Calculating Overlaps and BRE/Crusader Uptime

[Iskato 0]

Hey guys,

I currently find myself in trouble finding out the actual uptime of procc based buffs like Crusader or the one of BRE.

My issue in fact is taking account overlaps of procced buffs into the calculation.

 

Anyone with some experience on that matter or someone that can explain how the spreadseet works on that?

 

Thanks.

 

[Storfan 24]

The Procwatch-addon got a PPM-function that will take into account overlapping procs. Crusader is 1 PPM with a duration of 15sec

15/60=0,25

25% uptime x 100str = avg of 25str or 50 constant AP.

 

Important to note is that while Crusader got a 1PPM, internals CDs was not "a thing" in vanilla meaning Crusader ACTUAL uptime will go up with a slower weapon and WF-totem and other modifiers.

[FullRetard 1]

Procwatch doesn't take into account overlapping procs. It just counts the number of proc messages in the combat log and divides them by the total time in combat. Uptime of crusader will be slightly under 25% if you are just autoattacking with no haste, due to some being lost to overlapping procs.

 

None of this really helps Iskato with real world situations...

Edited February 6, 2017 by FullRetard

[Iskato 0]

Thanks for the replies,

 

as Fullretard said, overlaps is the issue that kills me if I wanted a 100% accurate calculation or simulation.

The thing is, I know how to "exploit" the mechanics of PPM and PPH to increase the uptime, but I can't calculate the actual uptime of the buffs to simulate my DPS in order to compare items.

 

I'm afraid it has to go in this direction: https://iam.yellingontheinternet.com/2012/10/16/theorycraft-101-how-to-compute-uptime-of-a-proc-based-buff/

But since I'm a pleb in maths someone has to break it down for me ;)

Edited February 6, 2017 by Iskato

[Zarant 2]

Modeling crusader uptime is kinda difficult not only because of the overlap issue, estimating the average number of attacks per second is not easy, it includes a lot of other variables, like flurry uptime and extra rage spent with heroic strike and hamstring.

[Zarant 2]

Modeling crusader uptime is kinda difficult not only because of the overlap issue, estimating the average number of attacks per second is not easy, it includes a lot of other variables, like flurry uptime and extra rage spent with heroic strike and hamstring.

[Vilius 1]

IMO, the best way to calculate this sort of thing is with what's called a Monte Carlo simulation.  There are other techniques but not as accurate and tough to estimate how far off they are without doing the Monte Carlo, so might as well do the Monte Carlo.  Unfortunately excel isn't really the best tool for this, but I have some code that will do it.  If you can estimate your # of attacks per minute (yellow and white) I can tell you your approximate crusader uptime.  Even this is an approximation of sorts, because in reality your attacks are not at equal intervals.  You might have a burst of attacks and then fewer for a time, and this pattern can affect the results.  But Monte Carlo with the regular intervals is better than anything except Monte Carlo with the irregular intervals (more complex to model).

I'm actually working on a more complete warrior simulator, but not ready yet.

Edited February 7, 2017 by Vilius

[Zarant 2]

If you have some kind of way to simulate a static 1 minute fight you don't really need to care about it, you just run the simulation a million times and figure out the average number off attacks per minute. I guess that's what the monte carlo method is all about, but if you manage to do that, the crusader uptime information kinda loses its meaning because you could extract much more useful information from that. Most spreadsheet models are all about ballpark math and thus, only accurate for really long fights.

[Vilius 1]

3 hours ago, Zarant said:

If you have some kind of way to simulate a static 1 minute fight you don't really need to care about it, you just run the simulation a million times and figure out the average number off attacks per minute. I guess that's what the monte carlo method is all about, but if you manage to do that, the crusader uptime information kinda loses its meaning because you could extract much more useful information from that. Most spreadsheet models are all about ballpark math and thus, only accurate for really long fights.

Yes that's essentially what monte carlo does.  And the proc part of that sim is almost trivial.  Of course lots of other things can be extracted, if they are part of the model, and I'm working on that.

I'm all for spreadsheet models for some things, but this is one case where they aren't really in the ballpark at all.