Abnormal hit/miss sequences

[Yellow_Horror]

I know that i am not the first person saying this: something is wrong with Wesnoth RNG. While bulk hit/miss ratio seems correct, it generate too high number of several hit/miss sequences. For example, five day ago i see 6 sequental hits against elvish archer in the forest (probability <0.07%). As a single fact it is nothing, of course. But that's not a single fact. Three day ago i see that in the first mission of HttT campaign orcs and trolls without orcish assassins or slayers during one turn beat out 4 friendly elvish archers/rangers (all 4 in the forest) having only one miss in the process. I do not count exactly number of hits, but taking into account that even lvl 3 troll needs 2 hits to take out an elvish archer, that was at least 8 of 9 hit ratio against 70% defence (probability <0.007%). Yesterday i attack an orcish assassin in the forest with 4 elvish fighters and score only 1 hit from 16 strikes - the first strike hits then 15 sequental misses (probability <0.5%)

Today... I know that it is "normal" to see how a mage miss 4/4 (probability 0.81%) once in a while. But seeing this three times in a hour is abnormal. So, i quit play go to this forum and write all these letters just to calm out.

You can say that i count only "negative" examples (my or friendly units miss or taking hits). The main cause of this is that negative examples are easy to remember. Be sure, there is also a number of "abnormal positive" hit/miss sequences. But this only proves the main point: the Wesnoth RNG is "bad". Probably, from time to time it falls in a cycle and generate long sequence of almost equal numbers - this is the known flaw of some easy-and-fast RNG algorithms.

So, i beg the game developers to change RNG algorithm to a better one in the future versions of Wesnoth.

[Ken_Oh]

New Wesnoth acronym: Replay or Please Refrain from Speaking

[ancestral]

So, i beg the game developers to change RNG algorithm to a better one in the future versions of Wesnoth.

This topic has already been frequently and thoroughly discussed before. Please search and read through previous discussions.

[Gaiyamato]

While I have observed this myself (though not that many times in such a short period of time) I must say that with the current algorithm it should not happen.
Something else is going strange somewhere.

What OS are you running on?
What version of Wesnoth are you running?

I am collecting data and replays to try and make a solid case and hopefully present a solution (or at least a plausible argument for the reason) as well as the primary argument.
Do you have the saved replay?

[Cloud]

Today... I know that it is "normal" to see how a mage miss 4/4 (probability 0.81%) once in a while. But seeing this three times in a hour is abnormal. So, i quit play go to this forum and write all these letters just to calm out.

You can say that i count only "negative" examples (my or friendly units miss or taking hits). The main cause of this is that negative examples are easy to remember. Be sure, there is also a number of "abnormal positive" hit/miss sequences. But this only proves the main point: the Wesnoth RNG is "bad". Probably, from time to time it falls in a cycle and generate long sequence of almost equal numbers - this is the known flaw of some easy-and-fast RNG algorithms.

So, i beg the game developers to change RNG algorithm to a better one in the future versions of Wesnoth.

Three times in an hour is not, I would say, abnormal. Unlikely perhaps but abnormal no... just no. Sure it's a small chance of happening, but there's still that small chance.

If there is as many abnormally positive results as there are abnormally negative results (which is very [for want of a better, less pun-able word] probable) then it is balanced out anyways and you either accept that that's life or rage quit over some bad luck and never play again. Oh wait if there really are as many abnormal positives as there are abnormal negatives almost proves that the RNG is working well. If it was really producing far more of one abnormal as another, or if it wasn't producing abnormal results, then something would be wrong with it.

The quote "if you can't stand the heat, get out of the kitchen" comes to mind whenever someone whines about the RNG being unfair to them.

[Max]

I am collecting data and replays to try and make a solid case and hopefully present a solution (or at least a plausible argument for the reason) as well as the primary argument

wrong, that's cherry picking and wont "make a solid case".

[cretin]

mmm rng is fine, ive gone to miss 6 or 7 times in a row. ive also gone and hit 6 or 7 times in a row too, repeating of courseheh wonder if anyone knows what this is from . people just remember when they miss, or when the enemy always hits.

[Tonepoet]

Today... I know that it is "normal" to see how a mage miss 4/4 (probability 0.81%) once in a while. But seeing this three times in a hour is abnormal. So, i quit play go to this forum and write all these letters just to calm out.

You can say that i count only "negative" examples (my or friendly units miss or taking hits). The main cause of this is that negative examples are easy to remember. Be sure, there is also a number of "abnormal positive" hit/miss sequences. But this only proves the main point: the Wesnoth RNG is "bad". Probably, from time to time it falls in a cycle and generate long sequence of almost equal numbers - this is the known flaw of some easy-and-fast RNG algorithms.

So, i beg the game developers to change RNG algorithm to a better one in the future versions of Wesnoth.

Three times in an hour is not, I would say, abnormal. Unlikely perhaps but abnormal no... just no. Sure it's a small chance of happening, but there's still that small chance.

If there is as many abnormally positive results as there are abnormally negative results (which is very [for want of a better, less pun-able word] probable) then it is balanced out anyways and you either accept that that's life or rage quit over some bad luck and never play again. Oh wait if there really are as many abnormal positives as there are abnormal negatives almost proves that the RNG is working well. If it was really producing far more of one abnormal as another, or if it wasn't producing abnormal results, then something would be wrong with it.

The quote "if you can't stand the heat, get out of the kitchen" comes to mind whenever someone whines about the RNG being unfair to them.

