Larian Banner
Previous Thread
Next Thread
Print Thread
Page 4 of 5 1 2 3 4 5
Joined: Sep 2020
addict
Offline
addict
Joined: Sep 2020
Originally Posted by Niara
... https://docs.google.com/spreadsheets/d/1tdyBoQNS_vwEGZGBgFRQex7b-Ma8S6P7zvEMK5wh9n4/edit?usp=sharing

I laid out my old numbers, and took 200 fresh numbers for BG3 using a character that wasn't a halfling (my previous rolls for BG3 were with a halfling and that did impact things; these numbers are fresh).

The left columns are the raw numbers as rolled, in the order they were rolled, for each game.
Beside that, is a small table showing the average result, the median result, and the roll count for each result, for each game.

Below that, are four charts that show a plotting of the results if mapped in the order they were rolled.
The three lowest charts, NWN, NWN2 and S:CotM, are almost indistinguishable from each other. The top chart is BG3. You can see the wave pattern.
Huh, that is indeed a wave pattern.
That's really bad that Larian didn't catch this. Like, really bad. If this holds true for all rolls (and isn't just local weirdness in the rng that you happened to catch), this basically explains all of Larian's opinions that "normal rolling" is not fun. Because they weren't ever rolling normally, they were getting frequent streaks of multiple misses.

Based off of your data:
a roll of 1-5 is followed by an average roll of 9.3
a roll of 6-10 is followed by an average roll of 10.2
a roll of 11-15 is followed by an average roll of 11.3
a roll of 16-20 is followed by an average roll of 11.8

Disclaimer: all the above averages have large errors

Last edited by mrfuji3; 23/02/21 05:53 AM.
Joined: Jun 2020
Niara Offline OP
addict
OP Offline
addict
Joined: Jun 2020
As with all random, pseudo-random and statistical analysis of this nature.... What I've got here could very easily be offset by someone else testing the same thing and getting different results, but these are mine. It absolutely could be a local weirdness of some sort, and I can't prove that it isn't. We can't see under the hood, so it will never be possible to say for certain. I am almost certain that other players will see this and say "But I never had streaks or swings like that", and they most likely will not be lying.

Joined: Oct 2020
enthusiast
Offline
enthusiast
Joined: Oct 2020
Originally Posted by Niara
As with all random, pseudo-random and statistical analysis of this nature.... What I've got here could very easily be offset by someone else testing the same thing and getting different results, but these are mine. It absolutely could be a local weirdness of some sort, and I can't prove that it isn't. We can't see under the hood, so it will never be possible to say for certain. I am almost certain that other players will see this and say "But I never had streaks or swings like that", and they most likely will not be lying.

If only there was some kind of two way communication here so we could know that Larian has been made aware of this. Fixing this is far more important than loading the dice.

Joined: Nov 2010
B
journeyman
Offline
journeyman
B
Joined: Nov 2010
That does indeed look like a pattern by a really bad RNG and seriously needs to get fixed. Good work finding out!

Joined: Jun 2020
Niara Offline OP
addict
OP Offline
addict
Joined: Jun 2020
Thank you, and I apologise for snapping at you with my response. I had a moment of frustration.

Joined: Nov 2020
B
Banned
Offline
Banned
B
Joined: Nov 2020
Great job! While its nothing that Proves something 100% it is still hinting at a underlying problem here. However.... since Larian gives Shit about community its sad that we wont see a real reaction.
But this also shows that LARIAN with their "Loaded Dice" arent even listening to people and instead always do their own weird thing. frown

Joined: Sep 2020
addict
Offline
addict
Joined: Sep 2020
Originally Posted by Niara
As with all random, pseudo-random and statistical analysis of this nature.... What I've got here could very easily be offset by someone else testing the same thing and getting different results, but these are mine. It absolutely could be a local weirdness of some sort, and I can't prove that it isn't. We can't see under the hood, so it will never be possible to say for certain. I am almost certain that other players will see this and say "But I never had streaks or swings like that", and they most likely will not be lying.
Have you summitted these results as a bug report? IIRC, you can attach an image to bug reports.

It probably has a higher likelihood of being seen by Larian than, uh...buried 4 pages in a Panel from Hell II Synopsis Forum post. Lol we're a bit off topic.

Joined: Oct 2020
D
member
Offline
member
D
Joined: Oct 2020
Originally Posted by Niara
I haven't, no, sorry. Doing these tests has pretty much eaten up most of my stamina for die-rolling statistics... Even this post was mostly made out of frustration at being told I was making it up...

I'd generally expect simpler systems to use simpler, free to acquire RNGS, while systems for whom that d20 roll is the backbone of their game to fork out for more advanced ones that aren't free.

I should probably request that this all be branched off into its own thread, however, since this thread is meant to be for the synopsis of the stream, and for discussion of the stream, and potentially the contents of the upcoming patch... not another debate about RNGs and dice...

Thanks for taking the time to do this. I'm actually still skeptical, but I do appreciate the effort involved here.

Note that my skepticism is based on personal experience... as a graduate student ~25 years ago I noticed a similar pattern in some simulations that I was running for my thesis, and I had an absolute panic attack over them. My advisor said "run more simulations" and as more were added the weird sine pattern went away.

