You forget about the law of averages. Whenever you're talking about statistics a larger sample size will more closely reflect the true odds than a small sample size. You may have a day where you hit a 10% upgrade chance on the first try but you're gonna have days where you never hit those 10% odds. In the end it will all balance out to leave you with a 10% chance of getting lucky.
If you really wanna post stats then do a 1000 or 10000 tries list. You'll be bankrupt but at least the numbers will be accurate.
0
plasticbatMember, NW M9 PlaytestPosts: 12,456Arc User
You forget about the law of averages. Whenever you're talking about statistics a larger sample size will more closely reflect the true odds than a small sample size. You may have a day where you hit a 10% upgrade chance on the first try but you're gonna have days where you never hit those 10% odds. In the end it will all balance out to leave you with a 10% chance of getting lucky.
If you really wanna post stats then do a 1000 or 10000 tries list. You'll be bankrupt but at least the numbers will be accurate.
You won't get bankrupt because you do that in the test server. One of guild mate actually did exactly that for something else last night.
*** The game can read your mind. If you want it, you won't get it. If you don't expect to get it, you will. ***
Today I tried to upgrade a 10% item and had 60 successive failures. Yes, 60. Someone in my guild said that RNG is basically broken in PE. I went to a regular map and it succeeded on the 3rd time.
Ok now that is interesting Subnocte, not statistically significant, but interesting. Could you repeat the attempt 1000 times in PE and then in another zone please? Also at any time did you take RNGesus' name in vain? What was the the sock count of the room you were in and which direction were you facing?
Ok now that is interesting Subnocte, not statistically significant, but interesting. Could you repeat the attempt 1000 times in PE and then in another zone please? Also at any time did you take RNGesus' name in vain? What was the the sock count of the room you were in and which direction were you facing?
Yes, as it's really, really common to get 60 failures in a row on a 10% chance. I mean, I'm sure you'd be pedantic enough to say how it could happen 500 million times in a row. Have you ever considered that the generator may actually be bugged, and that the randomness is affected by zone activity? I guess not, as it's much easier just to be a total arrogant HAMSTER.
0
thefabricantMember, NW M9 PlaytestPosts: 5,248Arc User
Ok now that is interesting Subnocte, not statistically significant, but interesting. Could you repeat the attempt 1000 times in PE and then in another zone please? Also at any time did you take RNGesus' name in vain? What was the the sock count of the room you were in and which direction were you facing?
Yes, as it's really, really common to get 60 failures in a row on a 10% chance. I mean, I'm sure you'd be pedantic enough to say how it could happen 500 million times in a row. Have you ever considered that the generator may actually be bugged, and that the randomness is affected by zone activity? I guess not, as it's much easier just to be a total arrogant HAMSTER.
You had a 0.17% chance to fail that many times, which means that out of 1000 people, approximately 2 would fail that many times in a row. Hello there, unlucky Mr 500! I am pleased to meet you, I am lucky Mr 100 who succeeded on the 1st try at upgrading a 1% chance with no coal ward!
People just need to think more before hitting the upgrade button -And probably at least use a preservation ward-. That 10% chance is a direct warning. Even worse is people trying to do a 1% upgrade. It says 10% chance to pass, doing it 10 times won't mean 100% chance to pass! If something is at 1% chance it is a "use coal ward" hint.
0
minotaur2857Member, NW M9 PlaytestPosts: 1,141Arc User
I'm also losing confidence in the rng actually being random. I've just done a load of 20/25% chances upgrading rank 7s to rank 9.
28 attempts 210 pres wards used so 28/238 and not a single one succeeded first time. More of them took >10 tries than took less, the worst took 32, next worst was 17.
My personal records for bad luck here are:
60% 11 straight fails (this is a crafting thing and I'm 24/62 on 60% chances) 20% 46 fails 10% 67 fails
The crafting one is interesting (I'm a statistics graduate), and using one of the standard confidence limit checks via the normal approximation to the binomial (which the sample size is large enough to do, you need about 40) you would not be confident that the mean really is 60%. And these are over a period so don't fall foul of what I talk about next.
My suspicion is that if the server is busy, it will sometimes not generate a new random number and use the last one. Since after a success, you will need to move some stuff around before trying again on the next enchant, you will get a new number (or new seed if that's the issue). If you try again straight after a failure, you will do this more quickly and may not.
I'm also losing confidence in the rng actually being random. I've just done a load of 20/25% chances upgrading rank 7s to rank 9.
28 attempts 210 pres wards used so 28/238 and not a single one succeeded first time. More of them took >10 tries than took less, the worst took 32, next worst was 17.
My personal records for bad luck here are:
60% 11 straight fails (this is a crafting thing and I'm 24/62 on 60% chances) 20% 46 fails 10% 67 fails
The crafting one is interesting (I'm a statistics graduate), and using one of the standard confidence limit checks via the normal approximation to the binomial (which the sample size is large enough to do, you need about 40) you would not be confident that the mean really is 60%. And these are over a period so don't fall foul of what I talk about next.
My suspicion is that if the server is busy, it will sometimes not generate a new random number and use the last one. Since after a success, you will need to move some stuff around before trying again on the next enchant, you will get a new number (or new seed if that's the issue). If you try again straight after a failure, you will do this more quickly and may not.
