|
Diamond Flask is 12.5str on average but on a 3min fight its 25str avg and best of all it stacks with death wish and execute so its effect is amplified. 75str for 1 min is pretty nice for a level 50 blue trinket. The same reasoning makes all other activated trinkets better on bosses than straight averages would suggest.
As far as mod flurry chance, my understanding is that it represents the % of time that flurry is up. This can not be 100% by definition. I dont have the formula on me but basically you have to figure out how many bonus swings per minute you are getting which help keep flurry up more. I'm sure someone else has the correct formula
As far as modeling eskhandar, it is an excellent agro weapon and is approximately equal to hakkar loot for dps. I actually do use it for level 60-62 trash mobs due to very high speed giving me a higher chance of getting 2 or 3 executes off before the mob is dead instead of 1 or 2. Faster weapons have lower std dev for rage gen. Similar arguements help explain why +hit is better than mathematically expected. Warriors need very consistant rage generation to maximize dps. Any rage spikes or troughs can really screw up your sustained dps. Fast mhs also let you burn off rage faster if you happen to get stunned and hit 100 rage (brood power green). There are 2 haste procs, 1 haste trinket and 1 juju haste that I can think of currently, Blizzard may very well add more as its an interesting mechanic. Modeling it may be more important if a haste proc weapon drops in Naxx that we don't know about.
You may want to add std dev calcs to figure out how often abilities can really be used. You can't just assume consistant rage.
I don't know if you are interested in modelling threat per second for dps spec warriors. Knowing how to maximize our agro when needed for tanking is useful.
Either way, i highly enjoy this kind of thing, can be contacted at myname @gmail.com if you are interested in getting help or perhaps turning this particular spreadsheet into an addon (that could link into real time damage meters to avoid messy stddev calcs + other unmodelable things such as out of range). If I had more time I'd have already written it but I'm certainly interested in testing/debugging such a thing.
|