|
member
|
member
Joined: Jul 2023
|
It’s also worth remembering that there is a particular type of selection bias in games like this. If you have a less than 50% chance to complete a roll, you probably won’t even bother. You would almost certainly succeed on as many 10% chances as you fail on 90% chances, but you never attempt a 10% chance success action, so you only see failures on the 90% successes.
Not saying that the RNG definitely isn’t broken, but I see this exact same topic in every game that uses RNG, and it is never the game. It is just that some people getting unlucky is a statistical inevitability. I remember the XCOM devs saying they tweaked the percentages shown to players from the true stats because humans have terrible comprehension of probability.
|
|
|
|
stranger
|
OP
stranger
Joined: Aug 2023
|
I've looked into the combat log to find out about the actual rolls, but while I could find the damage rolls, I couldn't find the to-hit rolls. Aren't they shown somewhere? Both the attack and dmg rolls are listed in the combat log. Generally, there are two sentence - the first one says whether you hit or miss. If you mouse over, it tells you the Armour Class of the character being attacked, and it also states your modifiers of your Attack Rolls and what your total Attack Roll is. The second one says how much dmg you have dealt. If you mouse over, it tells you your Dmg Roll.
|
|
|
|
stranger
|
OP
stranger
Joined: Aug 2023
|
I'm a big believer in stats and human biases, but for me the rolls in BG3 definitely feel bad. There have been countless times where I've rolled multiple rolls in a row under 4, often with advantage. Of course, this is likely just be confirmation/negativity bias, where I'm only noticing the negative events and not the times I've rolled multiple >16s in a row. However, it could also be a real problem with Larian's rng.
In order to test the actual randomness of a d20, you need >500 rolls. >1000 is preferable. If someone(s) wants to sit down and record their rolls, I'll happily perform statistical analysis on them. If a few people each do 200 rolls then those data sets can be combined. The most useful dataset is a continuous string of rolls (don't reload the game) where each roll is recorded in order of appearance, ideally all against the same or similar target DCs/ACs. And of course, note whether you're using Karmic Dice or not. I guess there's also the question of....how random is something alleged to be random. In theory, nothing is truely random. Even a traditional physical dice roll is subject to how a player holds and tosses the dice, and its orientation prior to being thrown. In theory, if a player say.....likes seeing the number 20 face up, and throws the die a certain way, the roll then becomes 'more' biased, and not perfectly random. The term 'random' is merely assigned by us humans to explain an outcome that we personally are unable to predict. In the case of electronics and computers, I guess the same question can be asked? Is the computer truly returning a random result? But i get what you mean. The unbiased nature (assuming it really is the case) of a die can only be ascertained over a large sample. It's entirely possible for initial results to be skewed to one side of the mean, but if this does happen, it takes a large sample for the opposite to happen, eventually exhibiting the expected normal distribution curve.
Last edited by MetricTensor; 10/08/23 02:11 AM.
|
|
|
|
old hand
|
old hand
Joined: Sep 2020
|
It might not even be the rolls, but other hidden mechanics that aren't displayed.
We could know for certain if Larian would just add the option to view combat rolls and see if I really was rolling 10 or less 20 times in a row on a 50% chance.
Last edited by DumbleDorf; 10/08/23 01:27 PM.
|
|
|
|
veteran
|
veteran
Joined: Sep 2020
|
It might not even be the rolls, but other hidden mechanics that aren't displayed.
We could know for certain if Larian would just add the option to view combat rolls and see if I really was rolling 10 or less 20 times in a row on a 50% chance. You can see the rolls if you expand the combat log and hover over the hit/damage/miss text. It shows what number you rolled on the d20, all the bonuses, and the target AC/DC. Or do you mean something else?
|
|
|
|
veteran
|
veteran
Joined: Dec 2020
|
I turned on karmic dice yesterday and played a bunch (too much :D) and it seemed better, not nearly as many nasty streaks. I didn't notice it working against me either, although in Act 3 now, everything is fairly high level, so many the karmic dice are working in my favor. I'm only level 9, and a lot of boss type encounters seem to be 10/11 with minions between 8 - 9.
|
|
|
|
enthusiast
|
enthusiast
Joined: Mar 2022
|
RNG in this game is broken. No karmic dice sleight of hand Check on shadow heart DC 30 stealing 5000 gold. I did 50 reloads- expected probability dictates at 40 roles to hit 1 nat 20. 50 is 10% above reasonable sample …. Not once did I pass this check. My results where all 10-17 with rare cases of 2-8 and 6,19. Most common number was 11,12 accounting for 15 of 50 entries, the rest being 2-3 repeats of 13,10. My sample size shows rng is not working correctly
|
|
|
|
enthusiast
|
enthusiast
Joined: Jun 2020
|
That’s XCom, baby!
Oh, sorry, wrong forum.
|
|
|
|
veteran
|
veteran
Joined: Jan 2018
|
RNG in this game is broken. No karmic dice sleight of hand Check on shadow heart DC 30 stealing 5000 gold. I did 50 reloads- expected probability dictates at 40 roles to hit 1 nat 20. 50 is 10% above reasonable sample …. Not once did I pass this check. My results where all 10-17 with rare cases of 2-8 and 6,19. Most common number was 11,12 accounting for 15 of 50 entries, the rest being 2-3 repeats of 13,10. My sample size shows rng is not working correctly No, that actually sounds perfectly reasonable.
