Backstab: Two Rolls? - Page 5

Faytte · 11/27/06, 2:24 PM

Wouldnt something simple be...

Cold Blood Eviscerate on a level 55 (or something lower level) with 0% hit gear. Crit rate is 100%, but it can still miss?

Maybe I'm entering the discussion to late;)

Ahiru · 11/27/06, 2:33 PM

Originally Posted by Faytte

Wouldnt something simple be...

Cold Blood Eviscerate on a level 55 (or something lower level) with 0% hit gear. Crit rate is 100%, but it can still miss?

Maybe I'm entering the discussion to late;)

A Cold Blood-affected attack can be dodged, parried and possibly blocked, if position allows, but if you have sufficient +hit, it will never miss. Rogues have long complained about dodge causing CB to be wasted, but I have never seen a rogue actually miss a CB attack unless they had less than +5% to hit. Furthermore, with sufficient +hit gear, I've never seen myself miss a BS, SS or Evis even without Cold Blood, except when my hit rate was debuffed. All evidence I've seen suggests that there is no cap to the effects of +hit on special attacks.

Faytte · 11/27/06, 2:39 PM

I suppose one could say that crit pushes hit off the table but not miss/dodge/parry for specials. So the best argument would be to get someone to have a very high miss rate, like from the cthun trash "dust cloud" and watch the crit % of their special attacks over time through it vs normal hits and see if either is being pushed off the hit table?

Having a crit rate similar to normal (on successful hits) would imply a 2 roll system, where as showing up all crits or all hits on succesful attacks would imply 1 roll. Flaws with my theory?:P

pf · 11/27/06, 3:19 PM

Originally Posted by Faytte

I suppose one could say that crit pushes hit off the table but not miss/dodge/parry for specials. So the best argument would be to get someone to have a very high miss rate, like from the cthun trash "dust cloud" and watch the crit % of their special attacks over time through it vs normal hits and see if either is being pushed off the hit table?

Having a crit rate similar to normal (on successful hits) would imply a 2 roll system, where as showing up all crits or all hits on succesful attacks would imply 1 roll. Flaws with my theory?:P

No flaws in your theory, just flaw in your post for not reading the previous page where I ran a test with 60%+crit and less than 60% hit. If hit was pushed off the table, this would be the case. Hit was not pushed off the table and it follows that special attacks are not on a 1-roll system. The problem is that people refuse to look at the overwhelming evidence in this thread against a 1-roll special attack (or backstab in almost all the tests) system. They are digging for flaws in the logic, but refuse to test it out themselves and only add useless and baseless conjecture.

◊ Kalman · 11/27/06, 3:24 PM

Originally Posted by pf

Originally Posted by Faytte

I suppose one could say that crit pushes hit off the table but not miss/dodge/parry for specials. So the best argument would be to get someone to have a very high miss rate, like from the cthun trash "dust cloud" and watch the crit % of their special attacks over time through it vs normal hits and see if either is being pushed off the hit table?

Having a crit rate similar to normal (on successful hits) would imply a 2 roll system, where as showing up all crits or all hits on succesful attacks would imply 1 roll. Flaws with my theory?:P

No flaws in your theory, just flaw in your post for not reading the previous page where I ran a test with 60%+crit and less than 60% hit. If hit was pushed off the table, this would be the case. Hit was not pushed off the table and it follows that special attacks are not on a 1-roll system. The problem is that people refuse to look at the overwhelming evidence in this thread against a 1-roll special attack (or backstab in almost all the tests) system. They are digging for flaws in the logic, but refuse to test it out themselves and only add useless and baseless conjecture.

Now, a more interesting question is if the hit/crit rate actually follows across *landed* swings, or if it's more complicated than that, for special attacks. Are we seeing a case where miss and crit overlay each other, for example, similar to some of the theories on warrior shield block table filling? Your data probably doesn't have enough samples to determine reliably whether crit % is across landed blows or not, but it's something to check before definitively saying it's 2 rolls; it could conceivably be one roll with a different table-filling algorithm.

(It's probably two rolls, I agree, but we're not *quite* there yet.)

pf · 11/27/06, 3:54 PM

Originally Posted by Kalman

Originally Posted by pf

Originally Posted by Faytte

I suppose one could say that crit pushes hit off the table but not miss/dodge/parry for specials. So the best argument would be to get someone to have a very high miss rate, like from the cthun trash "dust cloud" and watch the crit % of their special attacks over time through it vs normal hits and see if either is being pushed off the hit table?

