Guild Wars 2

Everyone complaining about the damage proc from Ai/Fire. Maybe stop playing zerk/glass if you can’t keep alive.

sample size???

test used???

p<0.05???

Zuik.7158:

sample size???

test used???

p<0.05???

About 100% crit chance on ele (perma fury, 6 in air, rampager’s amulet, eagle’s runes and signet of fire), attacking almost twice per second. In 487 tries, I never get a proc before 5 seconds has passed since the last proc.

Oh god pls make air anyhow more worthy than blood pls

It has a critical chance!! It CAN procc. Lol, this guy trolled you all.

zone.1073:

I’ve been testing this on necro using death perception, and the internal cooldown on air is indeed 5 sec.

Zuik.7158:

sample size???

test used???

p<0.05???

Necro with death perception, 100%+ crit chance, golem group in HotM with life blasts that pierce through 8+ golems each life blast.

30 minutes of using life blast on said golem group.

Number of activations of air sigil in < 5 seconds = 0 times.

Everyone here is wrong and obviously has never heard of Schrodinger’s Cat.

Because the sigil has a 50/50 chance of proc, it is both alive and dead at the same time.
#schooled

/thread – You’re welcome.

manysongsgw.2931:

Everyone here is wrong and obviously has never heard of Schrodinger’s Cat.

Because the sigil has a 50/50 chance of proc, it is both alive and dead at the same time.
#schooled

/thread – You’re welcome.

why people are trying to prove that the sigil CD is indeed 5 seconds when the proc has only a 50% chance of activating per hit, even if crit rate is 100%, is beyond me.

It’s not possible to validate that the sigil rate is working unless you have a perma 100% crit rate, clock every proc of Air sigil, and happen to catch a time gap between them of 4.99 seconds or less, assuming you attack (unfailingly, mind you, in a game with cast times and lag) once every second.

(and even then it would be broken. You’d need to catch one at -precisely- 3.00 to confirm it is “working as intended”.)

That’s why I asked Sunfish for how he came to that conclusion. Instead people are trying to prove it’s broken, even though he’s the one who made the claim and clearly has no desire to back up what he’s saying.

Yeah, I’m also wondering if people are confused of the ICD itself and actually have a stopwatch on it.

The proc won’t go off every 3rd second even if the sigil was 100% chance on crit. If you proc it on the 1st second the CD would only be available on the 5th second as 2nd 3rd and 4th seconds are the CD.

I did test it, and it didn’t seem to be triggering any faster than every 5s. Even more, when equipped with both fire/air (removing accuracy so only 86% crit rate), air procced at the exact same frequency as fire.

The damage, however, was higher for air due to a greater damage coefficient.

manysongsgw.2931:

Everyone here is wrong and obviously has never heard of Schrodinger’s Cat.

Because the sigil has a 50/50 chance of proc, it is both alive and dead at the same time.
#schooled

/thread – You’re welcome.

“How can we possible determine 50/50….yea, let’s just flip a coin, that’s heads or tails…We would need 50 coins, then 50 more: 50-50….then my computer can analyze it…with science”

BlackBeard.2873:

manysongsgw.2931:

Everyone here is wrong and obviously has never heard of Schrodinger’s Cat.

Because the sigil has a 50/50 chance of proc, it is both alive and dead at the same time.
#schooled

/thread – You’re welcome.

“How can we possible determine 50/50….yea, let’s just flip a coin, that’s heads or tails…We would need 50 coins, then 50 more: 50-50….then my computer can analyze it…with science”

http://www.youtube.com/watch?v=MTLp14MKDDU

You’d actually need a random number generator (for crit chance) to determine whether you flip that coin or not.

Previous page comments are making me smile.

Exedore.6320:

For all the people that don’t want to believe this, here’s a simple scientific setup:

Hypothesis: Sigil of Air has a 5sec ICD

Test Setup: Engineer with as high a crit chance as possible (rampager or assassin amulet, 6 into precision line, precision runes). Equip Sigil of Air and flamethrower. Auto-attack indestructible dummy (~4 attacks/second and you can let it keep doing).

