Cataclysm Mechanics Testing - Page 6

• Aldriana · 10/16/10, 11:21 PM

Whoops, missed that. So either a) the dodge rate is somewhere around 5.6% (if it were 5.6% you'd expect one miss per 13569 attacks at 172 expertise, making that data set unsurprising), or b) The dodge rate is about 6.5% but expertise converts about 26.5 points to the percent as opposed to the 30.75 its supposed to. Of those options the first seems rather more likely, but either way it's a bit odd. Have we confirmed whether the same effect occurs on beta?

Crtlaltdel · 10/16/10, 11:43 PM

Very crude test! Only 1999 auto attacks. Was going to do more, but I got called away to do CoD4 stuffs. This was done with 0 expertise as you can see. Dodge rate doesn't add up to the tooltip. Figured I would post this in response to the discussion that is now going on! Yes my character is not wearing pants/chest/wrists...>_> don't judge me.

Rallik · 10/16/10, 11:44 PM

It may even be slightly under 5.6%, which is very strange since at least 5.6% has some significance. My test and Kortwa's seen here: Cataclysm Warrior Changes had a total of 62,026 swings without a dodge. The chances of that occurring if the dodge rate were 5.6% is only 1.03%. I suppose it's also possible that the dodge rate is 5.6% and expertise is slightly more effective than it should be. If it took only 30.715 expertise per 1% instead of 30.755, that would be enough to cap it at 172 instead of being under.

As for the possibility that expertise rating is much more effective than before, I ran a test at 0 expertise to get a ballpark for dodges without any expertise involved. The confidence interval of the results puts 6.5% outside the range even at 99.95% confidence, so it's fairly certain 6.5% doesn't hold up. Also, the 42,000 swing test had a parry rate of 8.4%, which is consistent with only a 5.6% reduction in parries from expertise.

Edit: Adding Vladz's numbers to my own puts the dodge chance being 6.5% outside the range with 99.998% confidence.

Docrev · 10/16/10, 11:47 PM

I can confirm Crtlaltdel's results, 1000 auto-attacks and only 47 dodges, not the 65 expected. I'll let it keep going until my off-hand dagger turns red, though.

Glokki · 10/17/10, 12:00 AM

I ran with 25/22 Expertise (175 rating) on Thursday night.

Obviously the focus would be on whether my OH was ever dodged by a boss. Here is the log of all melee swings by Glokki vs Bosses for the entire night:

World of Logs - Real Time Raid Analysis

Despite a relatively small sample size you will notice not a single dodge by a Boss.

I understand that 172 rating is the supposed 'magic number' (not 175) but believe this log is still useful for the purposes of this discussion.

Rallik · 10/17/10, 1:01 AM

I just checked on beta to see if this carried over. On a level 85 against the Raider's Dummy with 671 expertise rating(~5.587%), I had 3 dodges already at 307 swings and stopped testing since that made it clear the chance wasn't 5.6%.

Docrev · 10/17/10, 1:24 AM

Here's some additional data from auto-attacks on the raid dummy in Thunder Bluff:

darkmisfit · 10/17/10, 4:19 AM

Originally Posted by Aldriana

....Have we confirmed whether the same effect occurs on beta?

I know about... 2 builds ago I noticed that I wasn't doging near as often as I should have been. After a lot of reforging, etc.. I was able to narrow it down to "bugged" at 3.25%. Right about the time this thread started I set out to test it again but with the new build to the Beta and on the PTR things "seemed" to have leveled off at 6.5%, so I didnt push forward with any serious testing and just chalked it off as an unintentional bug that got fixed.

Screen shot of the bug at the time:

-
WoL parses: Test 1 - 1.4 speed - Dalaran Axe - 90 Minutes -- 1.4 Speed - Dalaran Axe - ~120 Minutes

Both are from 9/23 - 9/24 so they are a bit old..

During raid tomorrow I'll idle in SW and run a 3-4 hour parse to get some solid numbers.

Ramathor · 10/17/10, 5:30 AM

In an attempt to further study the mysteries revolving around the way Expertise is behaving at this time, I sat on a level 80 dummy at 0 Expertise Skill in order to find out whether or not the chance to be dodged facing a level 80 target has changed at all. My findings were that it has not.

From this there are two possibilities that I can think of for what this means to our discussion.

The jump from a level 80 target to an 83 target is only a .5% jump, roughly .16 per level.

or

The current test dummies are set to reflect a level 81 instead of 83.

At this point, our best course of action is to confirm the cap to be 172 on an actual boss to set the theory in stone. I will continue testing on the level 80 test dummy to further validate my findings.

EDIT: After further testing, my previous findings appear to be correct

• Aldriana · 10/17/10, 11:42 AM

Given that testing puts the actual expertise value much closer to 5.6 than 5.5, I think we can rule out the "level 81 target" option in favor of the "only .2% per level" option.