In any case, you might be on to something. I intend later to do some additional analysis on your data if I can find the time. I may also try to collect some more rng data to supplement this set.

Joined: Sep 2020
Banned
Offline
Banned
Joined: Sep 2020
This is probably the most important thread on these forums as it demonstrates the inability of Larian to do anything other than what they know, including being bullheaded about bad designs, and also their lack of communication with the dedicated community.

Joined: Dec 2020
R
member
Offline
member
R
Joined: Dec 2020
Is there a quick way to get the game to roll a bunch of d20s and export the values?

Joined: Oct 2020
D
member
Offline
member
D
Joined: Oct 2020
Originally Posted by RBarbare
Is there a quick way to get the game to roll a bunch of d20s and export the values?

I could be wrong, but I *think* Niara reloaded and rerolled each time, and wrote the value down each time. This is pretty tedious, and that's why there are only 200 entries.

Joined: Oct 2020
D
member
Offline
member
D
Joined: Oct 2020
A log file that dumps all of the rolls would be really awesome, since we could investigate much larger samples in that case.

Joined: Oct 2020
enthusiast
Offline
enthusiast
Joined: Oct 2020
Originally Posted by Niara
He also frequently experiences the well known and much reported character selection bug where you need to click multiple times to get it to select. He blames it on his mouse, but we all know better. Why doesn't he?

The issue with UI cursor/click responsiveness is pretty major, since it seems to show up just about everywhere.

This is usually the one that breaks me and has me quitting the game after like an hour.

If it was just for character selection that'd be one thing, but its pretty pervasive and touches nearly every part of the game. I've only seen one thread where it was brought up and highlighted as one of The main problems with the BG3 UI (something that makes every other UI problem worse) here...
https://forums.larian.com/ubbthreads.php?ubb=showflat&Number=752140#Post752140

People have described any number of things about the UI as feeling 'clunky' but I agree with OP of that thread, that the main clunk factor seems to be coming from this.

Joined: Dec 2020
Location: CA
S
addict
Offline
addict
S
Joined: Dec 2020
Location: CA
Originally Posted by Black_Elk
Originally Posted by Niara
He also frequently experiences the well known and much reported character selection bug where you need to click multiple times to get it to select. He blames it on his mouse, but we all know better. Why doesn't he?

The issue with UI cursor/click responsiveness is pretty major, since it seems to show up just about everywhere.

This is usually the one that breaks me and has me quitting the game after like an hour.

If it was just for character selection that'd be one thing, but its pretty pervasive and touches nearly every part of the game. I've only seen one thread where it was brought up and highlighted as one of The main problems with the BG3 UI (something that makes every other UI problem worse) here...
https://forums.larian.com/ubbthreads.php?ubb=showflat&Number=752140#Post752140

People have described any number of things about the UI as feeling 'clunky' but I agree with OP of that thread, that the main clunk factor seems to be coming from this.

Agreed. It’s incredibly frustrating. But hey the cinematics are pretty...

Joined: Jan 2020
enthusiast
Offline
enthusiast
Joined: Jan 2020
Originally Posted by dwig
Originally Posted by RBarbare
Is there a quick way to get the game to roll a bunch of d20s and export the values?

I could be wrong, but I *think* Niara reloaded and rerolled each time, and wrote the value down each time. This is pretty tedious, and that's why there are only 200 entries.

Well, if the numbers come from reloading, then the analysis might not stand up to scrutiny, as you may be resetting the RNG each time, and there is no guarantee that RNG seeds in a random way.

You need to take a sequence within the context of the game, which may well be awkward if looking for patterns, as you do not know when the RNG is used by the game without you seeing the results.

Having said that, any arbitrarily sampled set of rolls within the game should exhibit randomness.

Joined: Oct 2020
D
member
Offline
member
D
Joined: Oct 2020
Originally Posted by etonbears
Originally Posted by dwig
Originally Posted by RBarbare
Is there a quick way to get the game to roll a bunch of d20s and export the values?

I could be wrong, but I *think* Niara reloaded and rerolled each time, and wrote the value down each time. This is pretty tedious, and that's why there are only 200 entries.

Well, if the numbers come from reloading, then the analysis might not stand up to scrutiny, as you may be resetting the RNG each time, and there is no guarantee that RNG seeds in a random way.

You need to take a sequence within the context of the game, which may well be awkward if looking for patterns, as you do not know when the RNG is used by the game without you seeing the results.

Having said that, any arbitrarily sampled set of rolls within the game should exhibit randomness.

I'll let Niara speak for their analysis, because I might be wrong about the way they did it. My *guess* is that this is not a problem with the analysis (but to be clear, I really am guessing here). An exportable log of all rolls would be much better though.

Joined: Jun 2020
Niara Offline OP
addict
OP Offline
addict
Joined: Jun 2020
Originally Posted by dwig
Note that my skepticism is based on personal experience... as a graduate student ~25 years ago I noticed a similar pattern in some simulations that I was running for my thesis, and I had an absolute panic attack over them. My advisor said "run more simulations" and as more were added the weird sine pattern went away.