Having a crit rate similar to normal (on successful hits) would imply a 2 roll system, where as showing up all crits or all hits on succesful attacks would imply 1 roll. Flaws with my theory?:P

No flaws in your theory, just flaw in your post for not reading the previous page where I ran a test with 60%+crit and less than 60% hit. If hit was pushed off the table, this would be the case. Hit was not pushed off the table and it follows that special attacks are not on a 1-roll system. The problem is that people refuse to look at the overwhelming evidence in this thread against a 1-roll special attack (or backstab in almost all the tests) system. They are digging for flaws in the logic, but refuse to test it out themselves and only add useless and baseless conjecture.

Now, a more interesting question is if the hit/crit rate actually follows across *landed* swings, or if it's more complicated than that, for special attacks. Are we seeing a case where miss and crit overlay each other, for example, similar to some of the theories on warrior shield block table filling? Your data probably doesn't have enough samples to determine reliably whether crit % is across landed blows or not, but it's something to check before definitively saying it's 2 rolls; it could conceivably be one roll with a different table-filling algorithm.

(It's probably two rolls, I agree, but we're not *quite* there yet.)

I did not do enough parsing to definitively tell how the system works. I agree with that. What I did find out, without any doubt, is that the 1-roll system that we use for auto-attacks most definitely does not work for special attacks. You could come up with a 1-roll algorithm that perfectly mimicked a 2-roll system, but whats the point? When testing, I had a ~60% backstab crit rate and it felt like every other landed backstab was a hit. The only way to really get a solid test would be to kite a servant to swamp of sorrows and have it be tanked while you beat on it and get aggro from one of the mobs there to get the debuff and then scan through combat logs. I dont know 2 other people (tank and healer) that would be willing to do this, and as such, I am going to assume a typical 2-roll system (one roll for mitigation, one for hit/crit) until I see evidence pointing to something else.

• Vulajin · 11/27/06, 4:22 PM

Originally Posted by Kalman

Now, a more interesting question is if the hit/crit rate actually follows across *landed* swings, or if it's more complicated than that, for special attacks. Are we seeing a case where miss and crit overlay each other, for example, similar to some of the theories on warrior shield block table filling? Your data probably doesn't have enough samples to determine reliably whether crit % is across landed blows or not, but it's something to check before definitively saying it's 2 rolls; it could conceivably be one roll with a different table-filling algorithm.

(It's probably two rolls, I agree, but we're not *quite* there yet.)

But here's a question...how would we tell the difference between misses and crits "overlapping" versus crit% being across landed swings? It seems like, unless Blizzard is using some really whacked out strategy to build their combat tables, the two results would be the same.

Additionally, do you have any thoughts as to how "overlap" would occur in an "average" case where you're simply Backstabbing with ~60% chance to crit against a mob with standard miss and avoidance rates?

I think we'll be well-served in obtaining more data from "average" cases rather than trying to look at extremes.

◊ Kalman · 11/27/06, 4:27 PM

Originally Posted by Vulajin

Originally Posted by Kalman

Now, a more interesting question is if the hit/crit rate actually follows across *landed* swings, or if it's more complicated than that, for special attacks. Are we seeing a case where miss and crit overlay each other, for example, similar to some of the theories on warrior shield block table filling? Your data probably doesn't have enough samples to determine reliably whether crit % is across landed blows or not, but it's something to check before definitively saying it's 2 rolls; it could conceivably be one roll with a different table-filling algorithm.

(It's probably two rolls, I agree, but we're not *quite* there yet.)

But here's a question...how would we tell the difference between misses and crits "overlapping" versus crit% being across landed swings? It seems like, unless Blizzard is using some really whacked out strategy to build their combat tables, the two results would be the same.

Additionally, do you have any thoughts as to how "overlap" would occur in an "average" case where you're simply Backstabbing with ~60% chance to crit against a mob with standard miss and avoidance rates?

I think we'll be well-served in obtaining more data from "average" cases rather than trying to look at extremes.

Assume 35% miss rate, 70% crit rate. Assume that miss replaces crit in this table (i.e., overlaps).

If miss and crit overlap, we'd expect 35 of 100 swings to miss, 35 of 100 to crit, and 30 of 100 to hit.

If crit is across landed blows, we'd expect 35 of 100 to miss, 45.5 to crit, and 19.5 to hit.

In an average case, something like 10% avoidance vs. 65% crit rate, it'd be:

Two roll: 10 miss/dodge, 58.5 crits, 31.5 hits.
Single roll, overlap: 10 miss/dodge, 55 crits, 35 hits.

Much harder to differentiate, in that case.

• Vulajin · 11/27/06, 4:47 PM

Originally Posted by Kalman

Assume 35% miss rate, 70% crit rate. Assume that miss replaces crit in this table (i.e., overlaps).

If miss and crit overlap, we'd expect 35 of 100 swings to miss, 35 of 100 to crit, and 30 of 100 to hit.

If crit is across landed blows, we'd expect 35 of 100 to miss, 45.5 to crit, and 19.5 to hit.

