Probabilities of skills rising

Cyram
Emperator
Posts: 41
Joined: Wed 06 Jan 2016, 16:02
Realm: Filgar

Probabilities of skills rising

Postby Cyram » Tue 02 May 2017, 16:54

Just thought I'd share a quick observation,

I've recently had a series of unlikely (mind you, far from impossible) chains of bad luck. These add up to quite low probabilities of happening, so I'm putting a post here in case it actually is an error (or lack of randomness or whatever implementation thing may cause this).

So, in no particular order:

1) failing salancedar cutting ten times in a row. At a presence of 32%, the chance for this to happen is about 2.1%
2) failing salancedar cutting 8 times in a row. The chance for this to happen is 4,6%
3) trying to level stonecutting from 86/86 to either 87/86 or 86/87. The game says the chance for a skill up cutting ornaments is 4,5%. I cut 53 volcanite blocks so far, bringing the probability for NO skillup down to 8,7%.

Obviously, it is also possible i'm just very unlucky, but I think there are two more likely causes of error possible:
1) the given probabilities for a skill up / success are wrong
2) the random number generator is not really random. This is a surprisingly common problem in many computer programs.

On the bright side: on average, the system does work. When our guild cut lots of grimalstone monoliths for grimalpur we netted about the 10% you'd expect. So, I think there likely is a random number system that is not really uniformly random. At least for me, it would reduce frustration if it were.
However, it is not in any way game breaking as the spikes up and down seem to average out (as you'd expect from the limit theorem).

User avatar
Cloudigger
Mystique
Posts: 1491
Joined: Fri 20 Nov 2015, 20:43
Realm: -

Re: Probabilities of skills rising

Postby Cloudigger » Tue 02 May 2017, 17:12

Hello!

The "dice" are rolled every time you make a check. So if there 32% of chance to get something, the game throws the dice every time... so it means that each time you have about 1 chance about of three to succeed. On 8 checks... it seams plausible that you failed every time.

Cyram
Emperator
Posts: 41
Joined: Wed 06 Jan 2016, 16:02
Realm: Filgar

Re: Probabilities of skills rising

Postby Cyram » Tue 02 May 2017, 17:21

Possible yes, but not probable. Especially when you consider that these chains occurred three times now this week.

It is still possible that i'm just unlucky, but having three of these in short order amounts to a chance of 1 in 10.000, which sounds less plausible. On the other hand, people do win the lottery sometimes ;)

User avatar
Balok
Bâtisseur
Posts: 140
Joined: Wed 06 Jan 2016, 16:45
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Balok » Tue 02 May 2017, 18:39

There is no such thing as a truly random number when dealing with computers. This is because they are deterministic by design. Simulations, implemented in random number generation libraries, vary in how well the simulate randomness. I have wondered about the randomness myself on the occasion of a large string of unlucky outcomes, but the truth is that you'd have to run a chi-squared test against a really large batch of random numbers generated by the Therian code to answer this question definitively.
Gentlemen! You can't fight in here!! This is the War Room!

User avatar
Céline
Mystique
Posts: 2286
Joined: Fri 20 Nov 2015, 20:42
Realm: -

Re: Probabilities of skills rising

Postby Céline » Tue 02 May 2017, 19:05

When I test new tasks or loots on test-server, I do 3 or 4 series of 1000x action/loot. Globally, I have something a little better than expected (near 1% better). But even with 3000 or 4000 actions, it's not enough to test it correctly, but I have other things to do :P
---
~Contact : support@virtys.com~

blah
Défricheur
Posts: 8
Joined: Tue 02 May 2017, 19:41
Realm: Fauconnia

Re: Probabilities of skills rising

Postby blah » Tue 02 May 2017, 19:46

I feel like this game's RNG is weighted on top of it being a RNG. for instance, if something has a chance to produce 1-5 items, there is a much larger chance that you will get 1 or 2 rather than 5. If that makes sense.
~I am Blah, hear me roar!~

Ambrielle
Emperator
Posts: 70
Joined: Fri 15 Jan 2016, 12:55
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Ambrielle » Tue 02 May 2017, 20:10

Cyram wrote:Obviously, it is also possible i'm just very unlucky, but I think there are two more likely causes of error possible:
1) the given probabilities for a skill up / success are wrong

My feeling, not based on a mass of empirical data, is that the probabilities are pretty accurate until your skill level gets over 80. Then they start to diverge with reality, as in the real skill increase chance is rather less than stated skill increase chance.

The number of items you get when it is random (mentioned by "blah") definitely averages out over a long run.

