However, Sleign, that does not explain the discrepancy I observed in the data I initially presented in this thread. It is not that the argument is based on the occurrence of a special event. Rather, the argument is based on statistical analysis of data collected over a rather significant period of time. The occurrence of special events simply shows that we are on the right track.
True, but by your own admission the data you presented is hardly a controlled experiment. It is unreasonable to expect outcomes based on a controlled situation from an uncontrolled system. Variations such as mob levels, buffs/stat changes, etc are going to significantly alter results.
Specifically, your undead slaying gear (which, why would you wear it in strat and not scholo?). It's difficult to say how much this would impact your crit % without knowing just how much you used it. Using it in one instance does not say much because you may run that particular instance 1 in 10 runs, or half the time. Your stats are also odd, in saying that you have parried/blocked backstabs, however minimal they may be.
True, but by your own admission the data you presented is hardly a controlled experiment. It is unreasonable to expect outcomes based on a controlled situation from an uncontrolled system. Variations such as mob levels, buffs/stat changes, etc are going to significantly alter results.
Specifically, your undead slaying gear (which, why would you wear it in strat and not scholo?). It's difficult to say how much this would impact your crit % without knowing just how much you used it. Using it in one instance does not say much because you may run that particular instance 1 in 10 runs, or half the time. Your stats are also odd, in saying that you have parried/blocked backstabs, however minimal they may be.
Variations were present, true, but let's look at that undead slaying gear, for example. My tooltip crit rate, even with that gear on, is greater than the observed crit rate. I also included, in the post, a calculation showing the probability of observing such a low crit rate if I had worn the undead slaying for the full duration of the parse. I found that probability to be significantly small. I also showed that regardless of whether I overestimated or underestimated the typical level of the mobs I was fighting, the discrepancy was too great to be accounted for by target level variation.
Most of the potential sources of deviation "should" have skewed my observed crit rate above my tooltip value. Instead, I observed a value nearly 3% lower. You're right that the data was not part of a controlled experiment. However, we would simply be hamstringing ourselves if we limited any analysis of special attacks to carefully controlled data sets, because we just don't have any of those (and they take plenty of time to gather; it took me three months to gather 5900 Backstabs' worth of data).
It would not necessarily be difficult to do a controlled test. It could simply be set up with a low level rogue (for duration's sake only) with enough crit to work with the numbers, and a tank engaging a higher level mob. Then simply have the tank hold the mob while the rogue spams sinister strike or backstab from behind to negate block and parry. You could muster as many number as your patience would allow, then compare the given values to what a 1 roll or 2 roll system would yield. I personally do not have the patience to run such a test, but I know there are crazy people that enjoy doing such things.
It would not necessarily be difficult to do a controlled test. It could simply be set up with a low level rogue (for duration's sake only) with enough crit to work with the numbers, and a tank engaging a higher level mob. Then simply have the tank hold the mob while the rogue spams sinister strike or backstab from behind to negate block and parry. You could muster as many number as your patience would allow, then compare the given values to what a 1 roll or 2 roll system would yield. I personally do not have the patience to run such a test, but I know there are crazy people that enjoy doing such things.
I ran a controlled test that, with 100% conclusiveness, shows that backstab does not follow the 1-roll system that we thought. With the 1-roll system, there was 0 chance for a backstab hit, and yet, i received a backstab hit. Thus the 1-roll system cannot be in place for backstab (and presumably all special melee attacks).
Looking over the information, it seems that the primary basis for the 2 roll argument on special attacks is that things still occur that should not occur given base math in a single roll system. This is not necessarily indicative of a 2 roll system, however. We could take a given example already known in the game to show that the base math need not apply to more extreme situations: defense vs crits.
Given a warrior with a high enough defense that a normal mob (something without an augmented crit %) would have a 0% chance to crit him, we know this is not the case; you can always be crit. This leads to 2 possible conclusions: minimum hard cap, or diminishing returns. Unfortunately I don't know how exactly the numbers run once you get past a certain point of low crit %. But, it is not unreasonable to say that there could be minimum caps or diminishing returns on other forms, such as chance to normal hit. All other possible outcomes could simply truncate at 95%, 96%, whatever, and a minimum x% is reserved for normal hits.
Just because you land a normal hit, it does not necessarily dictate that it is not a 1 roll system.
