Its the rubber neck effect of people driving by a horrific car wreck.

-Expected lapse between WF procs is exactly (6-1)S

Honestly, if you want people to believe and accept what you're saying you're going to have to show some real world proof via parses and what not. Its already been simulated and tested that a certain way is better and to disprove that you have to show results.

It can not be ?

Hi - I'm challenging one of the basic assumptions of how your class works with some very simple, very clear math. This seems worth discussing. Address my logic?

No, you have to take the CD into account any time your attack rate goes below 1 per 3 seconds.

Consider a weapon with speed 0.7 - say it procs at time zero.

The first strike (0.7) cannot proc, as it is within the cooldown
The second strike (1.4) cannot proc, as it is within the cooldown
The third strike (2.1) cannot proc, as it is within the cooldown
The fourth strike (2.8) cannot proc, as it is within the cooldown

Now, tell me whether you honestly believe that the fifth strike will ALWAYS proc? Because that is what your model is assuming?


Next, look at a weapon with speed 1 - say it procs at time zero

The first strike (1.0) cannot proc, as it is within the cooldown
The second strike (2.0) cannot proc, as it is within the cooldown
The third strike (3.0) cannot proc, as it is still just within the cooldown
The fourth strike (4.0) can proc, since it is outside the cooldown
The fifth strike (5.0) can proc, since it is outside the cooldown

Your model says that the 5th strike will always proc - that's not the case.


Finally, to make this as simple as I possibly can - let's use the analogy of a dice roll. If you throw a six, you get a WF proc.

You are modelling it that you roll a 6 on every sixth roll of the die, which works fine if you ignore the cooldown. Now imagine that whenever you throw a 6, you have to ignore the next five rolls.

So, you roll a dice and throw a six
You roll it five more times and ignore the results
What do you think will happen on the sixth throw? Are you guaranteed to throw a six? After all, you've rolled six times, it must be "time for another six" by now. Bad news - you're not.


To all the people defending him and saying "this is deep" - NO IT'S NOT. It is an ELEMENTARY STATISTICAL FALLACY, and is taught around the age of 11 or 12. Give up, now.

The idea that a 3.1 sec weapon is outside the cooldown range of WF challenges some basic assumptions?

I'm putting this thread on ignore. If something remotely note worthy comes out of it, I'm sure someone will post it elsewhere for inclusion into the enhance thread.

No. A faster weapon doesn't equal more WF procs because WF procs are capped in time via the 3 sec cool down. You might want to re-read the Windfury tooltip.

Are you going to tell us that it's better to have a 1.3 speed main hand dagger and use WF totem instead?

heel if you completely ignore flurry and any sort of dual wielding mechanics and assume that you see exactly a 20% proc rate (which you don't) then the DPS increase from Windfury for weapons between 1.6 and 2.9 is the same.

Not even, because of the AP to weapon speed multiplication on WF swings being not normalized.

Good, good, I'm glad that's out of the way. Another question: why wouldn't you see an average procrate of 20% outisde of the 3s cooldown? That's what's in the tooltip.

For the very specific example he's talking about the proc rate and increase in DPS from WF non-normalization will cancel each other out. This isn't how it shakes out in practice because of Flurry, however.

He said 20% of all swings, not 20% of swings outside the cooldown.

We lost the opportunity to have your greather input then

Yikes.

As someone who has also written a Sim, and I am sure others will agree, one of the first things that really surprised me about my data is how much shaman dps can vary given the same inputs. Time samples (10-20 minutes) that I would have assumed would yield somewhat flat results were not flat at all, they still varied by 10+%.

Even running a sim for an interval of an entire day will still show variations of 1-2% at times. What this means is that no one is ever going to solve this problem because it is simply not possible to run an ingame test that would confirm any sim, the data will always be too noisy.

Aside from seeing source code from bliz getting within 5-10% is pretty much the best shamsims can hope for.

Great, how much big is your timer tick ?

But that isn't correct. When you keep rolling a dice, there is no void rolls, there is no ignore all 6s for the next 3 sec after rolling a 6.

Hi - I don't believe gravity works like everyone agrees it does, and until you address my points one by one I'm going to keep posting about it.

This thread was dumb when it started and it's still dumb. The answers for the behavior of DPS modeling are addressed elsewhere with far greater understanding than the OP. Rather than learn from those models, the OP wants to reinvent the wheel and only asks that we come and prove to him where he's wrong. Pass.

Thread ignored.

Edit: Although I'll admit that I'm glad this thread is a stand-alone one, or else this discussion would be fagging up another useful thread right now.

Do you agree that 2 dice rolling are not correlate event? So I can throw the dice for each swing and the ignore a WF proc if last WF was less than 3 secs from now? And that will not change anything ? Time to raid now...

20% of eligible attacks will proc Windfury. In other words, on average, every 5th eligible attack will proc Windfury.
Attacks within 3 seconds of a previous Windfury proc are not eligible attacks. You do not account for this.

Ok, I will stop this because clearly you know all better than I do and all you have already written all on the matter, nothing can be written more than you or someone else did. If you are interested in the simulator output or have the code jsut send me a private message.

Good luck with the your own theoricraft, I thougth EJ was a forum open mind, it seems is not.

From what I have read, you are completely unwilling to admit that your initial assumptions about the way the mechanic works might be wrong.

Just above this post is a post by lactose explaining, very concisely, where your assumptions err.

The windfury proc change is a piecewise function (albeit a simple one), and not a static 20% chance per swing.