I was impatient so i was looking to get some result for my monks, specially about the “Conflict and Strife” one.
So i decided even during GloNet running to do my own sims. And not one or two but a bunch of them.
An so, i have preliminary result while this is running in my background.
Here is the first result i got:
So let’s imagine this is all the data i can get.
If you are looking to the essence “the Ever-Rising Tide” it represent 75% of the solution withing 2% of the best HPS (for the major Essence).
BUT, it’s no the “best one”, the best one is “Life-Binder’s invocation” and in margin of error “the Ever-Rising tide”.
Because in 75% of the case “the Ever-Rising tide” is the very good (~withing 2%) i would take this one and in “nich” time take “Life-Binder’s invocation”.
But i dont think it’s what AMR is doing.
Or it is ?
If i have the full choice (all unlocked Essence), AMR would provide me with Life-binder’s invocation as a go to even so, 75% of the case, “the Ever-Rising tide” is very good whatever the other essences are.
So, this is probably a coincidence for my specific case.
I wonder if it could be generalized to:
- Reduce the number of iteration to do ?
- Avoid some edge cases where gear/iLvl goes nuts ?
Well this thread was just for me a good way to write down some of my thought ! I hope this isn’ t a waste of time for people to read and could improve AMR ! =D
P.S: i had another idea ot share from a discord discussion we had…
What if. you use a genetic algorithm base on Neural Network that can “evol” by changing one gear each time, sim it and get the next best gear set you can have.
My head is exploding it could be SOOOO cool (in my head it’s cool ok ? xD)
i would need some number to calculate if it could generate less iteration and be faster for AMR to do