User avatar
Neku
Bourlingueur
Posts: 61
Joined: Wed 06 Jan 2016, 04:57
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Neku » Wed 03 May 2017, 03:08

Last time I did send a mail to the support about leveling my engineering skill with the samponar task. I would like to share it here:

My first email:
"This is my third/fourth time making samponar to level up my engineering. The chances of level up is 76.8%, but I have been doing it for three times and no level up at all. That level up percentage is pretty high as well. So, could you please check that for me?"

They replied:
Moderation : Thank you to not post private answers made by the support team.


Me replied:
"Let's think about it the other way around. In my case, now my success rate to get level up is 21.4% (100%-76.8%). That means I was really really lucky to get level up from those four that have 21.4% successful rate.
What do you think? Am I the most luckiest person? I still think there is an issue. Please check carefully."

Their final response:
Moderation : Thank you to not post private answers made by the support team.

Many people noticed the same issue when skill reached 70 and above. I guess they still do not agree that there is an issue with the RNG skill level up.

Cyram
Emperator
Posts: 41
Joined: Wed 06 Jan 2016, 16:02
Realm: Filgar

Re: Probabilities of skills rising

Postby Cyram » Wed 03 May 2017, 08:31

Neku wrote:Last time I did send a mail to the support about leveling my engineering skill with the samponar task. I would like to share it here:

My first email:
"This is my third/fourth time making samponar to level up my engineering. The chances of level up is 76.8%, but I have been doing it for three times and no level up at all. That level up percentage is pretty high as well. So, could you please check that for me?"

They replied:
Moderation : Thank you to not post private answers made by the support team.


Actually, the chance to NOT level up is 23.2%, so the chance to get no level up three times in a row is actually 0.232^3 = 0,012. This means the chance for that to happen is a tiny bit more than 1%.

Once again, as in my experience posted before, this is certainly possible, but not very likely.

Many people noticed the same issue when skill reached 70 and above. I guess they still do not agree that there is an issue with the RNG skill level up.


It seems we do have quite a number of very unlucky people. Even considering psychological biases (you notice failure way more), imho there's quite a few one digit probability failures happening. Taking all of these into consideration, the chances for this to happen go down to almost 0.

Since the dice roll itself is probably only a few lines of code, nothing bad can come from checking it just to make sure we're a bunch of star crossed skillers ;)

User avatar
Neku
Bourlingueur
Posts: 61
Joined: Wed 06 Jan 2016, 04:57
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Neku » Wed 03 May 2017, 11:05

Yup, very unlucky. :evil:

There is nothing I can do other than grinding and complaining.

Hope this thread would change their view points.

User avatar
Cloudigger
Mystique
Posts: 1491
Joined: Fri 20 Nov 2015, 20:43
Realm: -

Re: Probabilities of skills rising

Postby Cloudigger » Wed 03 May 2017, 11:31

We already checked the codes many times. We also asked two programmers that did not work for us to verify if there was anything wrong it. No one found anything. We absolutely have no interest to have a system that will disadvantage you.

blah
Défricheur
Posts: 8
Joined: Tue 02 May 2017, 19:41
Realm: Fauconnia

Re: Probabilities of skills rising

Postby blah » Wed 03 May 2017, 11:40

Cloudigger wrote:We already checked the codes many times. We also asked two programmers that did not work for us to verify if there was anything wrong it. No one found anything. We absolutely have no interest to have a system that will disadvantage you.


Sooo just bad luck! haha :D
~I am Blah, hear me roar!~

Cyram
Emperator
Posts: 41
Joined: Wed 06 Jan 2016, 16:02
Realm: Filgar

Re: Probabilities of skills rising

Postby Cyram » Wed 03 May 2017, 14:06

Cloudigger wrote:We already checked the codes many times. We also asked two programmers that did not work for us to verify if there was anything wrong it. No one found anything. We absolutely have no interest to have a system that will disadvantage you.


No doubt about that. Actually, all my experiences with Virtys have been positive :)

is it possible that the skill up chance displayed in game is simply too high? that would also explain the experiences we had.

User avatar
Cloudigger
Mystique
Posts: 1491
Joined: Fri 20 Nov 2015, 20:43
Realm: -

Re: Probabilities of skills rising

Postby Cloudigger » Wed 03 May 2017, 14:21

Cyram wrote:
Cloudigger wrote:We already checked the codes many times. We also asked two programmers that did not work for us to verify if there was anything wrong it. No one found anything. We absolutely have no interest to have a system that will disadvantage you.


No doubt about that. Actually, all my experiences with Virtys have been positive :)

is it possible that the skill up chance displayed in game is simply too high? that would also explain the experiences we had.


I will try to investigate if it's a SL display or something.

