Get back to old stat weight setup?


#1

Is there a way to go back to the old way of inputting stat weights? I use Simc myself to find Stat weights and used to put them into AMR and the old way worked great. I tried using the ‘Imagination’ mode and it messes my gear way up. Is there any way to go back or make the new modes act like the old? I can edit strategies I made before the change went down but can’t make any new ones.


#2

You will have to use the new way, which requires telling us how good all of the special items are relative to your stat weights as well.

It was more or less just luck that plugging custom stat weights into one of our gearing strategies resulted in good gear rankings… they just aren’t designed to work together like that.

If you don’t want to enter things in by hand, you could create a custom gearing strategy with our simulator, which will generate a coherent set of data that ranks all the special item effects and stats together. If desired, you could then enable stat weight mode instead of machine learning mode, and then tweak the stat weights that it produces slightly to get a different balance of stats.


#3

Hi yellowfive and sorry zeke911 for hijacking your thread, but i thought it fits the question you raised and doesn’t really need an own thread.
I don’t really get why it is necessary to put in the weights for legendaries, trinkets and so forth by yourself at imaginary, even when the engine may need it.
Either you take a guide as base for those weightings (what probably most would do) which most likely will result in the same input every time except of small variations based on whom the guide taken as base is from. Or you sim all the items in the list by yourself in all variations and get your weightings that way.
The first case cries for defaults, because there isn’t much variation, the second case isn’t practicable, because it probably takes the same amount of time to sim all those variations by yourself as it would if you simulate a whole gearing strategy. You even could just take those defaults per spec (which would probably make most sense) and let them be overridden per item IF another value is assigned within the mask.
Maybe you could think about this suggestion, I’m sure the imaginary option anyways isn’t how you really want people to do things, but it still comes in handy to quickly put in some weights based on a snapshot and get a quick look on the next best thing instead of whole gearing strategies sometimes and no offense, but at the current state letting a gearing strategy run through just to alter the stat weights to the values you like afterwards isn’t much of an option either at the current state, because it simply takes way to long, which for sure has it’s reasons, but doesn’t help it’s case within the given context.

Kind regards,
Tectas


#4

You could run a custom strategy once, and then just keep changing the stat weights whenever you feel like it. Don’t need to re-run the strategy to modify the stat weights on it.

Or, you could enter in some reasonable values for all the special items once on an Imaginary strategy, and never change them – just keep editing the stat weights when you want to change them.

Either way – you need to get yourself a coherent starting point. Very soon all of our defaults will be replaced with our new adaptive strategies, and they just don’t work with stat weights – they are a completely different method. You need to generate some kind of starting point for yourself that makes sense.


#5

Well, yes, sure, but still defeats the purpose having to run a ~4 hour gearing strategy (has been ~1 hour about a month ago btw) when I anyways use different values afterwards.
And yes, I could simply enter the values myself, but just take a guide, even there I first have to translate the text of “nice for survival” to an weighting, which can make a for sure not really accurate approach even less accurate.

I’m appreciating the effort and thoughts you put in the base for sure, but for most people I talked to the huge difference in time for generating “stat weights” (even while it for sure is way more at amr, because it is a whole gearing strategy and takes way more into consideration) is a big advantage for simcraft over amr. The reason is pretty simple, 5 minutes, even when you have to do them over and over again aren’t much of an issue, you can even kill a boss and if you get some loot can have an updated string at the next boss already. You wouldn’t need to do that with AMR, sure, but that’s unfortunately no use if you want to update your pawn string quickly.
Unfortunately it even can be needed to spend not only once the ~4 hours initially, but multiple times, if there is more than one talent setup which can be used, not to mention different boss scripts.
An simple way to use your other tools based on stat weights, even while not accurate that way would be a way to get more people on board.


#6

I guess part of this is that I just don’t understand the motivation here…

So you want to use our ranking data for legendaries, set bonuses, relics, and trinkets. But you want to use stat weights to choose how much of each stat to get?

Or are you saying: you don’t care what we suggest for legendaries, set bonuses, relics, or trinkets – you want to just use it to optimize with your stat weights, and then you will pick the other stuff yourself?

If the latter is the case, just go ahead and do that… for example, here is a retribution strategy I just made with some random stat weights: http://www.askmrrobot.com/wow/strategy/PaladinRetribution?strategy=e3e73e19a369458e9327c76fe0f75908

I put in no value for special stuff, it will all be worth zero. It optimizes just fine, picks gear based on stats alone. You could then lock in the special items you prefer and it will pick stats around it.


#7

Basically yes, especially at legendaries there isn’t much variation, they anyways have the highest gear score in the game and there are pretty much fixed BiS legendaries in many cases at least based on which talent setup is used (if even).

Trinkets are for sure another beast, but there are still many where the effect is either useful or not.
Sure at both cases the stats itself are of interest also (and it get’s more complicated when the effect grants a stat), but depending on how good the effect is it takes or gives value to the importance of the stat (but whom do I tell this anyways, probably the effect is already just some kind of modifier for the stat value of the given item or the other way round).

Letting those values empty and just considering the stats is for sure a possibility that at least tackles what I was talking about, but unfortunately isn’t really a solution because nothing else has a value that way, which sure, is a way to look at it, but that’s even to rough. You do have your basic strategies set up anyways (even while I must confess that I don’t understand why the current default strategies, at least for tanks, are mythic+ based, but that’s another thing and doesn’t really matter anyways) where the weightings for legendaries, trinkets, relics,… are present already, even while they won’t fit perfectly for sure, they will in most cases at least fit way better than just taking the stats into consideration.

When I think about it, the easiest solution would be a way to simly clone your default strategies and make it possible to alter those clones.


#8

Our adaptive strategies (which will replace all defaults very soon) can’t really be cloned… they are massive. They use a huge number of data points under the hood, and use a statistical technique that factors in all the special effects in conjunction with stats. You can’t alter one piece of it (just the stats) to something arbitrary like stat weights… that would break it.

Also, there is a lot more variation in legendaries than most people think. For example, there are 6 different legendary combinations for a retribution paladin that simulation within 2% DPS of each other in a BiS setup. And for any particular player, unusual combos can become as good or better than the “normal” ones – the stat gain on legendaries is pretty massive and can more than make up for a weaker special effect. It can depend heavily on how good your non-legendary alternatives are.

That is why we created this more in-depth, data-driven model in the first place… it can figure all these things out for you, and account for all these interactions. But I understand that some people just want to do their own thing, and that’s fine. Thus the “imaginary” strategies.

I really wanted to get two things across with the imaginary strategies: firstly, you can customize the optimizer to do almost anything. It is a completely separate system from our simulator, and you can plug in any values you want from any source you want. Secondly, it doesn’t make sense to only enter stat weights to rank gear. You need to think about the value of all the special effects and how those relate to/scale with your stat weights.

The imaginary strategies are very easy to play around with. Here’s a couple rules of thumb to get you started:

  1. A “good” legendary effect is typically worth around a 5% DPS increase for damage specs. You can ballpark that at around 2000 rating of a secondary stat. So start by making it worth around 2000 * your weight on a secondary stat, and tweak from there. Some effects are certainly stronger, many are weaker. But that gives you an idea of the magnitude you should be using.

  2. Leave stuff blank that you don’t care about – items you think are bad or will never use.

  3. For trinkets, start by just putting the tooltip value of the special effects at the specified item levels in each box. Then multiply or divide them all by the same number to get a good starting estimate.