If Sigil of Air procs once within less than 5 seconds, the hypothesis is disproven. Let’s say we have a 50% crit chance and a 50% proc rate. On average, one of those 4 attacks within a second should proc the sigil if it’s off ICD. Let it go for 10-15 minutes and see if the interval between any proc was less than 5 seconds.

It’s not rocket surgery.

thank the lord. why did I have to wait so long to see someone say this? this was literally my first thought on the matter

Azure The Heartless.3261:

why people are trying to prove that the sigil CD is indeed 5 seconds when the proc has only a 50% chance of activating per hit, even if crit rate is 100%, is beyond me.

My test used 100% crit piercing life blasts that pierced through 8+ golems per life blast.

8 golems = 50% + 50% + 50% + 50% + 50% + 50% + 50% + 50% chance of a proc per life blast (cast time: 1 second).

Please understand the math. The results are conclusive.

zone.1073:

Azure The Heartless.3261:

why people are trying to prove that the sigil CD is indeed 5 seconds when the proc has only a 50% chance of activating per hit, even if crit rate is 100%, is beyond me.

My test used 100% crit piercing life blasts that pierced through 8+ golems per life blast.

8 golems = 50% + 50% + 50% + 50% + 50% + 50% + 50% + 50% chance of a proc per life blast (cast time: 1 second).

Please understand the math. The results are conclusive.

Did you account for the skill note and latency?

Notes

Attack speed is approximately 1.4 seconds per attack.
And even so, that result is theoretical (and reasonable), but not conclusive. Flipping a coin eight times doesn’t guarantee an outcome.

Azure The Heartless.3261:

Did you account for the skill note and latency?

Notes

Attack speed is approximately 1.4 seconds per attack.

Irrelevant.

If the air sigil has an internal cooldown of 3 seconds, it would proc before 5 seconds using my test. That is absolute and irrefutable.

The fact that my data yielded 0 (zero. ZERO) procs before 5 seconds for 30 minutes of testing means that the test is conclusive.

End thread.

BlackBeard.2873:

I did test it, and it didn’t seem to be triggering any faster than every 5s. Even more, when equipped with both fire/air (removing accuracy so only 86% crit rate), air procced at the exact same frequency as fire.

The damage, however, was higher for air due to a greater damage coefficient.

manysongsgw.2931:

Everyone here is wrong and obviously has never heard of Schrodinger’s Cat.

Because the sigil has a 50/50 chance of proc, it is both alive and dead at the same time.
#schooled

/thread – You’re welcome.

“How can we possible determine 50/50….yea, let’s just flip a coin, that’s heads or tails…We would need 50 coins, then 50 more: 50-50….then my computer can analyze it…with science”

http://www.youtube.com/watch?v=MTLp14MKDDU

It does go back to probability and statistics.

The expected value of a sigil proc’ing with 86% crit chance will be different from 100%. People with lower crit chances should see a longer time between procs.

Also, I would like to see a stop watch and video of peoples’ experiments. I don’t disbelieve the realm of possibility the sigil could have a longer ICD than stated tooltip. Just there’s no hard evidence produced so far to prove it . There’s the issue with the base eV of 100% crit chance and 100% proc rate in theory only yielding a proc every 4th second after the 1st proc.

zone.1073:

Azure The Heartless.3261:

Did you account for the skill note and latency?

Notes

Attack speed is approximately 1.4 seconds per attack.

Irrelevant.

If the air sigil has an internal cooldown of 3 seconds, it would proc before 5 seconds using my test. That is absolute and irrefutable.

Wishful thinking. You’re looking at the math yourself. I don’t know how you’re deluding yourself into believing that a theoretical chance becomes absolute if you do it eight times, but have fun with that.

Skill rate and your latency are also very relevant, because your description of 5 seconds could be offset due to those factors. (It would be helping your case of air CD being > 3 seconds, but at least we’d have a solid frame of reference.)

