I began using Mr. Robot for a new class I leveled up and it said that you do not need to be fully hit capped. As long as your under 0.20% your ok. (0.2%>0.19%). Is this right?
I began using Mr. Robot for a new class I leveled up and it said that you do not need to be fully hit capped. As long as your under 0.20% your ok. (0.2%>0.19%). Is this right?
Well you never need to be hit capped. It's just generally your strongest stat until you cap it. What class is this? It definitely doesn't sound right that hit would be worthwhile until an arbitrary number like that and then no more, unless that's the closest you can get with your gear without going way over and wasting a bunch of stat.
It depends on your class. There have been classes in the past (and there are classes now) where one of their other secondary stats is better than Hit. Therefore, they don't really need to bother with hit cap at all (just making sure they don't go over it).
In this specific case, Mr. Robot is not exact enough to calculate that last 0.2%. I recommend using Mr. Robot for gear and gems and enchants, then heading over to http://www.wowreforge.com/ for Reforging because their reforging is 10x more accurate and can get you exactly to hit cap.
Imagine you had X amount of resources as a website to spend on calculations. Mr. Robot uses that X for gear, gems, enchants, and reforging all at once, but to do that it sacrifices a few percent in accuracy (I think they claim like 98% accuracy). WoWReforge ONLY does Reforging, so they can be extremely accurate with it.
Global Moderator | Forum Guidelines
Aff locks don't need to be hitcapped.
And my Rogue isn't, as she gets free Envenom uptime with less Hit.
Isomorphic for LoL EU WestW/L/Death count: Wolf: 0/1/1 | Mafia: 0/5/5 | TPR: 0/2/2SK: 0/1/1 | VT: 1.5/3.5/5 | Cult: 1/0/1Legendary Overlooked for WoW EU-EN.
This is a question that's fundamentally about the underlying math of statistical modeling and the analytical methodology behind how stats can be numerically valued in such a way that should predict dps performance.
If you produce an accurate model for how a dps class works that will stochastically produce parses of performance based on modeling an accurate rotation that takes stat values as input, you can measure expected baseline performance. Then what you do is reproduce that modeling experiment by changing each stat by a small amount one at a time, and finding what the statistically significant difference in mean dps is.
If you repeatedly do this for each stat, assuming that your model is accurate and your trials have statistical power, you can relatively assign each stat point a number that represents something about how much dps per additional point that stat should contribute.
This is somewhat treacherous because there will always be the possiblity for nonlinear stat interactions and inflection points, i.e. every single combination of stats produces slightly different relative values for the marginal improvement of 1 point of each stat to additional dps.
Now this is where a tool like Askmrrobot comes in. It uses a heuristic method to attempt to optimize your total "score" based on static values for each stat produced by a modeling procedure such as above.
So the answer to the OP's question is, we don't really know. We're making educated guesses based on assuming that stat weights being used are accurate, and that the underlying model that produced them is accurate, and the parameters of the original model are close enough to your gear configuration to produce relevant results.
That's basically how all this works.
Hmm. If I am not capped, I stand a chance of missing. If I miss, I might not crit. If I don't drit, my numbers look bad. Guess I better cap, just in case.
The website explains it. That, based on the maths of your chance to miss, your stats are often better allocated elsewhere. That is, you'll gain a dps increase over the fight.
A missed attack is a dps loss, but statistically the benefit from your other stats would outweigh that miss.
Within a reasonable margin of error.
I still stand by copy-pasting Mr. Robot's weights over to WoWReforge and having reforges done there. When I reimport my character into Mr. Robot afterwards, it increases my score and says I'm optimized (which is what it said before too, but it couldn't tell there was a better solution).
Global Moderator | Forum Guidelines
Using ReforgeLite in-game can expedite the reforging process, and it's worked better for me than Mr. Robot or wowreforge.
The biggest 'problem' with being just below the hit cap is that you are more at the mercy of RNG. Essentially, you are gambling that the benefits of the extra non-hit stat will outweigh the loss of hit (relative to being hit capped). For some specs (i.e., affliction warlocks), this gamble is much more likely to pay off than others (i.e., elemental shaman). Furthermore, when applicable, hitting haste breakpoints and/or capping expertise can be more important than 0.2% miss chance.
This is pretty much a giant argument about how many angels can dance on the head of a pin, because no one actually can answer this question with any empirical rigor whatsoever. The reason this is still a "debate" is because there is no real answer.
We simply do not know whether being a minute amount under hit cap in exchange for a miniscule amount of more secondaries produces more average dps than some arbitrary amount above hit cap without specifically putting our exact gear configuration through a very accurate simulation and iterating to enough trials to have statistical power.
We're basically making assertions here about a swing of +/- a number of secondary stat points that is orders of magnitude too precise for the real way that stat weights are produced for general use on these kinds of tools. If you're really simming your character and iteratively optimizing with a significant amount of rigor, that's a different story, but for what people are talking about here, this difference is complete nonsense.
Thinking that there is an empirically-defensible difference between for example 7.53% hit versus 7.48% hit plus a tiny amount of secondary stats is basically to ignore that this entire thing is an exercise in garbage-in/garbage-out when it comes to where these stat weights actually come from.
---------- Post added 2012-11-14 at 08:12 PM ----------
If you care enough to go through this trouble, then you should be determining your own stat weights using some kind of statistical modeling, instead of using some generic stuff from askmrrobot.
Using imprecise stat weights throws away the benefit of what you think you're getting by using a method that you believe is better.
Last edited by underdogba; 2012-11-15 at 01:10 AM.
Isomorphic for LoL EU WestW/L/Death count: Wolf: 0/1/1 | Mafia: 0/5/5 | TPR: 0/2/2SK: 0/1/1 | VT: 1.5/3.5/5 | Cult: 1/0/1Legendary Overlooked for WoW EU-EN.
Yes, I use custom weights as well, almost always from SimulationCraft but sometimes generics sets for my alts.
I only said what I said because the OP doesn't seem like the type of person who would go through the trouble of setting up SimulationCraft for their character and waiting 10+ minutes for accurate weights.
Global Moderator | Forum Guidelines