Soul

Don't forget that there is crit reduction versus bosses. I don't know if anyone ever figured out what the reduction was or the nature of it, but it's definitely there.

 

Pheroz

Please support this statement. I have never heard, seen or encountered anything like this and I would like clarification.

 

Shurik

You should check your WWS logs and have a look at your average crit rate and compare it to the one expected. It should be ~5% lower.

Success isn't a result of spontaneous combustion. You must set yourself on fire.
- Arnold H. Glasow

 

galzohar

I don't remember anyone actually managing to figure out the crit reduction VS bosses, if it even exists - though I wouldn't be surprised if it existed just like bosses have more parry/dodge/to be missed due to their higher level.

Regarding "RNG" and "...but you're assuming ideal conditions where you actually get your crit rate" - what do you suggest we do instead? Should we assume we get 0% crit for worst case? Should we assume we get <average crit> - <crit standard deviation>? I mean no matter how you look at it, it just doesn't make sense to do anything other than taking the average DPS and to a lesser degree its standard deviation.

Saying you seem to be always on the lower end of the RNG is rather silly, since we all know RNG is always random and if you got on the low end you probably just looked at the time(s) when you got unlucky and ignored the ones where you got lucky (or just didn't play enough to get lucky). Also keep in mind that while hot streak specs are sensitive to RNG (compared to what other classes / pre-WOTLK specs were), arcane specs and frost specs are not very different - arcane has MBarr procs which make a significant DPS difference and have not far from 50% chance per rotation to occur, and even frost has FoF procs. If you want theorycrafting to take RNG into account, you have to say how you want it to be taken into account.

In the end this game has very little RNG that you can do something about other than just taking averages.


Regarding interrupt modeling, I think it would be better to calculate it with a random DPS window that has, say, a standard distribution around some value or a uniform distribution with a set range of possible timeframes. If you just set the DPS window to a fixed number you're definitely going to get weird results with 1 haste rating being worth a lot when you get that extra cast in and then haste rating worth 0 when you don't - in reality cast windows are not at all constant - even a 0.2s variation from minimum human reaction time is enough to make the "fixed time cast window" calculations rather pointless.

 

◊ manly

If crit reduction existed, then explain me why a fully naked mage will still crit 2% of the time on a skull level mob.

Log on with different model:
1- Create a character of the desired model. Log on/off.
2- At character selection screen, select your actual character; mouseover the new, desired model character, and hold down left click; hit enter and release left click at the same time.
bug Arcane Potency only applies to the first Arcane Missile bolt.

 

Soul

We found out that there was crit reduction back in TBC when we were trying to figure out if there was a one or two roll system for spells. We (well, Seandammit) threw thousands of Arcane Missiles at Dr. Boom with a level 64 Arcane Mage with and without Arcane Focus and found that regardless of whether there was a one or a two roll system, it was at least 99.5% certain that the paper doll crit level was significantly higher than the actual crit level on that boss.

We don't know the nature of the crit reduction, we just know that it is there. For example, if it was a scaling factor applied to INT and crit rating, then your base crit would remain the same, but the amount of crit added per point of rating/INT would be smaller, giving rise to the observed drop in crit rate.

I haven't done testing on this sort of thing because with the change to downranking, this sort of testing is not really practical any more. We needed thousands of tests to find out with that kind of certainty that crit depression existed and we were only 90% sure the casting system is a two-roll system. You need tens of thousands to get a feel for the nature of crit scaling with level and good old Arcane Missiles simply won't cut it for testing any more.

 

Enthorn

(Link) Kavan had an untalented level 80 mage with no gear on and crit rate showing 2.07%. After 30 casts, he scored a crit. This was to disprove other findings stating that there is a 2.1% crit reduction, which was supposedly found via this test from that link:

Meanwhile I got a crit wearing 2.16 after 1200 casts and now running 2.05 for 2100 casts without a crit and counting.

These two findings are obviously contradictory. How do you go 2100 casts without a crit at 2.05%, and then get a crit in 30 casts with 2.07%? And if the reduction is 2.1%, that would make sense why it took ~1200 casts to get a crit (with a supposed .06% crit rate).

But then equally so, I find it interesting that, even if Kavan's reported crit rate of 2.07 was off by even a bit, it's still highly unusual to get a crit after only 30 casts. Furthermore, I've been told by others that there is a base crit reduction of 0.4% per level above you. Thus, a boss level target would have a 1.2% crit reduction versus you.