Looking over the information, it seems that the primary basis for the 2 roll argument on special attacks is that things still occur that should not occur given base math in a single roll system. This is not necessarily indicative of a 2 roll system, however. We could take a given example already known in the game to show that the base math need not apply to more extreme situations: defense vs crits.
Given a warrior with a high enough defense that a normal mob (something without an augmented crit %) would have a 0% chance to crit him, we know this is not the case; you can always be crit. This leads to 2 possible conclusions: minimum hard cap, or diminishing returns. Unfortunately I don't know how exactly the numbers run once you get past a certain point of low crit %. But, it is not unreasonable to say that there could be minimum caps or diminishing returns on other forms, such as chance to normal hit. All other possible outcomes could simply truncate at 95%, 96%, whatever, and a minimum x% is reserved for normal hits.
Just because you land a normal hit, it does not necessarily dictate that it is not a 1 roll system.
Assumption: With the 1-roll system, there will be 44.4% misses and 55.6% crits and 0% hits.
Having just 1 single time where a hit lands contradicts the above and thus the above cannot be the case. A single hit was found (multiple times actually) and thus contradicts the assumption. Thus a 1-roll system is not used for backstab. QED
The 1-roll system has been proven and shown to work with white melee attacks (there are extensive tests specifically dealing with crushing blows to prove this). What we do not know with the yellow attacks is how many rolls are made. Is there a roll for all forms of mitigation? Is block seperate? Questions like that have not been answered, but the point still stands, the 1-roll theory cannot work with backstab.
Looking over the information, it seems that the primary basis for the 2 roll argument on special attacks is that things still occur that should not occur given base math in a single roll system. This is not necessarily indicative of a 2 roll system, however. We could take a given example already known in the game to show that the base math need not apply to more extreme situations: defense vs crits.
Given a warrior with a high enough defense that a normal mob (something without an augmented crit %) would have a 0% chance to crit him, we know this is not the case; you can always be crit. This leads to 2 possible conclusions: minimum hard cap, or diminishing returns. Unfortunately I don't know how exactly the numbers run once you get past a certain point of low crit %. But, it is not unreasonable to say that there could be minimum caps or diminishing returns on other forms, such as chance to normal hit. All other possible outcomes could simply truncate at 95%, 96%, whatever, and a minimum x% is reserved for normal hits.
simple testing results from letting 3 mobs outside stormwind beat on my priest for an half an hour. My priest is lvl 45 with 225 defense being beat on my mobs lvl 5-8. It is possible to become uncritable by low level mobs by having high enough defense.
Having just 1 single time where a hit lands contradicts the above and thus the above cannot be the case. A single hit was found (multiple times actually) and thus contradicts the assumption.
Correct, but it does NOT prove there is a second roll, it only proves that if there is a single table, you cannot completely push hit off the table. This however does not require a two-roll system.
Correct, but it does NOT prove there is a second roll, it only proves that if there is a single table, you cannot completely push hit off the table. This however does not require a two-roll system.
Granted that even if hit can't be pushed off the table, I still shouldn't experience 3% less crit than expected over a sample of 5924 Backstabs, even given the "lack of control" of the sample.
Having just 1 single time where a hit lands contradicts the above and thus the above cannot be the case. A single hit was found (multiple times actually) and thus contradicts the assumption.
Correct, but it does NOT prove there is a second roll, it only proves that if there is a single table, you cannot completely push hit off the table. This however does not require a two-roll system.
excepting that it has been proven by people doing glancing blows tests that there is most definitely not a cap on +hit. There was a warrior who got enough buffs to be above the crit cap and only had crits/glances. Not to mention the people who tested stuff on loatheb.
excepting that it has been proven by people doing glancing blows tests that there is most definitely not a cap on +hit. There was a warrior who got enough buffs to be above the crit cap and only had crits/glances. Not to mention the people who tested stuff on loatheb.
By saying that, you're claiming that:
1) Special attacks are different from white attacks (two rolls versus one roll)
2) Special attacks are not different from white attacks (since one has no hit cap, the other can't)
If they're on a different system, you can't correlate the two at all.
Before you start to drift, and your soul begins to scream.