Again, your data is reasonable, but not absolute. Don’t market it as such. Like I said, I really don’t care if it’s broken or not, so I have no need to prove it is working as intended.

If you’re going to say it is broken though, you need to tie up loose ends.

Phantaram.1265:

I tested in about 5 minutes of my free time. Less than than most of you probably took typing the things you typed.

86.5% crit chance channeling Air #1 on scepter (elementalist)

1st 1 minute trial with fire sigil (5 second cooldown)
procced 11 times

2nd 1 minute trial with fire sigil
Procced 11 times

1st 1 minute trial with air sigil (listed 3 second cooldown)
procced 11 times

2nd 1 minute trial with air sigil
procced 11 times

Oh no. I’ve provided proof when the burden was not mine. Did I commit some kind of treasonous act?

You need to find the minimum intervals between strikes not the number in a minute, you haven’t proved anything I’m afraid. (It looks like they are the same ICD but your method of proof doesn’t show anything)

Azure The Heartless.3261:

Wishful thinking. You’re looking at the math yourself.

Did you really stop reading mid sentence?

zone.1073:

The fact that my data yielded 0 (zero. ZERO) procs before 5 seconds for 30 minutes of testing means that the test is conclusive.

I was also using rage sigil, meaning that every 30 seconds I would rapidly cast a chain of life blasts that had 0.5 second cast times.

The point is that within 30 minutes of testing, there should have been at least 1 data point of an air proc before 5 seconds. Again, the results are conclusive.

But go ahead and keep posting and bumping this thread. Perhaps the devs will read it.

zone.1073:

Azure The Heartless.3261:

Wishful thinking. You’re looking at the math yourself.

Did you really stop reading mid sentence?

zone.1073:

The fact that my data yielded 0 (zero. ZERO) procs before 5 seconds for 30 minutes of testing means that the test is conclusive.

I was also using rage sigil, meaning that every 30 seconds I would rapidly cast a chain of life blasts that had 0.5 second cast times.

The point is that within 30 minutes of testing, there should have been at least 1 data point of an air proc before 5 seconds. Again, the results are conclusive.

But go ahead and keep posting and bumping this thread. Perhaps the devs will read it.

Hopefully they do, then it would get a buff (or a tooltip fix), both of which are reasonable options.

No, I did not stop reading mid sentence. I conceded that your results were reasonable. I was directly addressing your assertion that the data is irrefutable simply because you got a 0 proc rate for repeated testing on an outcome affected by random number generation.

No, it’s not absolute. It’s convincing, but not 100% conclusive.

All thats going to happen if it IS 5 seconds and is changed (fixed?) to 3 is a massive backlash when people are already frustrated with Air/Fire.

You guys are still talking about silly maths! Crazy guys! Just crazy! This can be resolved so easily! Either it’s bugged and who cares because it’s strong enough as is! Anet has a huge list of bugs that turn into features! Or it’s not bugged in which case why the frickfrack are you guys typing in this thread!

Wahoo! Bye frands!

ArrDee.2573:

You guys are still talking about silly maths! Crazy guys! Just crazy! This can be resolved so easily! Either it’s bugged and who cares because it’s strong enough as is! Anet has a huge list of bugs that turn into features! Or it’s not bugged in which case why the frickfrack are you guys typing in this thread!

Wahoo! Bye frands!

+1

ArrDee.2573:

You guys are still talking about silly maths! Crazy guys! Just crazy! This can be resolved so easily! Either it’s bugged and who cares because it’s strong enough as is! Anet has a huge list of bugs that turn into features! Or it’s not bugged in which case why the frickfrack are you guys typing in this thread!

Wahoo! Bye frands!

-1

Talking maths is fun!

Turk.5460:

You would need a huge sample size even if it was a 99% chance. That’s the point of probability and statistics.

