[Shaman] Experimental combat simulator - Page 3

Bink · 08/08/07, 9:31 AM

It really is just the joy of statistics messing with you. Take a look at it like this:

4 speed weapon:
Swing1 (1 in 5 chance)
Swing2 (1 in 5 chance)
Swing3 (1 in 5 chance)
Swing4 (1 in 5 chance)
Swing5 (1 in 5 chance) (proc)
Swing6 (1 in 5 chance)
Swing7 (1 in 5 chance)
Swing8 (1 in 5 chance)
Swing9 (1 in 5 chance)
Swing10 (1 in 5 chance) (proc)
Swing11 (1 in 5 chance)

1.5 speed weapon:
Swing1 (1 in 5 chance)
Swing2 (1 in 5 chance)
Swing3 (1 in 5 chance)
Swing4 (1 in 5 chance)
Swing5 (1 in 5 chance) (proc)
Swing6 (NO chance)
Swing7 (NO chance)
Swing8 (1 in 5 chance)
Swing9 (1 in 5 chance)
Swing10 (1 in 5 chance)
Swing12 (1 in 5 chance) (proc)
Swing13 (NO chance)
Swing14 (NO chance)
Swing15 (1 in 5 chance)

This really is oversimplifying it, but it might help you see why the fomula has to account for weapon speed as well.

Kagekami · 08/08/07, 9:33 AM

Originally Posted by Bink

Both ways of saying it work. And they both mean the same thing: you have no chance of getting a WF for those swings.

If you have a 0% chance to proc or have a 20% chance to not proc... what does the terminology matter? You can't proc. If you can't proc, then that swing doesn't count to the 1 in 5 average you are doing.

The terminology matters, as in draghkar's example you would get loads more wf procs overall. His saying is that there has to be a wf proc in 5 swings, even if some are inside the WF cd.

Say 4 swings are inside wf cd, 5th outside. (theoretical 0.75 weapon speed.)

With draghkar's theory this 5th swing would always proc wf. In reality, it is not so. At least not as far as I know.

(Didn't dare use AFAIK, don't know what abbreviations and TLAs that are accepted yet.)

draghkar · 08/08/07, 9:36 AM

Those are the graph simulating 1milion seconds fight, considering WF CD

as you can see (without consider crit) you have 2 steps: 1.5 seconds speed, 3.0 seconds speed, after 3 seconds you total average damage is the same. Increasing the crit % you have another step at aroung 4.3 secs. After that nothing changes. See also this immage (33% crit graph zoomed) it shows that 1.7 speed is even better than 2.0 speed, then you gain again at 2.2 secs ( having better damage that 1.7 ).

EDIT: Both images are wrong due to some glitch on the simulator, more specificaly the second graph was affected by a too long resolution timer, 0.1 second instead of 15 ms.

Malan · 08/08/07, 9:39 AM

I've got to admit, this is entertaining.

draghkar · 08/08/07, 9:44 AM

Originally Posted by castille

dragkhar, just because you want something to be true doesn't mean it is.

I don't want something true, I'm just showing what my simulations are showing, I like that most of you are trying to explain me where my model is wrong but till now I still see it correct. As soon as the code is in good shape (WF at the moment is hard coded in the weapon for example and so I can not add to a weapon a generic enchant), I will put it in google code.

draghkar · 08/08/07, 9:52 AM

Originally Posted by Bink

It really is just the joy of statistics messing with you. Take a look at it like this:
This really is oversimplifying it, but it might help you see why the fomula has to account for weapon speed as well.

Were is written that the dices are thrown only when not in CD? You can have same effect with:
Dices are thrown at each swing but if in CD the WF doesn't proc.

This is what the simulator shows in his combat log in case of 2.3 speed

2.3 HIT
4.6 HIT
6.9 HIT
6.9 FLURRYON
6.9 CRIT
8.5 HIT
10.1 HIT
10.1 WF
10.1 WF
11.7 FLURRYOFF
11.7 HIT
11.7 FLURRYON
11.7 CRIT
13.3 HIT
14.9 HIT
16.5 FLURRYOFF
16.5 HIT
16.5 WF
16.5 WF
16.5 FLURRYON
16.5 CRIT
18.1 HIT
19.7 HIT
21.3 FLURRYOFF
21.3 HIT
21.3 FLURRYON
21.3 CRIT
22.9 HIT
24.5 HIT
26.1 FLURRYOFF
26.1 HIT
28.4 HIT
28.4 WF
28.4 WF
30.7 HIT
33 HIT
33 FLURRYON
33 CRIT
34.6 HIT
34.6 WF
34.6 FLURRYON
....

