BiS improvement - let's bring some stats

Hey !
I got an insane idea for BiS improvement.
AMR calculate the best items by getting a score for each item (or close to something similar).
What if we bring some stats with a probability to be at BiS.
What is the probably to reach BiS next week ?

We can take few hypothesis like 'if you unlock 3 MM+ slot there is 100 items then you have 3x1/100 chance to get the BiS weapon that you like.
But a crafted item has a probability of 100% because it’s deterministic.
Based on the hypothsis that some item are equivalent.

Here is a concret exemple :
BiS tell me this is best :


But the probability to get it is very low (1 item over all of the MM+ dungeon pool in a vault) So i excluded this item.
Now BiS suggest a combo between a crafted weapon and an offhand:
image
Great probability has increase but the off hand can be swapped between 3 (4 other off hand):

The probability for me to have the BiS configuration is 3 times higher.

We could event expand this to consider how many classes wants this item but it would be a cross spec BiS and would be too big to compute.

So what i would like BiS to do is to take this into account and bring me to a high probability build while minimizing the gap compare to the maximal performance achiveable

What do you think ? Is it absolutly horendous idea ?

Funny enough… Swol and I were just talking about some features like this that calculate the probabilities of obtaining certain items. It would of course be approximate as we don’t know the exact drop tables and chances, but even a ballpark number would be potentially useful.

We would like to add some features along those lines at some point – we have already started working on several improvements now that we have some time between major patches/expansions.

1 Like

Big brain time I read you mind !

Can AMR addon provide you with updated statistics? Not directly but maybe if you could note what were the item in the vault you could do some stats to see how the loot table evolve and use this current season to kind of predict the next. Maybe even use bayesian statistics to get and update assumption for the next.
Maybe too much data to store…

What and interesting topic ! Almost as interesting as simulation !