This is completely false, and the point of probability and statistics are to draw conclusions without HUGE sample sizes. In fact, as few as 3 trials for air/fire each would be enough (although 4 or 5 would be better) to give a relatively high confidence (p-value) that air/fire share timing mechanics. How? You could take 3 tests (1-2 mins) each to get a mean and standard deviation on number of procs. Your hypothesis is that they are from different populations, making your null hypothesis that they actually have the same timing mechanics. From here it is a simple t-test (justified b/c the proc-chances are completely random, so they will follow a normal distribution) to find your confidence value that they are in-fact the same or not.

:)

BlackBeard.2873:

Turk.5460:

You would need a huge sample size even if it was a 99% chance. That’s the point of probability and statistics.

This is completely false, and the point of probability and statistics are to draw conclusions without HUGE sample sizes. In fact, as few as 3 trials for air/fire each would be enough (although 4 or 5 would be better) to give a relatively high confidence (p-value) that air/fire share timing mechanics. How? You could take 3 tests (1-2 mins) each to get a mean and standard deviation on number of procs. Your hypothesis is that they are from different populations, making your null hypothesis that they actually have the same timing mechanics. From here it is a simple t-test (justified b/c the proc-chances are completely random, so they will follow a normal distribution) to find your confidence value that they are in-fact the same or not.

:)

When you are testing data that is like the data in air sigils, the bigger sample size the better. I dare you to go to a statistics professor and say that statistics is to draw info from limited data.

SobeSoul.6910:

BlackBeard.2873:

Turk.5460:

You would need a huge sample size even if it was a 99% chance. That’s the point of probability and statistics.

This is completely false, and the point of probability and statistics are to draw conclusions without HUGE sample sizes. In fact, as few as 3 trials for air/fire each would be enough (although 4 or 5 would be better) to give a relatively high confidence (p-value) that air/fire share timing mechanics. How? You could take 3 tests (1-2 mins) each to get a mean and standard deviation on number of procs. Your hypothesis is that they are from different populations, making your null hypothesis that they actually have the same timing mechanics. From here it is a simple t-test (justified b/c the proc-chances are completely random, so they will follow a normal distribution) to find your confidence value that they are in-fact the same or not.

:)

When you are testing data that is like the data in air sigils, the bigger sample size the better. I dare you to go to a statistics professor and say that statistics is to draw info from limited data.

I’m a professor in engineering…and yea that is what we teach. So, in the engineering sense (such as hypothesis testing), yea… Bigger sample sizes can give greater degrees of confidence (p-values), but the purpose of statistics is to draw conclusions with the minimum sample size. Just looking at the data, you should be able to reach a p-value of 0.05 with only a few samples.

I spent some time testing with a Longbow and stop watch (on warrior, 80% crit double shots) I actually think this may be the case, honestly. BUT~ THAT SAID, does not need a buff. It just needs to state 5 second cooldown instead of 3.

Jelzouki.4128:

For the love of all that is holy…..if you really want to test this out you don’t test it on an engineer exedore.

Hidden Killer on thief: 100% crit chance when striking from stealth, revealed buff is 3s. you have enough time to restealth and test the proc intervals. It may be off by 1 second or less.

The proc is not a 100% chance on a crit though…

Turk.5460:

Jelzouki.4128:

For the love of all that is holy…..if you really want to test this out you don’t test it on an engineer exedore.

Hidden Killer on thief: 100% crit chance when striking from stealth, revealed buff is 3s. you have enough time to restealth and test the proc intervals. It may be off by 1 second or less.

The proc is not a 100% chance on a crit though…

Well, with a longbow the 4th shot would always land between the 3 and 5 second mark and 5th would always land at about 5.8 seconds. After about 10 minutes of straight shooting and counting shots (with stop watch confirmation and longbow shooting double shots, 80% crit) I had a very frequent 5th shot proc and not a single 4th shot proc in 10 minutes of testing, and remember each test frame has a double chance to proc. That should be enough to make a pretty valid conclusion.

That said, it still needs a tooltip adjustment, not a functionality adjustment.

BlackBeard.2873:

SobeSoul.6910:

BlackBeard.2873:

Turk.5460:

You would need a huge sample size even if it was a 99% chance. That’s the point of probability and statistics.