In an average case, something like 10% avoidance vs. 65% crit rate, it'd be:

Two roll: 10 miss/dodge, 58.5 crits, 31.5 hits.
Single roll, overlap: 10 miss/dodge, 55 crits, 35 hits.

Much harder to differentiate, in that case.

Ah, I was misunderstanding your wording of miss and crit overlapping. However, I think even with a difference of 3.5%, there should be notable differences in samples of several hundred Backstabs at least, depending on which of those was the case. The difference in being able to determine statistical certainty here would be that we would need to control the crit rate more carefully than I originally did.

In the first post, where the goal was not even necessarily to prove a two-roll theory, but simply to show that the data did not support our existing one-roll model, the lack of control was permissible because I clearly showed that any discrepancies should have skewed the observed crit rate higher, rather than 3% lower. However, in order to differentiate between the two potential models you've suggested, we'll need more precision. Again, I think we'll benefit from more data, whether it's for an extreme case or for an "average" case. As long as there's a fairly good sample size and a carefully controlled crit rate, we're okay.

pf · 11/27/06, 4:49 PM

Originally Posted by Kalman

Originally Posted by Vulajin

Originally Posted by Kalman

Now, a more interesting question is if the hit/crit rate actually follows across *landed* swings, or if it's more complicated than that, for special attacks. Are we seeing a case where miss and crit overlay each other, for example, similar to some of the theories on warrior shield block table filling? Your data probably doesn't have enough samples to determine reliably whether crit % is across landed blows or not, but it's something to check before definitively saying it's 2 rolls; it could conceivably be one roll with a different table-filling algorithm.

(It's probably two rolls, I agree, but we're not *quite* there yet.)

But here's a question...how would we tell the difference between misses and crits "overlapping" versus crit% being across landed swings? It seems like, unless Blizzard is using some really whacked out strategy to build their combat tables, the two results would be the same.

Additionally, do you have any thoughts as to how "overlap" would occur in an "average" case where you're simply Backstabbing with ~60% chance to crit against a mob with standard miss and avoidance rates?

I think we'll be well-served in obtaining more data from "average" cases rather than trying to look at extremes.

Assume 35% miss rate, 70% crit rate. Assume that miss replaces crit in this table (i.e., overlaps).

If miss and crit overlap, we'd expect 35 of 100 swings to miss, 35 of 100 to crit, and 30 of 100 to hit.

If crit is across landed blows, we'd expect 35 of 100 to miss, 45.5 to crit, and 19.5 to hit.

In an average case, something like 10% avoidance vs. 65% crit rate, it'd be:

Two roll: 10 miss/dodge, 58.5 crits, 31.5 hits.
Single roll, overlap: 10 miss/dodge, 55 crits, 35 hits.

Much harder to differentiate, in that case.

would not be hard to disprove the single roll overlap theory. Repeat the same test I did using sinister strike instead of backstab. If you get a crit, your miss replacing crit assumption is contradicted, and thus cannot be the case.

suicuique · 11/28/06, 5:49 AM

Just to add some (very small) observation from our Loatheb Kill last night.

I was very wary about getting aggro world buffed, so i just autoattacked till first spore buff (I was in first group).

My character sheet crit rate was 43.41%. Spore buffed it was 93.41%.
Assume my crit rate against loatheb was 92.8 %. I had +17 hit equipped.
The dodge rate of Loatheb is difficult to measure, but assume the standard dodge rate for bosses minus weapon skill, that is 5.32%.

Of all my 110 heroic strikes when spore buffed none missed, 100 critted, 10 hit.

Take it as a hint.

Yes, I know that the sample was absurdly low. But it should not be difficult to collect the data of the subsequent kills too. Just takes time.

EDIT: Typos

Faytte · 11/28/06, 7:57 AM

Originally Posted by pf

Originally Posted by Faytte

I suppose one could say that crit pushes hit off the table but not miss/dodge/parry for specials. So the best argument would be to get someone to have a very high miss rate, like from the cthun trash "dust cloud" and watch the crit % of their special attacks over time through it vs normal hits and see if either is being pushed off the hit table?

Having a crit rate similar to normal (on successful hits) would imply a 2 roll system, where as showing up all crits or all hits on succesful attacks would imply 1 roll. Flaws with my theory?:P

No flaws in your theory, just flaw in your post for not reading the previous page where I ran a test with 60%+crit and less than 60% hit. If hit was pushed off the table, this would be the case. Hit was not pushed off the table and it follows that special attacks are not on a 1-roll system. The problem is that people refuse to look at the overwhelming evidence in this thread against a 1-roll special attack (or backstab in almost all the tests) system. They are digging for flaws in the logic, but refuse to test it out themselves and only add useless and baseless conjecture.

Sometimes at work I dont have the liberty or thoroughly reading threads. I can catch a post here and there while I wait for things to compile. Sorry:)