Leveret

Even against a +3 mob, there should be a 13% difference between the expected crit chances under the two possibilities. There isn't even evidence for +3 mobs?

 

Roywyn

Stuff that has been done by people:

Loatheb - the encounters gives you a +60% spell crit buff. People got to the point where miss rate plus crit rate were greater than 100% and still got normal spell hits. So, it can't be a one roll.
I have only read that, and do not have any first hand evidence of it.


Stuff that I have tested:

Cloak of Shadows - wowhead/thottbot listed it as -90% spell hit chance. I did a sloppy test with ~23% crit, 75% miss chance, and molten armour versus a SnD hasted dagger rogue. We had around 100 molten armour casts, with ~30 non-resists, 8 of which were crits.
It was just a quick test, the setup of numbers wasn't planned, and I don't know how CloS works exactly on the mechanics.
Still, it's a pointer that if it works how I think, it's an indicator for a 2-roll.

With 40% crit and 10% miss, you have 40% crit in a 1-roll and 36% crit in a 2-roll system. There is no level 73 Blasted Lands/Dr. Boom mob for easy testing, and in raid, people usually change gear/buffs/debuff/groups for different +hit in different encounters.
So, even if you manage to separate level 73 mobs, you can't really take a raid log and compare.

 

maxi

Well, the thing that people are already putting hundreds of hours in is raiding.

If someone could convince their raidleader to allow them to raid without nay +hit whatsoever for testing purposes and record hit/miss/crit data for every boss encounter, it could be interesting.

By the way, if i will try to test it myself, is there any good addon that records and counts hits misses and crits?

 

sp00n

/combatlog is the best. Of course you'll need a parser for this, WWS could suffice.
Or Recount, but then you have the data only ingame.

 

SeanDamnit

My mage is currently level 60, and if you have some patience I can set him up to do a Dr. Boom test to provide a ton of data at level 65.

I'm not too good with the math stuff, but I am interested in helping out the community. Once I hit 65 I'll put a good amount of time on him with /combatlog running and I'll post the data here.

How much data do you all think you'll need?

 

maxi

well anything less than a thousand casts would not really be sufficient -.-

So brace yourself for about an hour of scorch spamming.

 

PSGarak

I've posted a testing suggestion for this over in the 2.3 PTR sign-up thread. It sounds like someone also tested against CloS, and saw a 2-roll version for that, but I would not at all be surprised to find that CloS has it's own roll outside of the caster's attack roll, being a target buff instead of a caster debuff.

I would suggest going at Attumen with a ~30% crit rate, and a 60% crit rate. The first one will tell you whether it's a one-roll or two-roll system under a relatively small data set (70% vs 30% effective crit rate is easier to distinguish than 30 vs 26, not to mention potential level-based crit depression). The second one will tell asymptotic behavior on minimum normal-hit chance. BTW both tests can be performed simultaneously, since kara is a 10-man raid. You can have 3 versions of each going on, you should get a decent-size sample set pretty easily.

 

• Joanna

I'm unsure if there's some glaring reason why this wouldn't work, but as I envision it should provide information on the possible scaling down of personal crit rate vs higher level mobs, and also some insight into whether or not a one or two roll system is in play.

I suggest you simply find a warlock and a group of willing summoners to bring lots of relatively low level mages to the Hellfire side of the Dark Portal. I'm sure everyone recalls the endless waves of high level demons besieging the rampart, and picking a specific type and level (for instance only level 69 felguard type) of mob to attack, one could essentially (time and boredom permitting) gather endless amount of information regarding how spells function against higher level targets.

The Dark Portal strikes me as the perfect location to test this since you are effectively guaranteed an endless supply of a specific mob with no danger of pulling aggro (the guards will quickly pick up anything that level based aggro attracts) and so you can refine your testing procedure to the point where you can vary only those factors you choose - I for instance would be interested to see if a level 69 and a level 70 mage, with equal hit and crit rates (from the player stat sheets, since they're all we have to go on) achieve remotely similar actual results when attacking a level 69 mob.

Apologies for a slightly meandering post, this is my first here, but it's a topic that holds some interest for me, and most of the problems people seem to visualize involve the simple practicality of having a low level player be able to cast sufficient spells against a high level target in the first place. As I say, let me know if I've overlooked something fundamental.

Nulla in Mundo pax sincera.

 

PSGarak

Well, you're testing two variables at once: 2-roll system, and crit depression vs high-level targets. We know that one of them occurs, since we see low-level targets hit-not-crit vs high-level opponents, but we don't know which one, and the sample size to test relative difference with any certainty is prohibitive. We need testing against an even-level or boss-level mob with a hit chance debuff to test the one-roll/two-roll system, and then regardless of the outcome we can use that to test lowered crit rate against high-level targets.

"(there is) no sincere peace in the world"?

 

• Joanna

The singature is from a Vivaldi motet, don't ask, I was stuck for something to put there and it was on in the background.