This is completely false, and the point of probability and statistics are to draw conclusions without HUGE sample sizes. In fact, as few as 3 trials for air/fire each would be enough (although 4 or 5 would be better) to give a relatively high confidence (p-value) that air/fire share timing mechanics. How? You could take 3 tests (1-2 mins) each to get a mean and standard deviation on number of procs. Your hypothesis is that they are from different populations, making your null hypothesis that they actually have the same timing mechanics. From here it is a simple t-test (justified b/c the proc-chances are completely random, so they will follow a normal distribution) to find your confidence value that they are in-fact the same or not.

:)

When you are testing data that is like the data in air sigils, the bigger sample size the better. I dare you to go to a statistics professor and say that statistics is to draw info from limited data.

I’m a professor in engineering…and yea that is what we teach. So, in the engineering sense (such as hypothesis testing), yea… Bigger sample sizes can give greater degrees of confidence (p-values), but the purpose of statistics is to draw conclusions with the minimum sample size. Just looking at the data, you should be able to reach a p-value of 0.05 with only a few samples.

I am an econ major. In econometrics we do not like micronumerosity at all. When we need to correlate data from the real world we simply cannot be satisfied with small data sets. For instance trying to correlate wage gaps between sexes and races is better if you have a huge data set.

Back to the air sigil thing. If the chances of proc’ing air on crit is 50% and you have 40% crit chances. You should only see an air proc 20% of the times when it comes off CD. Given the fact that you fail to crit sixty out of one hundred it wouldn’t be rare to miss that proc a lot over a small sample size.

Not a single person who has presented their version of “irrefutable” evidence apparently has the confidence to actually submit a thread on the bug report forum…

SobeSoul.6910:

I am an econ major. In econometrics we do not like micronumerosity at all. When we need to correlate data from the real world we simply cannot be satisfied with small data sets. For instance trying to correlate wage gaps between sexes and races is better if you have a huge data set.

Back to the air sigil thing. If the chances of proc’ing air on crit is 50% and you have 40% crit chances. You should only see an air proc 20% of the times when it comes off CD. Given the fact that you fail to crit sixty out of one hundred it wouldn’t be rare to miss that proc a lot over a small sample size.

But we are testing with specs that have 80% crit chance and 50% proc on crit. When you have the large number of variables in a human population, you need large sample sizes. By minimizing the number of variables, we can use smaller samples sizes. If you are an econ major, really think about the difference between wage gaps in human populations (and all the variables that come into play) and a hard-coded machine system with only 1 variable. Does it make sense why small sample sizes can reveal a difference here, while larger sample sizes are required to compare populations that share a few similarities and many differences for each member?

More learning if you look up student-t test, which is a perfect application here. Good luck with your econ studies, gaining a deeper understanding of the limitations your professors have suggested will make you certainly more valuable in the workplace.

Turk.5460:

Not a single person who has presented their version of “irrefutable” evidence apparently has the confidence to actually submit a thread on the bug report forum…

My lack of “confidence” is because I don’t want the kitten buffed.

I recorded a few minutes of attacking some golems (multiple at once) at 105% crit chance with ranger short bow in HotM. Could anyone who knows anything about video editing point me towards some programs for editing in a stopwatch thingy for the air procs?

Here’s something different, a video test with 100% crit chance on wand air ele.
https://www.youtube.com/watch?v=i2XoifRhx1w

According to the result, seems likely to have 5s recharge. The last strike was not at 100% crit though the previous three seemed accurate.

Edit: also apparently kitten recharge is full of soft cuddles.

BlackBeard.2873:

SobeSoul.6910:

I am an econ major. In econometrics we do not like micronumerosity at all. When we need to correlate data from the real world we simply cannot be satisfied with small data sets. For instance trying to correlate wage gaps between sexes and races is better if you have a huge data set.

Back to the air sigil thing. If the chances of proc’ing air on crit is 50% and you have 40% crit chances. You should only see an air proc 20% of the times when it comes off CD. Given the fact that you fail to crit sixty out of one hundred it wouldn’t be rare to miss that proc a lot over a small sample size.