Juice · 08/08/07, 9:58 AM

I recommend you take your simulated performance and then attempt to repeat that performance in game with equivalent gear, weapons, etc. That's what Disquette (I'm leaving out others due to poor memory, sorry) and company did with their simulators. I think Disquette, et al., found their simulators to be within 5-10% of actual. That's a pretty good indicator that they are close.

Bink · 08/08/07, 9:59 AM

Originally Posted by draghkar

Were is written that the dices are thrown only when not in CD? You can have same effect with:
Dices are thrown at each swing but if in CD the WF doesn't proc.

Simulation:
I roll a 100 sided dice, and I get 18. Woohoo, that would be a proc! Oh wait, I'm in cooldown. Darn, no proc.

Hmm.. Now lets think about that. If I can roll every number on this dice, but I can't proc, wouldn't that mean I have no chance to proc? If I have absolutely no chance to proc on this swing, doesn't that mean I have 0% chance to proc?

If I have 0% chance to proc on this swing, then basing an equation on the idea that EVERY swing has a 20% chance to proc would be flawed.

Now, to be more helpful, for the non-dual wielding, no flurry having, not going to stormstrike shammie, here's the formula:

Damage(TimeInSeconds)=WeaponDamageRange*(TimeInSeconds/WeaponSpeed)+TimeInSeconds/(5+MAX(0, 3-WeaponSpeed)*(WeaponDamageRange*2+WindFuryBonusDamage*WeaponSpeed*2)

david0925 · 08/08/07, 10:07 AM

This is like one of those forum QQ posts saying X item should drop in Y runs because he did 1/Y dungeon runs, statstics doesn't work like that. When you take statistically like that, things get messy.

You need to simulate individual hits to have 20% proc rate, and if the most recently active proc and the next proc are within 3 seconds, ignore the proc. To space out every proc evenly means you can completely skip the 3 second rule, which is a pretty bad way to ignore mechanics. Especially right now, the concern is with Dual Wield, when both weapons are active with flurry (most of the time, if not all), it is quite possible that consecutive windfury proc, but are not triggered due to 3 second internal cooldown.

This entire graph is taking statistics in a biased interpretation. If you cannot accept that then there really is no point for this discussion to go on

heel · 08/08/07, 10:12 AM

Originally Posted by Juice

This thread needs to go away.

Absolutely not. The OP isn't stupid - he's just viewing the problem from a different angle. Any discrepancy in results will undoubtably be cleared up as his methodology is nitpicked further. He's probably wrong, but hey, he might have stumbled onto an aspect of the problem that has been ignored thusfar. In any case, like I said, he's not stupid, and he and everyone else will doubtless come to perfect agreement within a few days. No reason to remove this thread.

draghkar · 08/08/07, 10:16 AM

Originally Posted by Bink

Now, to be more helpful, for the non-dual wielding, no flurry having, not going to stormstrike shammie, here's the formula:

Damage(TimeInSeconds)=WeaponDamageRange*(TimeInSeconds/WeaponSpeed)+TimeInSeconds/(5+MAX(0, 3-WeaponSpeed)*(WeaponDamageRange*2+WindFuryBonusDamage*WeaponSpeed*2)

To be even more help full I believe that with WeaponDamageRange you mean WeapongSpeed * WeaponDPS, and also I believe second part of the formula is wrong, in case of WeaponSpeed > 3 then WF part is nullified.

What you tried to write is:

Damage(Y secs)= speed*dps*(Y/speed) + Y/(speed*5)*(speed*dps*2+WindFuryBonusDamage*speed*2) =
dps*Y+ Y/5*(dps*2+WindFuryBonusDamage*2)

and as you can see speed doesn't matter, of course this is valid for speed > 0.6 (othewise you have to consider WF CD as well.

Bink · 08/08/07, 10:18 AM

Max(0, 3-WeaponSpeed) is how you take into account when the weapon is faster than 3 seconds. It means take the largest number of 0 or 3-WeaponSpeed.

Bink · 08/08/07, 10:21 AM

Originally Posted by draghkar

Damage(Y secs)= dps*Y+ Y/5*(dps*2+WindFuryBonusDamage*2)

if you want me to write it that way then...

Damage(Y secs)= dps*Y+ Y/(5 + Max(0, 3-WeaponSpeed)*(dps*2+WindFuryBonusDamage*2)

Malan · 08/08/07, 10:21 AM

Originally Posted by heel

He's probably wrong, but hey, he might have stumbled onto an aspect of the problem that has been ignored thusfar.

No he's definitely wrong, and his 'model' is about as good as I could make with an Excel spreadsheet - which is to say, pretty bad. I'm terrible at math and even I can look at this and just see that its wrong.

Lactose · 08/08/07, 10:25 AM

Let's say you have a weapon that, after a Windfury proc, is able to land 4 swings within 3 seconds.
Is your belief, dragkhar, that swing #5 will proc Windfury?