I take your points that my rather simplistic proposal was flawed in terms of providing a straightforward answer to the question in hand, but going back to Chase's point from earlier, it should still provide some insight. If you bring a sufficiently low level caster with an appreciable crit rate to the Hellfire side of the dark portal, with 0 Hit rating, no points in elemental precision and a known crit rating, you could answer fairly quickly whether a crit can miss, or through some further mechanic be 'downgraded' to a hit.

I'm not sufficiently familiar with how hit chance vs higher levels scales past a difference of 7 levels or so, but I'm working on the assumption that someone somewhere has an idea of what level difference you need to make the chance for a spell to hit lower than the crit rate of the test subject. Supposing, just for illustration, as stated I have no basis for the actual numbers I'm choosing, a level 55 character has a 10% chance to hit a level 69 mob, but a 20% crit rate.

In a one roll system with no unknown further modifiers, I'd expect 20 of his spells to crit and 80 to be resisted completely.

In a two roll system, I'd expect to see 2 crits, 8 normal hits, and 90 full resists.

In a one roll system where crits are downranked by some unknown mechanic, I'd be downright amazed if Blizzard had matched the degree of downranking to be similar to the two roll result - if they have it makes this moot and the whole thing behaves like a two roll system regardless, but if they haven't I'd expect to see a spread of hits, crits and resists (with sufficient data) totally other.

As a small addendum, I'm truly awful at statistics and mathematical puzzles in general, my posts was initially intended simply to suggest a suitable venue, since the mechanics of this type of test seem to be the biggest stumbling block. Again, sorry if I'm overlooking something, or 'just don't get it,' (which is entirely possible)!

Nulla in Mundo pax sincera.

 

◊ Turik

GMs do not have privy to that sort of information.

 

Chardonnay

I was thinking it's a single roll system, but now I recall the weeks before BC when we farmed lvl70 normal mobs around Karazhan as lvl60...
On a mob 10 levels above my level I have 94% miss chance if I'm not mistaken. I specced 51 fire for fun. My gear included 10% hit (84% miss) and crit around 15-20%. Most of the time I wounded the mob it was a crit. Absolutely most of the time. But still I remember landing normal hits, most of them being partials. Probably this is a 2 roll system in some way.

Testing a 60ish mage against the mobs at the Dark Portal is a great idea.

 

chase

Where did the idea of "crit depression" come from? We know vs mobs +3 (bosses) that are crit rates are not devalued/deflated. Why should we assume that at +4 levels or greater there comes in another variable?

I'm not saying it isn't possible, just that I have seen nothing to date to suggest such a mechanic exists. The whole if it walks like a duck and quaks like a duck it isn't a zebra thing.

 

Gaborn

One thing that leads me to believe it's a 2 roll system is the way crits on partial resists work.

When you see something like this in your combatlog:

"Your Shadowbolt crits X for 2250 (500 resisted)"

This example shows that game first calculates if the spell is resisted in some way or not, and then uses the remain damage (if it's not resisted completly) to calculate the crit damage.

I know this doesn't proof anything beyond the fact that crits on partial resist work this way, but, this mechanism is easier to implement using a 2 roll system than a 1 roll system.

 

chase

Except from the blue post:
Not saying I trust blue posts any more than GM's, but with this kind of detail it suggests to me that Eyonix talked with someone who knows the roll system.

So while I agree that misses are determined first, a miss and full resist are different.

Since partials resists (full resist is simply 100% "partially" resisted) are calculated at impact (for DD spells) I don't think you can draw any inferance to the roll system from this.

WoW BlueTracker: +Spell Hit Chance

 

PSGarak

A tank needs 490, not 475, defense skill in order to become uncrittable because the additional 3 levels affect the mob's crit chance. Obviously, spells work way different than melee attacks; most of the damage reduction comes from glancing blows, not outright misses. Nonetheless, it's enough of a suspicion that we really should test it than leaving it to chance.

We actually don't know that +3's have no crit depression. We just assume they don't and we've never tested.

Chardonnay's story would lead me to suspect a one-roll system: a two-roll system would lead to around a 40-50% crit rate on attacks that landed. A 2-roll system is far more likely to have a higher-than-normal amount of your unresisted attacks crit, even if not all of them do. The numbers have to juggle just right, but it's still more likely imo.
Of course, "SS or it didn't happen" (or combat log parses in this case) and all that jazz. Enough suspicion to elicit a test, not enough evidence to prove anything.

 

Mageborn

Still no conclusive results?

I think its worth pointing out that although some people describe your crit chance increasing with one system over another, the question really is:
Given 25% crit chance and 10% miss rate;
Do 25% of your spells that land crit?
or
Do 25% of your total casts crit?

I'm fairly sure the answer is that 25% of your spells that land will crit, as I believe thats how SWstats measures your crit rate, and my crit rate on SW stats has always matched up closely with the crit rate that my character panel (plus talents) reports. Most of those results are from boss fights (eg nightbane/prince/gruul) where I had 8%+hit or less.