But we are testing with specs that have 80% crit chance and 50% proc on crit. When you have the large number of variables in a human population, you need large sample sizes. By minimizing the number of variables, we can use smaller samples sizes. If you are an econ major, really think about the difference between wage gaps in human populations (and all the variables that come into play) and a hard-coded machine system with only 1 variable. Does it make sense why small sample sizes can reveal a difference here, while larger sample sizes are required to compare populations that share a few similarities and many differences for each member?

More learning if you look up student-t test, which is a perfect application here. Good luck with your econ studies, gaining a deeper understanding of the limitations your professors have suggested will make you certainly more valuable in the workplace.

Ohhh kittentt shots fired

Again to sobesoul, an ele testing with air 1 on scepter with 2.8 hits per second with an 86.5% crit chance and a 50% proc chance, will have 1.211 hits per second that proc air sigil. With that kind of chance to proc you don’t need a huge sample size to see that its clearly not working right.

People always complain about sample sizes its a weak argument…

Combined hours of testing in this thread have yielded no Air Sigil hits under 5 seconds apart.

Maybe if we do it for 10,000 hours with a stopwatch, it can be also proven to the people that took Economics and know better.

BlackBeard.2873:

SobeSoul.6910:

I am an econ major. In econometrics we do not like micronumerosity at all. When we need to correlate data from the real world we simply cannot be satisfied with small data sets. For instance trying to correlate wage gaps between sexes and races is better if you have a huge data set.

Back to the air sigil thing. If the chances of proc’ing air on crit is 50% and you have 40% crit chances. You should only see an air proc 20% of the times when it comes off CD. Given the fact that you fail to crit sixty out of one hundred it wouldn’t be rare to miss that proc a lot over a small sample size.

But we are testing with specs that have 80% crit chance and 50% proc on crit. When you have the large number of variables in a human population, you need large sample sizes. By minimizing the number of variables, we can use smaller samples sizes. If you are an econ major, really think about the difference between wage gaps in human populations (and all the variables that come into play) and a hard-coded machine system with only 1 variable. Does it make sense why small sample sizes can reveal a difference here, while larger sample sizes are required to compare populations that share a few similarities and many differences for each member?

More learning if you look up student-t test, which is a perfect application here. Good luck with your econ studies, gaining a deeper understanding of the limitations your professors have suggested will make you certainly more valuable in the workplace.

There’s more underlining variables than the few you listed btw.

My question is, why are you so defensive that this couldn’t possibly be a thing when a plethora of evidence shows it COULD be the case, while no evidence has been shown that it is NOT the case.

As far as has been proven so far, it seems as if it very well may have a 5 second ICD. No amount of testing could possibly “prove” to you that it certainly has a CD of 5 seconds because “statistics” says it could just be bad luck. We could go around and around with this forever….

Assuming it is 5 sec icd, why are you trying so hard to deny it? Because you don’t want it buffed? Because I don’t either. Hence why I said I would rather them change the tooltip instead. But that’s no reason to try to sabotage the testing with insatiable needs for “proof”.

Lordrosicky.5813:

It honestly is the reason I quit. Getting insta gibbed by auto attacks and 4 fire/air sigils in a few seconds EVERY few seconds = no room for skill or counter play. It really is brainless. Unavoidable SPIKE damage. Celestial isn’t even that OP compared to these. I would make the icd 10 seconds for air and fire.

And people in general complain about passive play in this game with mesmer clone death traits debilitating dissipation and crippling dissipation as prime examples.

What’s even more funny is the fact that you can dodge those while you simply take the air/fire proc.

Still that’s for those that use zerker specs and are genuinely squishy to complain about.

Personally, if it is bugged, which it more than likely is, I mean….come on. I’ll honestly be happy with a tooltip change, even though the buff/fix of sigil of air will help me vs a lot of classes that dont use the same sigils, its still pretty powerful.