User avatar
Neku
Bourlingueur
Posts: 61
Joined: Wed 06 Jan 2016, 04:57
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Neku » Wed 03 May 2017, 14:40

Cyram wrote:
Cloudigger wrote:We already checked the codes many times. We also asked two programmers that did not work for us to verify if there was anything wrong it. No one found anything. We absolutely have no interest to have a system that will disadvantage you.


No doubt about that. Actually, all my experiences with Virtys have been positive :)

is it possible that the skill up chance displayed in game is simply too high? that would also explain the experiences we had.


That's a nice way to approach the issue. Never thought about that.

Coastis
Bourlingueur
Posts: 52
Joined: Mon 18 Jan 2016, 18:41
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Coastis » Wed 03 May 2017, 15:15

For every tale of woe, there is a tale of joy...

Image

....and I have many similar examples of this, such as taking just 5 volca blocks to skill up to 87. Of course, I have also experienced bad "luck" too.

That said, computer RNG is notoriously not random and features long strings of similar numbers, it is only when you average it out over tens of thousands of iterations do you get the true %.

Don't let RNG mess with your mind, it can be defeated!

Aeschylus
Citoyen
Posts: 40
Joined: Fri 29 Jan 2016, 20:12
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Aeschylus » Wed 03 May 2017, 18:38

Cyram wrote:1) failing salancedar cutting ten times in a row. At a presence of 32%, the chance for this to happen is about 2.1%
2) failing salancedar cutting 8 times in a row. The chance for this to happen is 4,6%
3) trying to level stonecutting from 86/86 to either 87/86 or 86/87. The game says the chance for a skill up cutting ornaments is 4,5%. I cut 53 volcanite blocks so far, bringing the probability for NO skillup down to 8,7%.

Well, the fact that RNGs are never really truly random (this is an unavoidable problem in all programming languages) aside, this is a good example of the gambler's fallacy (i.e. the logical problem of applying post-hoc probabilities to individual observations, #3 is a particularly good example). Basically the point is that strings of bad (and good) luck will occur that seem to defy probability, but aren't really all that surprising when you look at the individual chances.

So while you're right that it's not truly random, it's not really correct to say that there's something wrong with the RNG based on strings of hitting/missing on chances. Not that it doesn't suck, it's just that statistically speaking nothing you describe is out of the ordinary.

Cyram
Emperator
Posts: 41
Joined: Wed 06 Jan 2016, 16:02
Realm: Filgar

Re: Probabilities of skills rising

Postby Cyram » Wed 03 May 2017, 18:48

Ah, but that's just not true. The gambler's fallacy would be expecting higher chance to level on the third try because you missed two times.
I'm fully aware that the chance to skill up is 4.5% each and every time, regardless of what happened before.

There is nothing wrong, however, with calculating the probability a certain sequence of hits and misses will occur. In this case, the total chance of something like example 3 happening is - as given - 8%. As I stated, 8% is still easily within the realm of possibility. It's about the same as hitting heads (or tails, or really any 4 iterations preselected sequence of a coin toss) 4 times in a row. Of course this can and will happen. It is just not likely to happen. Now, if those unlikely sequences occur more often than they should, it does not hurt to point it out in case something is amiss.

Aeschylus
Citoyen
Posts: 40
Joined: Fri 29 Jan 2016, 20:12
Realm: Fauconnia

Re: Probabilities of skills rising

Postby Aeschylus » Wed 03 May 2017, 19:03

Fair, that's another way to interpret it. All I'm saying is that the way it's being approached in this thread is that anecdotal examples of bad and good luck in relatively small samples are being used as evidence of possible issues in the RNG, which I just don't see as being logically supportable. Yes, it's reasonable to calculate the probability of a given sequence happening, but frankly not on the sample level you're describing. If the overall distributions when you had sufficient statistical power to get a meaningful result were strange then yes I'd agree that something was wrong, but from what you describe you just can't draw that conclusion. With enough data you'd have the priors to fit a bayesian model to check the odds of the actual results happening given prior expectations, but I don't think anyone really has access to that data. :P

Cyram
Emperator
Posts: 41
Joined: Wed 06 Jan 2016, 16:02
Realm: Filgar

Re: Probabilities of skills rising

Postby Cyram » Wed 03 May 2017, 22:21

I quite enjoy this discussion :) it's not about drawing conclusions though, more like taking the data as a hint and inquiring if there could be more than random chance at work.

Anyway, the point about the amount of data is fair. Assuming you want to estimate the success chance within 1% and an alpha of 0.05 would require 6600 samples. I'm fairly sure I'm going to hit my level up sooner ;)

However, depending on how detailed the server logs the actions happening in game, that might actually be a worthwhile part of the "database" (or test files, whatever) to open up to toolmakers. I for one would love to dig through it.


Return to “Skills”

Who is online

Users browsing this forum: No registered users and 2 guests