now or not!

1 Like

Well my point is that generated pawn weight should iteratively adapt so that after you use best in bag feature and you export stat weights to pawn addon, it shouldn’t tell you to replace item to a different one. And this is what i was actually expecting from pawn weights generated from AMR. Right now they doesn’t seem to adapt that much and feel too much static.

I like the best in bag feature, but it takes more time to compare all items on web rather than in game. So I am really looking forward to a solution that would improve generated pawn weights by AMR after using best in bag feature, as a the end I want to compare most of my items in game and use AMR only every few days or when new Azerite gear/Trinket drops.

I guess I didn’t explain it that well… it already does adapt. Every time the gear in your bag changes, the Pawn string changes with it. The analysis is run each time you do the export.

But if I change items based on best in bag suggestions, export pawn string, pawn addon suggest to change the items back, so it means that those generated weights are not in line with best in bag suggestions.
What I’m expecting from stat weights generated by AMR, is that when best items from my bags are chosen based on those weights with pawn - they would be the same as ideal suggestions from best in bag feature (except complex items like trinkets and azerite gear). Currently this is not true.

I want to have a quick option to compare simple items in game, but when pawn suggests to wear 335ilvl cloak, best in bags suggests 340ilvl cloak, I will not be able to tell if another 340ivl is an upgrade or not without exporting game data to AMR on the site.

The Pawn string we generate is telling you the information that you want. It is telling you which stats are most likely going to lead to an upgrade. If you get one or two items that have the stats that trend best, chances are your BiB suggestion will move towards those stats.

I think the disconnect here, @eFse7en is that you use Best In Bags to find… the best gear in your bags. Then the Pawn string is for identifying potential upgrades while in game. You don’t want to do Best In Bags, then take the Pawn string and attempt to shift gear that you already have around, and the Pawn string isn’t really designed to tell you what gear to wear, just what gear might be better than what you’ve got.

I would suggest, based on your usage, that you do BiB, equip the gear, then import the Pawn string. Use that string to find new gear that might be an upgrade, and then at your convenience, run BiB again with whatever you have accrued. Obviously, the more regularly you re-do your BiB and Pawn string the better, and the ‘ideal’ in my mind is to just re-run BiB every time you acquire an item within 5 or so iLvL.

I hope this helps.

1 Like

I think you still don’t understand the problem with the generated pawn string.

If I parametrize Best in Bag setup to give me ideal setup (0% change threshold) and it will tell me that with my gear, Versatility and Crit is most valuable, then I would expect generated Pawn string to follow the same rule.
Otherwise if pawn addon at the moment shows that i should equip other item with more haste, and i should ignore it, then if other item drops with more versatility but much less haste, it might not be even shown with pawn addon as an upgrade but it could with Best in Bag.
So the conclusion is that currently generated pawn weight by AMR are inaccurate and i cannot trust them, which is a shame, because it’s one of the reasons why I bought the premium.

The only workaround I can see so far is to to manually modify generated stat weights so they are more in line with Best in Bags suggestions, but that will take time, it still won’t be perfectly accurate, so I hope there would be a better way implemented in future.

Sounds to me like AMR & Pawn are two conflicting animals that may have little chance of agreeing this side of Raid Gear being available… if at all.

I would ask why you use both… if they do similar jobs, but offer differing advice?
Also - were you a Pawn user before you came to AMR or the other way round?

Neither AMR nor Pawn “tell you what to do”, they suggest… you ultimately have the last say in which advice to accept.

I used pawn before using AMR. Pawn is faster as it gives me immediate answer whether item is an upgrade (if stat weights are correct):

With AMR I have to:

  1. Open AMR addon in game
  2. Ctrl+C
  3. (if the tab wasn’t opened that day) Open
  4. Click “Import from addon”
  5. Ctrl+V
  6. (if the tab wasn’t opened that day) Click “Best in Bag” feature (or scroll down).
  7. (if the tab wasn’t opened that day) Click “Find Best in Bags” on the left
  8. Click “Find Best in Bags” on the right
  9. If item is an upgrade - click on the item, wait for popup to appear, and check how much downgrade would previous item be.

As you can see Pawn in game gives me an option to compare item immediately in a tooltip vs up to 9 clicks with AMR.

Plus with ArkInventory+Pawn I don’t even have to mouse-over item to check whether it’s an upgrade, as green “up” icon will show up in such case:

Let me start by saying I’ve never used Pawn.

However - something tells me that Pawn is better suited to using one set of stat’s for all situations & is less suited to the adaptability that AMR works with… in that given a fixed set of rules, Pawn only allows for limited upgrade options. At the same time, it looks like AMR works on any item being an upgrade, as long as the core (primary) stat’s & equipment proficiencies are met. I’ve been wrong in the past & I don’t expect not to be in this appraisal.

If this is ‘about right’, it may come down to how much flexibility you’re happy to work with.

btw - how does Pawn arrive at any given start point for what to upgrade to? Where does it get it’s foundation Class/spec. stat. start point from?

I think that stock weights used in Pawn come from Wowhead.
As for the how the addon works, it simply calculates weighted sum of stat of given item and compares it using the same method with item you wear. It takes into consideration sockets with assumption you would put there best available gem best on stat weights (or manually configured gem).
You are right - Pawn will not factor Azerite armor, trinkets, set bonuses and won’t help you choose enchants. But as for enchants the best one doesn’t change often, sets are rare in BfA and Azerite armor + trinkets only take up 5 of your slots, so I can still use Pawn for every other.
AMR can of course simulate total amount of stats if this is the best combination to end up with, as pawn will just compare individual items. This of course can be addressed by using Best in Bag feature from time to time. Anyway it’s still more convenient to use Pawn for regular items as it’s just much faster.