I'm not sure what or who to believe (and not entirely sure it matters).

 

Zaldinar

If the crit reduction is less than 2.07% and greater than 2.05% this makes perfect sense?

Just because something has a 0.01% chance of happening, doesn't mean it wont happen within 30 attempts. How many people got Barons mount on their second time trying?


Edit: To establish a new lower limit we've seen a crit at, here's a 1.99% crit chance crit on the Darnassus boss dummy.

http://i142.photobucket.com/albums/r...treduction.jpg

Trinkets were [Foresight's Anticipation] and [Spark of Life] to make the test go faster. Specced Talent Calculator - World of Warcraft wearing mage armor.

To truly model the game, we first must research it.
http://forums.worldofwarcraft.com/th...icId=109841969
Proven TheoryCrafting Stuff, chain casting in a PTR near you soon.

 

Northerner

Well, Kavan continued on to say:That doesn't preclude a simple streak in probability but it makes it improbable. I'd guess that there is some testing condition being missed here but I really don't see much point in investing a lot of time into it since there is no meaningful impact on our playstyle or gear selections.

 

Silanre

I agree crit is looking more valuable then before point for point compared to haste but its all in the algebra.
RAWR is a fabulous tool and being tweaked atm, but it is tending to bump up crits value but keeping it close to hastes value and that is assuming a more static fight.


Crit reduction...
If crit were scaled back slightly with your gear that might explain the lower end differences(naked gnome example) but also why wearing 200-226 item level gear might still vary the average reduction from one player to another. I doubt it would be worth the time to analyze enough WWS to find the exact formula, but someone might make a good educated guess based on trying 3-5 crit tiers of gear and plotting the changes.

Hidden formula might be similar to: Crit = Total Crit value - [2% + 1%(caster's crit value)]
(not suggesting these numbers are even close)

Lots of variables might make this difficult if say for example it scales to you only versus other individual raid synergies.
The simplest thing might be the master value of all buffs consumables and even timers reduced by a fraction of a percent or to the player's base crit value.

There may also be hidden rules such as 1 out of every 1000 or 10,000 etc will always crit. As much as I enjoy looking at the math and how things work I don't envision Blizzard wanting everything to be too clearly calculated whether it is intentional or accidental on their part.

An interesting mystery nonetheless.

 

Zerstorung

I never said that dps needs to be calculated differently, the way it's calculated now seems the most logical. However, I feel like there may be some weighting factor when looking at crit vs haste due to hastes consistency.

 

Enthorn

Oh, don't get me wrong -- I understand this. I'm just looking for some clarification as to whether the crit reduction chance is around 2.1%, 1.2%, or if it exists at all. I'm merely looking for the data.

 

galzohar

The difference in the inconsistency between a mage with X% crit and Y% crit, assuming both of similar item level gear, is extremely small. In fact, increasing crit above 50% actually reduces inconsistency, as 50% is the most random.

 

Zaldinar

That's pretty easy to isolate as far down as is possible without violating the terms of the experiment... What race has the lowest int at 80? Should be a Troll right? ( WoW.com -> Info -> Classes -> Mage )

Do we have a volunteer naked troll in the house?

To truly model the game, we first must research it.
http://forums.worldofwarcraft.com/th...icId=109841969
Proven TheoryCrafting Stuff, chain casting in a PTR near you soon.

 

Soul

Setup and Procedure

So, just to verify that the whole crit reduction business isn't a figment of my imagination, I headed off to the trainer to pull out a spec without Arcane Potency.



Arcane Missiles only is affected by global crit talents, so the paper doll crit, which, buffed with AI and Mage Armor, is 16.80% for my mage in his normal gear (I don't raid much) would be the expected crit of Arcane Missiles.



I then proceeded to spam arcane missiles at the Heroic dummy until I got tired of it. I tossed some Arcane Blasts in for good measure, but they don't count because I was dumb and got Incineration. Whoops.

