Dodge stacking

[magicrectangle]

I just started a monk (second char), and wanted to know how dodge stacking works. A cursory search failed me, so I reverse engineered the formula, and thought I'd share it with you all:



So it isn't possible to get 100% dodge chance by stacking dodge buffs (unless one of those buffs is 100%, which doesn't exist in game). Dodge buffs are multiplicative, but not directly as dodge, rather chances to be hit are multiplied instead. At very low dodge numbers, this is very close to adding dodge percentages, but at higher dodge numbers you will see it falling significantly short of what you'd get by adding up your dodge bonuses.


Process of reverse engineering, spoilered for cleanliness:

Spoiler

Raw data:
17.2 base
29.6 guardian's path (+15)
40.2 base + guardian + mantra (+15 +15)
49.1 base + guardian + mantra + mantra temp (+15 +15 +15)
49.7 base + guardian + mantra + lightning flash (+15 +15 +16)
57.3 base + guardian + mantra + mantra temp + lightning flash (+15 +15 +15 +16)


Assumption: reasonable humans came up with the formula. Rather than using regression, we guess that if they aren't adding or multiplying dodge chance, they're probably adding or multiplying something else. If I were designing it, I'd multiply chances to be hit, so lets test that.


raw dodge = .172
raw chance to be hit = 1 - raw dodge = 0.828


likewise for guardian's path 1 - .15 = 0.85
0.828*0.85 = .7038
1 - .7038 = 0.2962 rounds to 29.6% which is what we see in the raw data.


add in the mantra, another 0.85:
.7038*0.85 = 0.59823
1 - 0.59823 = 0.40177 rounds to 40.2% dodge again what we see in the data.


etc.

[ElementEight]

Isn't an easier way of doing this just going by adding the % to whichever bracket you already fulfil?

Ie, I gain 15% from mantra, 15% from guardian's:

15 + (15*(1-.15)) = 27.75

Then, adding another 12% from whatever source would go by

27.75 + (12*(1-.2775)) = 36.42%

That's probably what you did, but iunno, my way feels simple.

[magicrectangle]

What you wrote is equivalent to what I wrote, I just used shinier notation.

[hubb]

Cool. So in essence each dodge check is done separately until one works (or they all fail). Like CS/DS in D2.

[oakienko]

on a parallel topic, do you guys thinking blinding flash 30% chance to miss goes before or after the dodge check?
in other words, does the monster swing and miss you because he's blinded or does the monster swing and miss you because you're dodgy AND he's blinded? :P

[Coriolis]

Why would that matter Oak? You're effectively multiplying the two chances, so it makes no difference which is first.

[magicrectangle]

Why would that matter Oak? You're effectively multiplying the two chances, so it makes no difference which is first.

From the point of view of your damage taken, it makes no difference, but if chance to miss comes first it means more misses and less dodges. If it comes second it means more dodges and less misses. I don't know that that matters for anything at the moment, but it isn't inconceivable that blizzard could introduce effects that proc when you dodge an attack.

Figuring out which is the case would be annoying to do without access to the backend. You'd have to collect a large tally of how many misses, hits, and dodges while blind was up, then check the miss rate. If the miss rate were lower than 30% by a statistically significant amount, one could assume that dodge was calculated first.

[Coriolis]

Actually there is already the rune effect backlash on matra of evasion that depends on dodging. But I don't recall ever seeing a "miss" so I assumed that the effect from blind is just another form of dodge. I suppose you could use backlash to test it if you really wanted to.