|
|
|
|
veteran
|
veteran
Joined: Dec 2020
|
RNG in this game is broken. No karmic dice sleight of hand Check on shadow heart DC 30 stealing 5000 gold. I did 50 reloads- expected probability dictates at 40 roles to hit 1 nat 20. 50 is 10% above reasonable sample …. Not once did I pass this check. My results where all 10-17 with rare cases of 2-8 and 6,19. Most common number was 11,12 accounting for 15 of 50 entries, the rest being 2-3 repeats of 13,10. My sample size shows rng is not working correctly On that specific conversation, if SH likes you, you don't even have to roll a persuasion check, but it's not at all clear what dialogue options to pick because of how SH has been acting the entire dungeon.
|
|
|
|
addict
|
addict
Joined: Oct 2020
|
RNG in this game is broken. No karmic dice sleight of hand Check on shadow heart DC 30 stealing 5000 gold. I did 50 reloads- expected probability dictates at 40 roles to hit 1 nat 20. 50 is 10% above reasonable sample …. Not once did I pass this check. My results where all 10-17 with rare cases of 2-8 and 6,19. Most common number was 11,12 accounting for 15 of 50 entries, the rest being 2-3 repeats of 13,10. My sample size shows rng is not working correctly 0.95^50 = 0.077 Its *likely* that 50 rolls will get you a success, but there is a 7.7% chance of going 50 without a success. I don't think that this proves that the dice are broken.
|
|
|
|
veteran
|
veteran
Joined: Jan 2018
|
RNG in this game is broken. No karmic dice sleight of hand Check on shadow heart DC 30 stealing 5000 gold. I did 50 reloads- expected probability dictates at 40 roles to hit 1 nat 20. 50 is 10% above reasonable sample …. Not once did I pass this check. My results where all 10-17 with rare cases of 2-8 and 6,19. Most common number was 11,12 accounting for 15 of 50 entries, the rest being 2-3 repeats of 13,10. My sample size shows rng is not working correctly 0.95^50 = 0.077 Its *likely* that 50 rolls will get you a success, but there is a 7.7% chance of going 50 without a success. I don't think that this proves that the dice are broken. And 50 is not a reasonable sample size. 200 rolls is okay. 300 is better.
|
|
|
|
journeyman
|
journeyman
Joined: Nov 2015
|
I recorded 78 d20 rolls as I was playing the game normally, then I gave up. =P (Keeping track of combat rolls is too annoying; you have to hover over everything in the combat log, some rolls can't be recorded because they don't show up at all, and you need to disregard advantaged/disadvantaged rolls, 'cuz those only report the final result.) My average roll came up to be 9.2, which seems reasonably average for that sample size.
When it comes to determining if 20 happens as often as it should, correct me if I'm wrong, but wouldn't that require a pretty monstrous sample size? E.g. if we decide 200 rolls is an ok sample size for determining if the rolls are skewed one way or the other, wouldn't it require 20 times that many, i.e. 4000 rolls?
|
|
|
|
stranger
|
stranger
Joined: Jun 2019
|
I agree, the dice are .. busted. I had to make a DC roll 10 and then a 15. On the 10 i got a one, rerolled, got a 1, rerolled, got a 7. i loaded that save 3 times, they were all single digits on trying to roll a 10. I rolled 3 1's and 4 2's i literally thought it was impossible to beat until i checked online. i finally got past the 10 and actually had way more luck with the 15, despite them using the same modifier. Something is wrong with the numbers, the number of times 95% hit has missed is... alarming
|
|
|
|
old hand
|
old hand
Joined: Jul 2023
|
I'm still in Act 1 but I turned off KD early on. At first it didn't seem to make any difference - you seem to get a session where nobody can hit anything then your next session and nobody can miss. Fingers crossed and touch wood, it seems to have settled down to hitting and missing frequency being 'acceptable'. I only hit level 4 this evening so level shouldn't have made much difference and we've had +1 weapons since early doors. I'm not 100% convinced about non-combat rolls though.
I think the percentage to hit figures you get when you aim at an enemy are suspect. I don't bother much with the actual figure, I just go for the highest, whether Target A is 25% and Target B is 30% or Target A is 80% and Target B is 85%.
Last edited by Beechams; 12/08/23 11:23 PM.
|
|
|
|
enthusiast
|
enthusiast
Joined: Apr 2023
|
It’s just probability. And you pay more attention to misses than hits, so it registers in your memory more.
It will even out. You’ll have your perfectly times crits and hits, do not worry.
|
|
|
|
old hand
|
old hand
Joined: Jul 2023
|
It’s just probability. And you pay more attention to misses than hits, so it registers in your memory more.
It will even out. You’ll have your perfectly times crits and hits, do not worry. Whether it evens out over ten rolls or over a thousand are two different things from a gaming perspective. The fact remains that for me this is the first RPG where the dice rolls have been an issue. Thinking about it, this is probably the only game I've played that has two different dice modes which must tell us something.
Last edited by Beechams; 14/08/23 01:12 PM.
|
|
|
|
veteran
|
veteran
Joined: Jan 2018
|
My experience with the dice rolls in combat and dialogue have been fine. Nothing at all unusual.
|
|
|
|
apprentice
|
apprentice
Joined: Mar 2021
|
It’s just probability. And you pay more attention to misses than hits, so it registers in your memory more.
It will even out. You’ll have your perfectly times crits and hits, do not worry. This is not true. You, Me and everyone else COMBINED has a near perfect statical distribution. You alone, Me Alone, Anyone Alone does not have PERFECT statical distributed dice rolls. This is the entire reason we need karmic dice.
|
|
|
|
apprentice
|
apprentice
Joined: Jul 2023
|
My next character is going to be named Unlucky Scumsave.
The dice hate me.
Not so much since I turned off karmic dice.
|
|
|
|
|