Results

Results of Run 1 are as follows:



Shortly after this, WoW crashed and took Recount with it, so I did another run. Results of Run 2 are as follows (the first image demonstrates that this wasn't a mere continuation of Run 1):




Analysis

We assume a two-roll system for spell casts. If crit suppression on higher level mobs is demonstrated assuming a two roll system, then it exists if casting was done on a one-roll table like melee. The number of critical hits for a number of landed spells is expected to follow a binomial distribution with the paper doll crit rate as the probability statistic, . For large (defined as ), can be approximated to follow a Normal distribution with mean and variance .

The expected crit rate is 16.8%.

In Run 1, 2491 missiles landed and 367 crit. It is expected that 418.488 missiles would crit, with a variance of 348.18. The difference is 51.5, or about 2.75 standard deviations from what was expected. So after Run 1, we are a little more than 97% confident that the paper doll crit rate is higher than the actual crit rate on a boss mob.

In Run 2, 3704 missiles landed and 554 crit. It is expected that 622.3 missiles would crit, with a variance of 518. The difference is 68.3 or 3 standard deviations. So, after Run 2, we are more than 98.5% confident that the paper doll crit rate is higher than the actual crit rate.

However, Runs 1 and 2 are independent of each other, as can be seen from the sets of screenshots. So, the odds that both runs 1 and 2 would occur if the paper doll crit rate were indeed the actual crit rate on a boss mob is less than 0.045%.

In other words, after these two runs, I can be more than 99.9% sure that the paper doll crit rate is higher than the crit rate you will actually see on bosses (at least for arcane missiles).

 

Mynak

I'm asking this as a legitimate question, what impact on theorycrafting does the existence of crit suppression have? Will crit become a less/more desirable stat?

 

Silanre

The hidden crit reduction assuming it exists (and is even measurable) is largerly if not entirely insignificant.

The larger issue for crit atm is how will its value shift and by how much in 3.1 relative to spellpower and haste especially.

 

Zaldinar

Soul, here's some interesting number fiddling...

Looking at your data, 2125 Hits + 363 Crits + 102 Misses = 2590 events. Observed crit rate of non-misses of 363/(2125+363) = 0.1459, vs your expected of 16.8. Observed crit rate of the whole shebang of 363 / 2590 =
0.1401

Now, let's take a walk on the wild side. Looking at your hit rating, 237/26.23 + 3 = 12.0354 Without Precision + 3 = 15.0354 with (Was it with precision?). Take your expected crit rate of 16.8, multiply it by your hit rate potentials, and you get 16.8 * (1 - 0.120354) = 14.7780 or 16.8 * (1 - 0.150354) - 14.2740

If we apply the same to my earlier thing of the 0.0199 crit rate, I would have an expected crit rate of 0.0199 * 0.87 = .017373.

Do you have a setup that hit caps you on the nose? I, too, will be conducting this test in a little bit, but I want independent confirmation.


Edit: Curiously, this is what I have so far...

Next test will be a level 80 dummy, then a few varied hit rates and talents to see if anything shows as a trend. But I'll confirm that we're seeing a 2%ish reduction from expected, at hit cap, against a level 83 target.


It will shift relative values a little. FFB mages will get more from that 2% crit than Arcane mages, so it changes it slightly in their favor, etc etc. Mostly thumb-sucking though.

To truly model the game, we first must research it.
http://forums.worldofwarcraft.com/th...icId=109841969
Proven TheoryCrafting Stuff, chain casting in a PTR near you soon.

 

Maje

According to rawr, crit's value for ttw/fire build is slightly above haste, both are well below spellpower. Looking at Scaling vs. Spellpower graph rawr provides shows an upward trend of both haste and crit relative to spellpower such that level of around 4.1k spellpower haste = crit = spellpower with haste scaling a bit better.

I'm guessing the culprit behind crit being better then haste is the new LB glyph which makes LB benefit from crit and almost no benefit from haste.

 

galzohar

Crit reduction changes theorycrafting just like putting in your new item in rawr changes theorycrafting. Every small change of stats has a very small change on the value of additional stats. Some people go as far as using someone else's gear setup to get their stat equivalences, and some will go as far to the opposite direction and make sure their stat equivalences are based on their exact current gear setup for maximum accuracy.

How much does it really matter? Very little, but if you want to be 100% accurate in calculating the value of additional stats you need to know your exact current stats.

 

Soul

No, I don't have a set that hit caps me, hence my rather unusual spec (I run heroics, so want 5/5 stability and 40% aggro reduction to deal with PUG tanks). Regardless of how hit and crit interact (one roll system, two roll system), we are still seeing crit reduction just as we were back in the TBC days.