Anyway I think I’ve found a way to do an approximation of desired stat weights based on what AMR is suggesting.

I have used this table of gem alternatives:
And put it in the spreadsheet:

I guess it’s the best workaround I can think off for now.

The main problem here is not that one or the other are inaccurate but the fact that both methods show completely different things that have never been and will never be “compatible” - for lack of a better word.

AMR shows the best combination of items through calculations of millions and millions of different gear sets. While a specific item might not be optimal in every respect, combining it with another item can be better than using two “optimal” items due to the complexity of stats and their interactions.

Stat Weights tell you how much DPS you will gain if you add stats on top of your currently equipped gear. Most of the time differences between stats are relatively small and will change with every new item causing different suggestions. Thus, Stat Weights can’t tell you if an item is universally better than another item because that’s not what they’re intended to show.

Let’s take critical strike chance as an example: If you absolutely need to get 100% crit and manage to reach it, additional crit will be worth nothing anymore. Your Stat Weights will then have a value of 0 for crit - which is correct, because having more crit does literally nothing for you.
Importing those weights into Pawn will now show every single item with crit as bad, because one of the two secondary stats is worth nothing. Does that mean that every item without crit is better than what you’re currently wearing? Obiously, it doesn’t - because when using that item you would drop below 100% crit which would be bad in this example.
AMR’s Best in Bags would then still recommend the item with crit which would be the correct decision.

Of course, that’s an extreme and frankly unrealistic example but it does show the differences between those two ranking systems - and why using them in conjunction is more diffcult than you might imagine.
In practice the differences are extremely small but that just increases the complexity of making optimal gearing decisions because even a couple mistakes while playing might have a bigger impact than say having 50 crit or 50 haste.

I understand, but then it means that stat weights exported by AMR only tell you about direction to follow when adding new stats while freezing the current ones. Such stat weights won’t be helpful when imported to Pawn.

Instead I am expecting to get stat weights that when followed, would advise me to equip the same items as suggested by Best in Bag feature (except for items where not only simple stats take factor - like azerite gear and trinkets, or when best in bag change threshold is different than 0%).

So, stat weights are very “coarse”. They provide a linear approximation to a non-linear solution. If you look at the potentially good stat distributions for any spec, you will see that there are very good stat distributions all over the place, due to the complex interaction between gear.

Best in Bags is looking at only the gear you have available and then finding the best solution. Maybe 75% of good solutions have high haste… but you just don’t have a lot of haste gear. In that case, the algorithm finds one of the good sets without high haste and recommends it, because it is the best solution available to you.

Now… does that mean that you don’t want to get haste gear? Do you want to continue gearing for the stat distribution that is potentially not as good? What you are asking is for Best in Bag to find the best stats available to you right now… and then set up Pawn to just keep you there. That doesn’t make sense!

The Pawn string that is generated shows you the stats that will have the highest probability to increase your gear’s value in the future as you get upgrades. And really, even if Pawn says an item is an upgrade or downgrade due to a set of linear stat weights… it doesn’t really matter - you still have to run BiB to figure out if that item is good for you or not. Stat weights give you a very rough idea of whether or not an item is good… but they will miss out on many potential upgrades because they cannot “see” local maximums (or maybe they are stuck in a local maximum), etc.

You’re absolutely right to question the practicality of using the data in such a way - but that’s not a problem AMR or anyone else can solve.

That’s just the way stat weights have been designed from the beginning. Doesn’t matter where you get them from, some simple numbers can and will never be reliable enough to accurately predict the interactions between different stats and determine an optimal gear set. It’s just not how stat weights and Pawn are supposed to work.
For Pawn to work like you want it to it would need to have an entire simulator built in which would be completely unusable due to the amount of time it would take to get a result.

By the way, that’s exactly the reason why AMR didn’t even have stat weights in the beginning and only added them in later when more and more people started to ask for them.

Been buried in work so haven’t been able to read this whole thread, but regarding the default weights in Pawn: he usually gets them from us. As soon as I finish these adaptive gearing strategies that I’ve been working around the clock to pump out for Uldir, I generate a good set of defaults for each spec that Pawn can use as a starting point.

I think he is talking about this scenario:

  1. Run BiB
  2. Equip the gear / recommended enchants / gems /etc.
  3. Run BiB again and see that you indeed have the BiB equipped.
  4. Export Pawn weights from AMR to pawn.
  5. Pawn, with the stats weights just generated for your character, still marks as an upgrade that haste ring you just unequipped instead of the crit one BiB told you to equip.
1 Like

I still miss functionality to generate stat weights which would lead me towards equipping the same gear as BiB. Even if you say that with such stat weights I could miss an opportunity in higher ilvls where haste is much better - that’s ok. From time to time I will be using AMR BiB and updating them anyway. My ilvl won’t be changing that rapidly, so I prefer to have in-game Pawn weight in par with last BiB.

And I still can’t figure out practical use of stat weights generated by AMR as even at higher item levels versatility is first and haste is 2nd at best:
(while AMR BiB weights suggests the opposite)