Since I saw that the vast majority (>95%) of ranked ferals at normal, heroic and mythic levels are using Niya’s Poison (to get Carnivorous Instinct) over Niya’s Burrs that AMR has consistently recommended, I played around with BiB and different conduits. When I changed to CI + NT:P, BiB said it is about a 4% increase to go back to NT:B. Pretty strong result, it is why I’ve not followed the cat pack thus far.
Looking at my logs from this week, the most NT:B did on a boss fight was 3% of my damage. But it was only 0.7% on second lowest boss and 0 on Rygelon. (zero! I died just over 2 minutes in but still, TF was up 33% of that time.) Across all kills, Burrs did 1.92% of my damage.
Tiger’s Fury was typically up 35-40% of the time (convoke can cast it). With CI at 252 ilevel, it is a 6% boost while TF is up, about a 2.1 to 2.4% increase overall. Right?
I did as suggested in another thread and customized values for NT:B (to .6) and CI (to 1.3) and BiB still went with Burrs. The Robot strongly recommends burrs! With CI at 1.4, it changed recommendations to CI – and to use Dreamweaver instead of Niya.
I’m not sophisticated enough to prove that NT:B is overvalued but it is suspicious to me that Burrs is considered a big increase when my (limited but actual) data has it as a decrease.
Those soulbind nodes are a bit difficult to rank… Burrs can be pretty variable from fight to fight, but it is theoretically quite strong in an optimal case. Definitely modify it if you don’t tend to get full value out of it.
Niya’s Poison is also hard to rank because some fights you can’t make use of it, others it’s great. When you can interrupt consistently it’s a good choice, so go ahead and take it or lock it in.
I think the customization tool might not be working with Niya’s Poison… I’ll see if I can tweak it so that you can put in a multiplier and force the optimizer to choose it without locking it in.
I know the default script has a chance of random movement for us but I don’t think the boss moves, so there shouldn’t be too much different between it and the Target Dummy result and there isn’t.
Now I’ve typed that I don’t know if the targets move in the M+ script either, they tend to move in the actual game but I don’t know if you modelled that in the M+ script.
While the differences aren’t huge but it’s been consistent over a few spot checks I’ve done, as I got tier gear. Also when you lock the Poison path in it shows a -2.93% result, which puts the combined difference at about 4%.
Given these results it surprises me that the optimiser is choosing Burrs, I know we can customise the rankings but this indicates we shouldn’t have to.
I think you should lower its value based what we are seeing in the “real” game world.
I looked at the top (500?) logs for feral on heroic Skolex with the idea that is the most Burrs friendly boss, and there were only 9 ferals who used it. The average damage it did was around 2.3 (8 values were 1.3 to 2.4 and then 4.34, very much an outlier; 2.06 avg w/o the outlier) which is about the expected value of CI. That isn’t 3-4% above Carnivorous Instinct (CI), plus those numbers are from, arguably, the best case raid scenario for Burrs. I think Burrs should have less value than CI.
Unfortunately it doesn’t work that way, someone loud will say something about what is “best” and then many players copy it, someone gets a good log and it gets reinforced as best.
Then it feeds on itself and there will never be enough logs to get a fair representation of what the other options are, in this case Burrs.
Having said that, if the simulator - which BiB uses - backs the third potency conduit and Niya’s Tools: Poison something screwy is going on.
e30eb9d054634df7bfb65c8b0e45e3be
Since the change it’s now telling me to use Dreamweaver.
Forcing Niya indicates -0.10% but Niya sims better. 11,664 DPS BiB since update 11,822 DPS BiB locking Niya
It isn’t that it’s more popular @yellowfive it’s that it isn’t matching your simulator results, it doesn’t match SimC either but that’s irrelevant as the optimiser is using your simulator.
As I have said – the optimization doesn’t exactly follow the simulator in all cases, sometimes we tweak things to cover more cases, etc.
It’s hard for me to say why it would prefer dreamweaver in this case… it could be that one of the dreamweaver soulbinds is ranking high. When I have time I can try to look into it… though we’re only talking about 1% difference or so here. Honestly you’re going to have a hard time feeling any difference in-game (which has been one of my big complaints about the whole soulbind system since day 1).
oh, I thought that was the point of GloNet and all the pre-run data, I knew you ran some fancy machine learning magic over the results, I didn’t realise there was more to it. I thought it would follow the sim data results.
I understand it’s a small margin and hard to notice, in cases like that I’ll have a think about what the abilities do in a game situation. In this case I’ll look Niya in as melee moves more than ranged and I can see the tank moving things out of the circle of goodness Dreamweaver drops for me to stand in!
It would be nice if the optimiser then chose the higher path through the tree but I’ll just lock that in too and run some times as I upgrade different things.
It’s also handy if Feral uses Niya so Terror Turkey can use Dreamweaver as I swap between them for trash because doing trash with ST Feral talents is horrible and it isn’t worth two Tomes between most bosses!