As for your numbers:

You would expect to have 992.4 crits, with a variance of 832.0. You have 863, which is 4.48 standard deviations away from the expected value. This implies that the paper doll crit value is higher than the actual crit value you see on bosses with a 99.999% probability.

I'm not making any guesses as to the nature of this crit depression yet. For one, our crit rates are very low - around 16-17%. A FFB mage would have a much higher crit rate and therefore, might see a differing amount of crit depression. It would be interesting to see what a full on scorch crit build with near BiS FFB gear would see spamming scorch on a Heroic target dummy.

As for how it would change how we value crit:

If the amount that crit is lowered is fixed, then the value of crit goes up because the less crit you have, the more valuable more of it is. On the other hand, if the crit reduction is due to the way crit rating applies to your chance of a critical strike versus a boss, then the value of crit rating is lowered. Of course, it could be some combination of both, which could make it a wash.

 

Mynak

I know that this test will be pretty hard to run but using this Arc/Frost spec Talent Calculator - World of Warcraft your Frostbolt would have 100% crit when you have a Totem of Wrath, Focus Magic, Winter's Chill, Arcane Potency, and Shatter available. If you have another mage that's willing to help you out, he could Frost Nova the target dummy when you have Clearcasting up to guarantee Arcane Potency/Shatter. If someone had a non-crit Frostbolt when they should have this 100% crit chance then we'd know that there is a crit suppression.
On the other hand crit suppression might not effect talents and raid buffs and could be a modifier of Critical Strike rating making this test invalid.

 

Dazl

I'm a new poster and relatively new lurker. I've read every post in this thread, I think, over the past several days. All very interesting and I'm slowly coming to grips with the math. Thanks, everyone, for the work you do.

Back on the 6th page of the thread, Manly first posted a question about changing the standard 0/53/18 FFB spec to 0/51/33, with a 3/3 shatter. Several people thought is was not a good change, but Manly defended the idea. Looking at the math, the changes in moving to this variation seem not to affect the majority of the spells & the spell priorities for most of the FFB spec, and it might add some interesting help with aoe if you are using blizzard more than flamestrike.

After all this time, has anyone tried it, and how did it work? It looks interesting to me as a possible variation. Might only affect trash pulls for the most part, but everything helps.

 

Zaldinar

My first iteration of FFB spec contained shatter vice Frost Channeling, which was an absolute dream for soloing and questing.

However, the primary concern of these boards is typically raiding, where shatter is solely useful in AoE situations and if you spec 1/3 Imp Blizz to get the frostbite procs, and the targets can be nova'ed (not always the case). To gain extra points in frost to get shatter you need to sacrifice range or pushback/threat reduction or flavor spells (BW/DB) or crit chance on LB/Pyro, all of which can help your single target output or your AoE Control options (gluth juggling, basically).

Shatter is great fun solo, Honestly it doesn't add an incredible amount to your AoE potential (15% chance for 50% crit = 7.5% crit chance gain, assuming that the frostbite doesn't break by the time the next waves crit chance is calculated), but does help provide a little more control than existed before via the imp blizzard snare.

Your mileage may vary, it's mostly a personality decision, like taking DB/BW over points in World in Flames.

To truly model the game, we first must research it.
http://forums.worldofwarcraft.com/th...icId=109841969
Proven TheoryCrafting Stuff, chain casting in a PTR near you soon.

 

Joneleth

I thought this was an interesting thought, so I decided to carry out the tests myself.

Scorch Critical Strike Chance: 34.58% (paper doll fire crit) + 6% (Incineration) + 10% (Improved Scorch Debuff) = 50.48%

Method: Data collected using Recount; Boss Dummy was scorched twice to apply debuff, then Recount was reset for the trial. Scorch was cast until OOM (ranging from 160-170 casts per run, 20 total trials). Gear adjusted to hit cap (swapped Embrace of the Spider for Mark of the War Prisoner), so all Scorches hit.


Total number of casts: 3288
Total number of crits: 1598
Observed critical strike chance: 0.4860
Expected critical strike chance: 0.5048


I'll let someone more qualified handle the statistics, if they're so inclined. These results seem to support the existence of crit suppression somewhere on the order of 2%, but I'd rather not be presumptuous.