Yeah, that's what I was getting at with some of my other comment.

Even a 200 count exposing a wave like that is worrisome, because it does seem to be exposing *something* of the algorithm behind the generation... (and seeming even to swing reliably up and down through a 'full' cycle every 50 rolls or so...) but it is entirely possible that as you carry the sample into the thousands and beyond, that distinction will dissipate. I don't know.

So people know, the number gathering I conducted was done using the goblin courtyard and sanctum; I tried to gather the results in one contiguous string, actively avoiding advantage and disadvantage (since the combat log doesn't show both rolls), but sourcing the rolls in order, and thus including enemy rolls as well as my own characters'.

But as pointed out, there's a lot we don't know, and can't know - we don't know what causes reseeds on the RNG, or whether it's using independent generation seeds and methods for enemies and players, just for example.

The results are disturbing because of how clear they show up, compared to how indistinguishable the other sets from other games are, but we can't know what is causing what I'm seeing, or why.

I have indeed submitted this as a bug report to Larian's bug reporting form.

Joined: Mar 2018
Location: Birmingham
journeyman
Offline
journeyman
Joined: Mar 2018
Location: Birmingham
If it wasn't so late at night I would be highly impressed as that's evidence right there..But on the other hand it is a PC game, RNG is often not so random as we would like no matter it's expense or developers..Me myself despite being a bit of a moaning Myrtle sometimes would prefer it if we all waited a while and see what the end product is like, then perhaps make our judgements based on that. I want them to get it out myself so I can then shout at my screen such things as "It's fixed, the f'in games fixed", "WHAT! That's a glitch", "Are you kidding me that guy had six turns, how come I don't get to use four special moves in a row", "AHH! Faaaark so it's innitiative is over 23 this level then huh? Goddamn it" amongst other favourites when I wipe. Being perfectly flawed is Larian games in so many ways, you will get upset, you will get angry, you will moan like hell as things won't go your way a lot of the time, storylines won't go the way you wanted, spells will fail, you'll scum save 19 times to get the roll you wanted, get blinded when you have a helm that makes you immune to blinded, traders will continually sell trash for the class you aren't playing at the time, then when you do everything you wanted on the last run will drop..Why was WoW so successful in it's day and even doing OK today? Because it never gave you what you wanted is the quick answer, unless of course you were living in Norway where theres a person per 20 miles or something and welfare benefits pay you as much as working so you could sit playing it all day, my point is if you got everything youw anted from a game you would be terribly bored of it within five minutes.

Last edited by Seleniumcodec; 24/02/21 01:33 AM.

But as they parted ways, saviour the second merely clapped saviour the first on the back and rode away...
Joined: Sep 2020
addict
Offline
addict
Joined: Sep 2020
Originally Posted by Seleniumcodec
If it wasn't so late at night I would be highly impressed as that's evidence right there..But on the other hand it is a PC game, RNG is often not so random as we would like no matter it's expense or developers..Me myself despite being a bit of a moaning Myrtle sometimes would prefer it if we all waited a while and see what the end product is like, then perhaps make our judgements based on that. I want them to get it out myself so I can then shout at my screen such things as "It's fixed, the f'in games fixed", "WHAT! That's a glitch", "Are you kidding me that guy had six turns, how come I don't get to use four special moves in a row", "AHH! Faaaark so it's innitiative is over 23 this level then huh? Goddamn it" amongst other favourites when I wipe. Being perfectly flawed is Larian games in so many ways, you will get upset, you will get angry, you will moan like hell as things won't go your way a lot of the time, storylines won't go the way you wanted, spells will fail, you'll scum save 19 times to get the roll you wanted, get blinded when you have a helm that makes you immune to blinded, traders will continually sell trash for the class you aren't playing at the time, then when you do everything you wanted on the last run will drop..Why was WoW so successful in it's day and even doing OK today? Because it never gave you what you wanted is the quick answer, unless of course you were living in Norway where theres a person per 20 miles or something and welfare benefits pay you as much as working so you could sit playing it all day, my point is if you got everything youw anted from a game you would be terribly bored of it within five minutes.
...what?
1.) The game is in Early Access now, a time where Larian has explicitly said they want to work "directly with players" and for people to "participate in community feedback." Thus, we should provide feedback now and not wait until the game leaves Early Access.
2.) Um, no? I don't want Larian to release a broken game full of glitches, "guys with 6 turns," useless traders, incorrect item descriptions, etc. That sounds like an absolutely terrible game.
3.) I gotta say, it's a bit weird that you brought Norway and welfare benefits into this o.O

Joined: Oct 2020
enthusiast
Offline
enthusiast
Joined: Oct 2020
Years in development, 4 Months into EA, TONS of highly detailed feedback... the UI SHOULD BE BLOODY PERFECT by now. NOTHING HAS CHANGED. You don't work on UI changes couple months prior to release.
This alone is proof we shouldn't even bother to give feedback, whats the point anymore.

Last edited by mr_planescapist; 24/02/21 09:49 AM.
Page 4 of 5 1 2 3 4 5

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.5