To clarify, the two roll system I envision would work as follows:
Given 25% crit and 10% miss rate:
Roll between 0 and 1, if less than 0.90, the spell hits.
If the spell hit, roll again between 0 and 1, if less than 0.75 then the spell crits.

The implementation I show here has 2 rolls, is very simple and intuitive, and the number of crits you get on unresisted attacks does not change from your crit rate.

Conclusive results from someone would be really great, and I'm surprised there aren't any :/

 

vokzhen

EDIT: Nevermind. Misreading stuff sucks.

 

alvinrod

Is that even a safe assumption to make?

We don't know if that trend continues or if at some point it changes or stops completely.

I think the best way to test it out would be to grab a mage that's around level 50 or so and escort him out to Dr. Boom. Give him all the good fire talents and a moonkin and he'll have a decent crit rate. Could use two different mages at the same level with the only difference being talents to affect the crit rate.

Just keep the mobs off of them and have them spam fireballs for a while. After about 1k casts from each we might have a better understanding of whether or not there's a 2-roll system or the % of spell hit you'd need to compensate for something with that wide of a gap.

 

vokzhen

Important thing is, you're running into the possibility of crit depression there. Tests MUST be done on equal (or very close) level mobs in order to make certain any supposed crit depression doesn't take place. In which case, you're running with ~5% miss rating, and will need thousands of samples (if not more) in order to make absolutely sure getting a few more crits than average doesn't skew the whole results.

On a slightly unrelated note... partial resists on crits do show up as crits still, right? They don't on SCT, which had me confused until I figured out what was going on, but I think (hope) that's a SCT-only problem.

 

Wuff

For now, crit depression is simlply an interesting assumption to invalidate the tests of low level casters with high crit against much higher targets. Right?
I think it would be much simpler to prove that there is no crit depression. I'm no expert in mages, but I think it could be possible for a sub 60 mage to reach 100% chance to crit against frozen targets. This way you would need a second mage to freeze a 70 mob while the lower mage spams frostbolts. If you ever see a non-crit I will believe crit depression exists.

 

mylek

This discussion has come up in the past. In summary, extreme increases in the chance for your spells to miss a mob or player do not seem to influence the rate of critical hits vs regular hits as expected of one roll. I realize that the data is far from what I would consider definitive proof but a tentative consensus on two rolls was reached. I have yet to see any supporting evidence of a one roll system for spells. (One roll meaning linked hit and crit chances.)

Spell Hit, Spell Crit and mages
Backstab: Two Rolls?

 

Soul

I'm pretty sure there is crit depression versus higher level targets, the question is what portions of spell crit are affected by level differences? Spell crit% comes from three sources: Int, crit rating and percentage crit modifiers such as talents and auras. We know crit % from int scales with target level, but we don't really know how the other portions of the crit calculation behave with regard to level differences.

My suspicion is that crit chance is calculated based on the target level (as opposed to the caster level) for int and crit rating and that the flat modifiers are added on at the end. That is, if you're a caster throwing spells at a level 70 target, then you need 22.1 crit rating or ~80 int to gain 1% crit against that target regardless of your level. However, this hasn't been formally tested yet, so maybe it's something that should be tested before we see if there's a one or two roll system.

And it's not too hard to test... just get two characters, one at level 66, one at 68 in identical gear with identical specs and a 1% miss rate versus level 68 mobs to chuck spells at Dr. Boom and see if the crit rate is any different across characters and different spell schools. A good test would be to use deep Fire mages with 2/2 Incinerate, 3/3 Critical Mass and 3/3 Pyromaniac focus to spam AM then Scorch at Dr. Boom and see how the crit rates differ.

 

SeanDamnit

Ok, my mage is now 65, and I'm prepared to venture over to Dr. Boom to provide as much data as my favorite albums on ITunes can sustain me through. My question is whether or not anyone still feels that this testing will be useful in working towards proving/disproving the one roll/two roll theories and crit depression. Seems like most thoughts on the best way to test this subject are kind of scattered, so I just want to make sure that if I spend time falling asleep on my Scortch button that someone will get useful info from.

I'm here now ready to be your instrument - just let me know what you would like me to do, and any details about spec or methods that you would like me to use.

Currently I am spec'd fire, and have 31.88% chance to crit with scorch (139 crit rating, molten armor, 342 intel self buffed). 0 hit from gear, and no +hit talents.

My character: The Armory



edit - note to self, don't test stuff on a PVP realm.

 

SeanDamnit

Here is 1000 rank 1 scorches Dangergnome - WWS

Crit is lower than expected. On the paper doll, I have 25.88% crit for fire spells. The scorch talent should make that 29.88%, and for some reason the addon Dr. Dmg puts my scorch crit chance at 31.88%. Anyway, as you can see here crit ended up being 23%.

I plan on getting more data, I just don't have time right now. I will post again when I gather more data.