Actually what matters isn't if we get an even number of positive/negative sequences because what's positive and negative is dependent on PoV. One side's good luck is another side's bad luck

What matters is that we get what's written on the box. It's not necessarily wrong to get combinations of 0.81% but it should only happen approximately 0.81% of the time, otherwise it's bugged and should be dealt with as such. If we get streaky and/or predictable hit sequences, that suggests that the odds may be skewed incorrectly.

In prior RNG discussions I've seen, we've never had a large pool of indiscriminate replays to check things over with and make sure that the odds were really congruent with what they should be. Now with the Public Replay Archive, we can potentially check all observable replays dating back to april if we want. It's only a matter of figuring out a method to see if there are oddsome tendencies which skew the odds out of favor. I think nothing but benefit could come to the Wesnoth community by checking, because we can either seek and destroy a potentially serious problem with the game or come up with solid evidence that there's absolutely nothing wrong with the game. The latter helps us because if can provide such evidence that nothing's wrong with a demonstrable method, it very well just might help to quell these repeated RNG related complaints and inquiries; especially if we provided a how-to sticky. I believe the main question would be how to do it; I'm a bit clueless myself. I'd be a bit surprised if there wasn't a statistical projection vs. results comparison algorithm we could use to double check somewhere though.

[Sauron]

Your problem is so reoccurring and old for Wesnoth, it got some solutions. Refer to my signature to get multiplayer solution. In the future maybe there will be also campaign patch.
I would like everyone aware of my solution to inform about it in such threads. Thank you in advance.

[HomerJ]

I think maybe some people go into this thread biased by the numerous rng-rant threads there have been already.
As I understand the OP, he simply states that if an event (no matter if positive or negative) is supposed to happen with 0.1 % propability, but happens with 10 % propability, there might be something wrong with the algorythm.
So, balancing out is not what this is about.
I totally support this statement, the only problem is proving it as long as nobody takes the effort to script something that analyses replays automatically in this context.
From my (limited) experince with propability, 10000 events (event would be 1 strike hit or miss) should give a first impression, but of course the more the more accurate.
Please kindly note that this is not the same thing as the numbers in the statistics panel (expected damage vs actual damage) because of the balancing out issue.

Greetz
HomerJ

[Sauron]

HomerJ, no RNG can change here a thing (despite in fact Wesnoth uses pretty outdated RNG, it was already discussed). Even if you used your fireplace as source of entropy the fight mechanics would remain prone to described situations.

[Soliton]

Your problem is so reoccurring and old for Wesnoth, it got some solutions. Refer to my signature to get multiplayer solution.

Even if you used your fireplace as source of entropy the fight mechanics would remain prone to described situations.

Could you please quote where in this thread the fight mechanics are discussed so that one can see how your contributions to this thread are any relevant? As the author of your mod I'm sure you're aware that it does not implement a new RNG for wesnoth or changes the current implementation.

[Sauron]

Your problem is so reoccurring and old for Wesnoth, it got some solutions. Refer to my signature to get multiplayer solution.

Even if you used your fireplace as source of entropy the fight mechanics would remain prone to described situations.

Could you please quote where in this thread the fight mechanics are discussed so that one can see how your contributions to this thread are any relevant? As the author of your mod I'm sure you're aware that it does not implement a new RNG for wesnoth or changes the current implementation.

As author of the mod and a person that worked a bit with various RNGs, including Mersenne Twister (It happens so that I implemented one in Java) I am well aware of the fact that the observed phenomena cannot be contributed exclusively to RNG. If you flip a simmetric coin 10000 times and get 100x heads in a row this does not prove the coin is malfunctioning or there is a flaw in the flipping process. The series required to generate reported 'bad luck' require just 3-4 high/low %100 numbers in a row. Thing that is bound to happen. I am referring to this:

Be sure, there is also a number of "abnormal positive" hit/miss sequences. But this only proves the main point: the Wesnoth RNG is "bad". Probably, from time to time it falls in a cycle and generate long sequence of almost equal numbers - this is the known flaw of some easy-and-fast RNG algorithms.

as sequences of hi-low numbers must appear, yet it is the fight mechanics that makes natural and expected in uniform distribution sequences "abnormally" positive/negative. The picture was to illustrate just that.

[Yellow_Horror]

Thank you all for attention. There are my answers for some questions in this topic:

1) I do not collect replays, especially of the games i lose, so i can't prove my experience. Furthermore, i found it pointless if you disbelief my feelings in the first place. Proving that a given RNG is "bad" is a solid mathematical task that can't be solved just by presenting some "strange" examples. But, AFAIK, one of the simplest ways to check a given RNG is to use human feelings against it. For example, filling "1x1 square" (or cube) with "random-coordinated" dots under visual control reveals some RNG weakness much quicker than mathematical search of self-correlations in the "random numbers" sequence.

2) I use Wesnoth 1.6.3 for Windows. I read some random-related threads and found that this (and all preceding) version of Wesnoth use "standard" RNG that can be implemented differently in each executable compilation. IMHO, this fact is bad enough by itself. So i repeat my appeal to the game developers:
please change the Wesnoth RNG in the future versions. I'm sure that there is free RNG algorithms that are checked against too systematical results. Implementing one of them in the game code will at least "equalize" experience of the game users under all OS.

[Sapient]

please change the Wesnoth RNG in the future versions. I'm sure that there is free RNG algorithms that are checked against too systematical results. Implementing one of them in the game code will at least "equalize" experience of the game users under all OS.

You obviously did not take the advice to educate yourself by reading prior discussions which refute this claim.

Therefore, I am locking this pointless thread.