I just wanted to tell you, that you're listening to a dream.
excepting that it has been proven by people doing glancing blows tests that there is most definitely not a cap on +hit. There was a warrior who got enough buffs to be above the crit cap and only had crits/glances. Not to mention the people who tested stuff on loatheb.
By saying that, you're claiming that:
1) Special attacks are different from white attacks (two rolls versus one roll)
2) Special attacks are not different from white attacks (since one has no hit cap, the other can't)
If they're on a different system, you can't correlate the two at all.
What I am saying is that the backstab system does not follow the one-roll white attack system. If it did, there would be no hit/crit cap just like has been proven to exist. (which is exactly what the other poster was claiming and its downright wrong). I am also claiming that the observed data I had shows that backstab critrate follows on landed attacks, and not swings. If there was some 99% crit cap or whatever the other poster is claiming, I would not have been able to repeat my test multiple times and get a hit about half the time. I was once a believer in the one-roll system for all melee attacks, and now i know this not to be the case. If anyone wants to run a test similar to what I did, they will come away with the same conclusion.
It does not totally prove that there is a 2-roll system. There are a few other things that could explain it.
There is some form of cap in place.
Debuffs (and buffs?) doesn't affect hitrate the way we assume.
Improved backstab is weird/bugged.
I can't say why this would be the case or why it would only affect specials though. More testing is needed. Is it possible to get a critrate 76% higher than hitrate to test it without debuffs?
While it is possible that special attacks differ from white damage attacks by having a fixed minimum-hit% on a 1-roll system.
The test to this is:
Get a rogue to backstab a target, while being under the effect of your proposed hit% cap(being in a situation, where a 1-roll system only would allow crit or miss). This will give both hits, misses and crits, as has been tested. The actual test is to raise the crit rate of the rogue testing.
In a 2-roll system, we will see an increase in crits, at the cost of hits.
In a 1-roll system, we won't see any change.
It might be a good idea to purposefully lower your critchance, for the initial testcase, so you can raise it enough to observe a sizable change. Ofcourse this doesn't rule out backstab being bugged, or any other parts of the game code.
I'll just throw something in here for someone else to think about: could it be that weapon skill behaves differently on attacks that can glance and attacks that can't? To the best of my knowledge weapon skill for ranged weapons (hunters in practise) significantly increases their hit% and crit% - I've heard mention of 0.24% per 1 skill.
To my understanding weapon skill also normally gives 0.04% hit crit -dodge -block and -parry per point as well as reducing the damage penalty of glancing blows. But what if that isn't quite the entire story? What if on every attack that can't glance weapon skill has a more significant effect on other stats? And perhaps on attacks that can't be parried or blocked it has yet a more significant effect on those that remain: hit, crit and dodge?
Conversely, those 15 points of defence a typical boss has over player's attack skill could then have a more significant impact on attacks which can not glance. In fact any slight difference advantage of level the mob has on player would have little effect on white hit and crit rate (while giving glances to white attacks) but might have a dramatic impact on yellow hit and crit rate.
1st Roll: Did it hit or miss?
2nd Roll: Damage loss through resistance (25%/50%/75%)
If combat log is to be trusted it seems resistances apply first and only then it's checked if the attack crits.
You can thus have written in combat log something like:
Your moonfire crits XX for 150 (100 resisted).
We all know 40% resistances don't happen, so this would be 50% resistance followed by 50% extra damage from critical hit. It might also just mean that combat log doesn't reflect the order in which the rolls are made.
I'll just add an observation I made a couple nights ago, not completely on topic, but I believe not totally off: I was in Scarlet Monastery helping a low level friend with my druid, and as you may remember in the library part, some mobs use an ability that works somewhat like shield block (or "improved blocking" as reported by SCT spell alert).
I noticed that when they were using it (sound and visual cues apply here), I couldn't land a rake, no matter how hard I tried, while claw landed all the times without problems. We're talking about a 60 druid vs a lvl 34 mob and 3-5 rakes not landing in a row.
That led me to believe that not all special attacks are calculated in the same way, or at least not all of them follow the same rules for connecting on a mob under special circumstances (in this case blocking). Food for thought?
For mere curiosity I'll test it again tonight and try to report here my findings.
While it is possible that special attacks differ from white damage attacks by having a fixed minimum-hit% on a 1-roll system.