Vaelaris · 10/17/10, 12:25 PM

I think all of the testing has absolutely confirmed that, at this time, level 83 dodge rate is only 5.5%. However, this does not change the fact that the Blizzard tooltip on Expertise still lists it at 6.5% for an 83/skull. Because of this, we have two options. Either Blizzard has actually changed the dodge rate to take into consideration increased dodge rates by 84-88 mobs and never updated the tooltip, or there is a bug in game that is accidentally decreasing the dodge rate. I would have to recommend a bug post on the official forums and see if we can't get some confirmation.

Edit:
Set up a bug report on the World of Warcraft Bug Reporting Forum. Feel free to add in any additional information if necessary.
World of Warcraft - English (NA) Forums -> [Bug?] Level 83/Skull Dodge Rate

Fyll · 10/17/10, 6:17 PM

I know we don't rely on dummie testing anyway. But I can confirm another bug of the dummie, might be already known. Just in case I post it short. If you shadowstep while a subt rotation to the dummie, you start to parry. Makes actual long time testing of the current talent variations basically impossible.

I only worry if shadowstep might also cause parry by a boss.

Armanewb · 10/17/10, 8:47 PM

Originally Posted by Fyll

I know we don't rely on dummie testing anyway. But I can confirm another bug of the dummie, might be already known. Just in case I post it short. If you shadowstep while a subt rotation to the dummie, you start to parry. Makes actual long time testing of the current talent variations basically impossible.

I only worry if shadowstep might also cause parry by a boss.

Shadowstep works in the same way as Killing Spree, in that it puts you so far inside the boss's hitbox that you can see parries. The solution is just to back up as soon as you use Shadowstep.

Rallik · 10/18/10, 3:24 PM

Originally Posted by Aldriana

How much armor do bosses have?

Assuming the Raider's Target Dummy continues to use boss armor levels and that armor must have an integer value, bosses have exactly 11977 armor.

I tested using Bloodthirst as fury spec(110% damage), in berserker stance(110% damage), using Glyph of Bloodthirst(110% damage), making the damage formula: .75*AP*1.1*1.1*1.1*Armor. Keeping in mind that WoW will only round a non-integer damage result after calculations to either of the neighboring integers(i.e. 5231.4 to either 5231 or 5232), I searched for AP values that provided border cases where a difference in 1 armor would result in a different potential damage range to set the upper and lower bounds.

First test at 8111 AP:
Damage at 11977 armor would be: 8111*.75*1.1*1.1*1.1*26070/(26070+11977) = 5547.97, meaning it can only deal 5547 or 5548 damage.
Damage at 11976 armor would be: 8111*.75*1.1*1.1*1.1*26070/(26070+11976) = 5548.12, meaning it can only deal 5548 or 5549 damage.

After 18 BTs at 5548 damage, I observed a 5547 damage hit. This sets the lower bound on armor to 11977 since any armor value lower than 11977 would have guaranteed a hit of at least 5548 damage.

Second test at 8693 AP:
Damage at 11977 armor would be: 5946.06
Damage at 11978 armor would be: 5945.91

After 13 BTs at 5946 damage, I observed a 5947 damage hit. This sets the upper bound on armor to 11977 since any armor value greater than 11977 would have guaranteed a hit no higher than 5946 damage.

• Aldriana · 10/18/10, 4:28 PM

Couple of additional points on armor.

Lets drop the the assumption that armor has to be a whole number for a moment and look at the actual range defined by that testing, according to the formula I derived earlier in this thread (A = P*(D-D')/D'). The first test tells us that armor has to be between 11976.81 and 11983.67; the second narrows it down to between 11971.01 and 11977.41. Hence, in combination, we know the answer has to be between 11976.81 and 11977.41; hence, even if you posit that it can be a half-integer (as the armor parameter can be) 11977 is really the only option.

Secondly, if I may quote a PM I received a few days ago:

(post-3.1) WotLK : npc.base=305*npc.lvl-14672
Beast Lore on 4.0 live : 10207@82, 10502@83
4.0 possible fit : npc.base=295*npc.lvl-13983
@88 estimate : 11977, matching well the value proposed by Aldriana

Of course, we still need additional data sets in order to verify this scaling pattern.

Hence, there is some external reason to believe that the answer is 11977. Thus, I think we can be really pretty confident that this is the exact value, and I have updated the first post accordingly.

If anyone is feeling truly insane, it might be worth verifying whether armor must be a whole number or not, particularly after debuff application. That is: with sunder/expose/FF up, 11977 armor would theoretically be reduced to 10539.76; does the game actually compute damage based on this number, or does it get rounded or truncated to 10539 or 10540? Testing this will require finding appropriate "critical values" as Rallik did, plus quite a bit of patience - but it should be possible to tell the difference.