To add to your post, a sample size of 40 seems very small. Isn't your confidence level too low with only 40 samples? Plus since your 40 tries is throughout a duration, don't you think your sample size is so much smaller compared that actual population that your confidence level is just too low at that point? Even at refinements at 60%, the number of tries people make in a day could easily break 500 in a day.
I guess number generation by the server could be an issue, but very few people actually report such issues about having 50-100 failures in a row. I think this issue only started when this thread began. On the other hand, we have lock box opening videos that seem to show that this should not be a problem. People open 1k-3k lock boxes and youtube and if number generation is an issue then there should be instances where they get 2 or more legendary results in a row, they do open those lock boxes fairly rapidly. I don't know, maybe cryptic should run a test on this to really know.
"As the good archmage often admonishes me, I ought not to let my mind wander, as it's too small to go off by itself." -Danilo Thann[/quote]
0
minotaur2857Member, NW M9 PlaytestPosts: 1,141Arc User
I'm also losing confidence in the rng actually being random. I've just done a load of 20/25% chances upgrading rank 7s to rank 9.
28 attempts 210 pres wards used so 28/238 and not a single one succeeded first time. More of them took >10 tries than took less, the worst took 32, next worst was 17.
My personal records for bad luck here are:
60% 11 straight fails (this is a crafting thing and I'm 24/62 on 60% chances) 20% 46 fails 10% 67 fails
The crafting one is interesting (I'm a statistics graduate), and using one of the standard confidence limit checks via the normal approximation to the binomial (which the sample size is large enough to do, you need about 40) you would not be confident that the mean really is 60%. And these are over a period so don't fall foul of what I talk about next.
My suspicion is that if the server is busy, it will sometimes not generate a new random number and use the last one. Since after a success, you will need to move some stuff around before trying again on the next enchant, you will get a new number (or new seed if that's the issue). If you try again straight after a failure, you will do this more quickly and may not.
To add to your post, a sample size of 40 seems very small. Isn't your confidence level too low with only 40 samples? Plus since your 40 tries is throughout a duration, don't you think your sample size is so much smaller compared that actual population that your confidence level is just too low at that point? Even at refinements at 60%, the number of tries people make in a day could easily break 500 in a day.
I guess number generation by the server could be an issue, but very few people actually report such issues about having 50-100 failures in a row. I think this issue only started when this thread began. On the other hand, we have lock box opening videos that seem to show that this should not be a problem. People open 1k-3k lock boxes and youtube and if number generation is an issue then there should be instances where they get 2 or more legendary results in a row, they do open those lock boxes fairly rapidly. I don't know, maybe cryptic should run a test on this to really know.
40 is the minimum for normal approximation to binomial, my sample size was 62.
n=62 p(success) = 0.6 mean is np = 37.2 variance is np(1-p) = 14.88 standard deviation = sqrt(variance) = 3.86
IIRC the 95% confidence limit for the mean is 2 standard deviations in the normal distribution so this fails comfortably (37.2-7.72 = 29.48 and I'm more than another SD off that)
I've been upgrading further today, 5 enchs 10-11. Average should be 95 pres wards, 100 attempts, I actually used 120 pres wards so not utterly unreasonable, but the 70 I used on one of the attempts did have me tearing my hair out. This was done in an empty instance. I finally got something to work first time, it was a 6->7 which then went to 8 and to 9 first time too immediately after, these are still after I did a load more 7-8-9-10s the only first time successes.
Comments
If you really wanna post stats then do a 1000 or 10000 tries list. You'll be bankrupt but at least the numbers will be accurate.
28 attempts 210 pres wards used so 28/238 and not a single one succeeded first time. More of them took >10 tries than took less, the worst took 32, next worst was 17.
My personal records for bad luck here are:
60% 11 straight fails (this is a crafting thing and I'm 24/62 on 60% chances)
20% 46 fails
10% 67 fails
The crafting one is interesting (I'm a statistics graduate), and using one of the standard confidence limit checks via the normal approximation to the binomial (which the sample size is large enough to do, you need about 40) you would not be confident that the mean really is 60%. And these are over a period so don't fall foul of what I talk about next.
My suspicion is that if the server is busy, it will sometimes not generate a new random number and use the last one. Since after a success, you will need to move some stuff around before trying again on the next enchant, you will get a new number (or new seed if that's the issue). If you try again straight after a failure, you will do this more quickly and may not.
I guess number generation by the server could be an issue, but very few people actually report such issues about having 50-100 failures in a row. I think this issue only started when this thread began. On the other hand, we have lock box opening videos that seem to show that this should not be a problem. People open 1k-3k lock boxes and youtube and if number generation is an issue then there should be instances where they get 2 or more legendary results in a row, they do open those lock boxes fairly rapidly. I don't know, maybe cryptic should run a test on this to really know.
n=62
p(success) = 0.6
mean is np = 37.2
variance is np(1-p) = 14.88
standard deviation = sqrt(variance) = 3.86
IIRC the 95% confidence limit for the mean is 2 standard deviations in the normal distribution so this fails comfortably (37.2-7.72 = 29.48 and I'm more than another SD off that)
I've been upgrading further today, 5 enchs 10-11. Average should be 95 pres wards, 100 attempts, I actually used 120 pres wards so not utterly unreasonable, but the 70 I used on one of the attempts did have me tearing my hair out. This was done in an empty instance. I finally got something to work first time, it was a 6->7 which then went to 8 and to 9 first time too immediately after, these are still after I did a load more 7-8-9-10s the only first time successes.