The test to this is:
Get a rogue to backstab a target, while being under the effect of your proposed hit% cap(being in a situation, where a 1-roll system only would allow crit or miss). This will give both hits, misses and crits, as has been tested. The actual test is to raise the crit rate of the rogue testing.
In a 2-roll system, we will see an increase in crits, at the cost of hits.
In a 1-roll system, we won't see any change.
It might be a good idea to purposefully lower your critchance, for the initial testcase, so you can raise it enough to observe a sizable change. Ofcourse this doesn't rule out backstab being bugged, or any other parts of the game code.
The original post in this thread contains a lot of this data you are looking for. And the evidence supports a 2-roll system.
Originally Posted by MatsT
It does not totally prove that there is a 2-roll system. There are a few other things that could explain it.
There is some form of cap in place.
Debuffs (and buffs?) doesn't affect hitrate the way we assume.
Improved backstab is weird/bugged.
I can't say why this would be the case or why it would only affect specials though. More testing is needed. Is it possible to get a critrate 76% higher than hitrate to test it without debuffs?
1) There is no cap in place, when testing with more crit than hit, a little more than half my landed backstabs were crits.
2) The debuff is bugged maybe? Well, it sure felt like the debuff was working as it usually took 2-3 backstabs to land a hit
3) I think we would have noticed something a long time ago
4) It would only affect specials because specials follow a 2-roll system and regular white attacks follow a 1-roll system. The 1-roll system was probably originally done to save processing time. However, with the use warriors have found for shield block, I doubt blizzard would go away from a 1-roll system, unless they did so upon the release of an expansion.
Instead of arguing with everyone in this thread that has actually gone out and tested it because they did not believe the initial post either. How about go out and test it yourself with whatever controlled test you want to use and see if you still support a 1-roll system for special attacks.
Assumption: With the 1-roll system, there will be 44.4% misses and 55.6% crits and 0% hits.
As another poster already pointed out, your assumption is manyfold.
Explicitely you are assuming many things:
1) It is a 1-roll system
2) 44.4% miss rate and 55.6% crit rate ... YES this is an assumption, as it does not have to be true (see e.g. hard caps on misses against lvl 1 boars; it could equally be applied to hits).
By getting that one hit you got, you just proved that the assumption i.e. 1)+2) does not hold true.
You did NOT prove 1) wrong.
And no, refering to glancing blows does not help your case.
Assumption: With the 1-roll system, there will be 44.4% misses and 55.6% crits and 0% hits.
As another poster already pointed out, your assumption is manyfold.
Explicitely you are assuming many things:
1) It is a 1-roll system
2) 44.4% miss rate and 55.6% crit rate ... YES this is an assumption, as it does not have to be true (see e.g. hard caps on misses against lvl 1 boars; it could equally be applied to hits).
By getting that one hit you got, you just proved that the assumption i.e. 1)+2) does not hold true.
You did NOT prove 1) wrong.
And no, refering to glancing blows does not help your case.
EDIT: Typo
1)There is no hardcap on misses against any lvl mob. See rogue that raided aq40 with 25% hit and received 0 misses the entire night. Also, it has been tested that there is no hit cap as people have reported 0% hits when over the crit cap (ie miss+dodge+crit+glance > 100). These baseless assumptions which have already been proven wrong do nothing towards invalidating the results of these tests.
2)All the evidence presented in this thread, which comes from a wide variety of testing, directly contradicts a 1-roll system. Instead of making inaccurate and baseless assumptions, why not design and follow through with your own test? This is exactly what I did when I saw the thread and did not believe the claim made by the original poster. I now fully back the original poster on this claim which is against everything we have thought to be true for the past 2 years.
1)There is no hardcap on misses against any lvl mob. See rogue that raided aq40 with 25% hit and received 0 misses the entire night. Also, it has been tested that there is no hit cap as people have reported 0% hits when over the crit cap (ie miss+dodge+crit+glance > 100). These baseless assumptions which have already been proven wrong do nothing towards invalidating the results of these tests.
pf, I am not going to touch 1 roll versus 2 roll here, as I have not looked into it enough. However, you are still making two assumptions that are incompatible.
You claim:
1) Special is different from white (i.e. "it's two rolls, not one!")
2) Special is same as white (i.e. "white has no crit/hit/whatever cap, therefore special doesn't!")
How can you so clearly argue that special is the same at the same time you say it's different?
Since the data has proven our first assumption was wrong - special attacks are not the same as melee attacks - you cannot correlate any data between the two. At a brief inspection it looks like the 2 roll system is how specials work, but I have not seen you prove it because you keep inserting faulty data from white attacks.
Before you start to drift, and your soul begins to scream.
I just wanted to tell you, that you're listening to a dream.
Just to restate what Pf is saying in a different way, keep in mind that several things we "know" about the one-roll system and combat mechanics are additional assumptions we've made based on the results of testing for white attacks. That is, I don't know of anyone ever doing a controlled (or even semi-controlled) sample of instant attacks and parsing the observed crit rate compared to the tooltip. It was done for regular melee attacks, and then we all simply assumed the behavior extended to instant attacks. Whenever things like "blocked crits" came up, we simply called them anomalies caused by lag, and moved on.
A lot of our assumptions about the way things work are just that. Trust me, I was as hardline a supporter of the one-roll system as anyone else until I started seriously analyzing these data. I can't stress enough the magnitude of that difference between my observed crit rate and my expected crit rate. Over 5924 Backstabs, the probability of experiencing a crit rate that low, given my tooltip and the expected variations, is very low.
Doing an even more controlled sample of this really isn't too hard. Just take a mod like KombatStats and parse any instance/raid from start to finish. Just make sure your crit rate doesn't change too much (a couple percent isn't a big deal, but you might want to leave out that fight against Loatheb :-P). Post your data and let's see. A couple hundred Backstabs should be sufficient.
Next time I run UBRS or ZG or whatever, I'll screenshot the parse myself and post the data again. Let's try to veer away from this idle discussion of assumptions and ideas, and get back towards hard data and facts.
1st Roll: Did it hit or miss?
2nd Roll: Damage loss through resistance (25%/50%/75%)
If combat log is to be trusted it seems resistances apply first and only then it's checked if the attack crits.
You can thus have written in combat log something like:
Your moonfire crits XX for 150 (100 resisted).
We all know 40% resistances don't happen, so this would be 50% resistance followed by 50% extra damage from critical hit. It might also just mean that combat log doesn't reflect the order in which the rolls are made.
Interesting. I never bothered to do the math whenever i got a partial resist crit to check that out. I'll start paying more attention to this.
But, like you said, maybe the combat log just doesnt follow event by event, and just outputs in the single line the end effects (You did THIS much and THIS much was resisted)
Eyonix's post does mention the resistance check being the second role, and if spells are on a 2-roll system then it doesnt make sense for crit to be on the second... how could one roll determine the resist chunk taken out and whether or not it was a crit?
Also, another thing I just thought about... does this mean basically all hunter shots are on a 2-roll system as well?
1)There is no hardcap on misses against any lvl mob. See rogue that raided aq40 with 25% hit and received 0 misses the entire night. Also, it has been tested that there is no hit cap as people have reported 0% hits when over the crit cap (ie miss+dodge+crit+glance > 100). These baseless assumptions which have already been proven wrong do nothing towards invalidating the results of these tests.
pf, I am not going to touch 1 roll versus 2 roll here, as I have not looked into it enough. However, you are still making two assumptions that are incompatible.
You claim:
1) Special is different from white (i.e. "it's two rolls, not one!")
2) Special is same as white (i.e. "white has no crit/hit/whatever cap, therefore special doesn't!")
How can you so clearly argue that special is the same at the same time you say it's different?
Since the data has proven our first assumption was wrong - special attacks are not the same as melee attacks - you cannot correlate any data between the two. At a brief inspection it looks like the 2 roll system is how specials work, but I have not seen you prove it because you keep inserting faulty data from white attacks.
Ok, since you seem to think that I am sticking to #2 as my basis for something (which im not), I'll give you some data that helps you understand the assumptions. I have not missed a special attack while raiding with over 10% hit in over a year (unless i had a -hit debuff). That means there is no misscap. Also, dagger rogues on loatheb with over 100% backstab critrate do not report hits, they get crits. Another thing, cold blood, increases your critrate by 100% on the affected spells, see http://thotbott.com/?sp=14177 However, I dont remember a single time me, or any other rogue reported not critting while having cold blood activated, thus there is no minimum on %hit.
11/25/06, 5:58 PM
Similar Threads
