Are the advantage/disadvantage(s) working properly?

[WereElf]

I've seen 2 Critical Misses happen in a row while having an advantage. While technically not impossible, the chances for that to happen are 1 in 160 000. But it's not just that. If you have a 40% chance to hit an enemy, and it's increased to 64% while you have an advantage, observing many hits in a row, while having the advantage, you'll notice that your hit rate is about 40%, not 64%.
Also, when you have an advantage on your stat rows, you still only row once, not twice. So it's either not implemented properly yet, either it's hidden (for the stat rolls).

[Chillearth]

There is a problem with the BG3 RNG. It does need tweaking and they have already messed with it in the options, but it needs a proper fix.

[id795078477]

Originally posted by WereElf:

I've seen 2 Critical Misses happen in a row while having an advantage. While technically not impossible, the chances for that to happen are 1 in 160 000. But it's not just that. If you have a 40% chance to hit an enemy, and it's increased to 64% while you have an advantage, observing many hits in a row, while having the advantage, you'll notice that your hit rate is about 40%, not 64%.
Also, when you have an advantage on your stat rows, you still only row once, not twice. So it's either not implemented properly yet, either it's hidden (for the stat rolls).

Problem is - you only take into account your playthrough. If you start thinking about other players and how many times they rolled that dice, with the 1/160.000 chance it doesn't look unlikely that someone got that situation. It just happened that "someone" is you.

Now, if you see this consistently through your game - then it's a fair statement. But in my 55 hours I never had this case. Again - that's also not telling anything. Point is - to try grasping the bigger picture and judge based on that.

[theslay66]

Well, you won the lottery. Congratulations.
Even when there is 1 in a million chance to win a lottery, if 10 millions tickets are sold, 10 of them (on average) are winning tickets. Although winning is improbable, it will happen to somebody. You just happen to be one of the winners.

As for the supposed problems with RNG in BG3, it's in fact a matter of perception. Our brain is very bad at handling true RNG, and will see streaks as strange, whereas they will happen in a true RNG system (and I say "true RNG" as in "unpredictable for the user", not in a mathematical or physical sense).
They are in fact, trying to reduce these streaks that appear naturally in a true RNG system, by taking into account previous rolls to modify the odds of the next roll. But it is quite hard to balance to a satifactory level for everyone, because this is a matter of perception and so, is in fact completely subjective, and perception of this problem will vary from player to player.

And that is the problem with systems entirely based around RNG. The simplest way to make it is just to have pure RNG, where each roll is completely independant of any previous roll. But it is also the most frustrating for the player, who feels he has no control over the game and will win or loose depending entirely on a whim of fate. Trying to "correct" these naturally occuring streaks is in fact a way to give back a bit of control and agency to the player.

[wirednight]

So there are two sides to this story.

One side is going to say "Sorry it's RNG and too bad" then go on a diatribe about how RNG needs to stay, it's all perception, long term math, stats, blah, blah, blah and you're just unlucky, suck it up.

The other side is going to say that the RNG is broken, that they can prove that it's broken from a small sample set, blah, blah, blah, Larian needs to fix it.

Truth is, Larian saw a problem and tried to fix it with the loaded dice in options they created, Whether the issue was perception wise or a problem with their RNG mechanic, and used the loaded dice option to correct the issue. Not sure how well it worked, that's not up to me to decide.

Either way, the defenders of RNG will defend it to their dying breath, regardless if it's broken or not, and justify that you're just an unlucky slob who pissed off RNGesus. Somewhere in their they'll toss a remark that DnD 5e is this way, and it is sacred and can't be touched.

Conversly the haters of RNG will attack it to their dying breath, regardless if it's broken or not. Somewhere in there they'll toss a remark that in DnD 5e tabletop, the DM has control over the dice rolls, and the computer game isn't going justice to the spirit of a tabletop rpg's DM.

Larian, deciding to do a game heavily based on RNG, so heavily in fact that you can barely sneeze without having to make a roll. Personally I think the RNG is perfectly fine as a mechanic. I think Larian's target numbers are way off. So it's not the advantage or disadvantage that's broken, it's that Larian decided to do consistant target numbers in the 15 target range. Which would be fine, if your companions could jump in and help out with the target numbers by giving using the Help mechanics from the rule book. Those don't exist.

[dolby]

does it really matter? so you missed twice buhoo unlucky hnext time you gonna get lucky and get 3 cirts in a row and i bet you are not gonna come and make a post about it...
the whole point of dice in dnd is that so just move on...

[wardibald]

RNG's have been working just fine for decades and are standard included in all half baked programming languages. I don't think Larian even built its own.

[WereElf]

Well, I did a little experiment on the Cambions in the tutorial area, and here are the results:
40% chance to hit (no advantage): 18 attacks - 4 hits, 1 crit, 11 misses, 2 critical misses.
Or 27,8% hit rate, with 5,55% crit rate and 11,11% critical miss rate.
51% chance to hit (30% with advantage): 88 attacks - 34 hits, 5 crits, 48 misses, 1 critical miss.
Or 44,3% hit rate (which is close to the 43,75% hit chance of a 25% hit attack with an advantage), with 5,68% crit rate and 1,13% critical miss rate.
64% chance to hit (40% with advantage): 74 attacks - 38 hits, 4 crits, 30 misses, 2 critical misses.
Or 56,75% hit rate (which is close to the 57,75% hit chance of a 35% hit attack with an advantage), with 5,4% crit rate and 2,7% critical miss rate.
So the conclusion from this limited test is:
- advantage does seem to work to at least some extent.
- the actual hit chances seem to be lower than they are supposed to be (at no point of time did I exceed the expected hit rate, after the 1-st 6 hits).
- the critical rate doesn't seem to be affected by the advantage, as it should. In theory, attacks with advantage should have around 10% crit rate. While in my test I got 9 crits out of 162 attacks with advantage, or 5,55% crit rate.
- since the advantage hit rate seemed to be about as high as if the enemy had +1 armor class, and the crit rate wasn't as high as it was supposed to be, I can speculate, that the extra roll is a 1d18 instead of 1d20. But this is just a speculation, that could explain these numbers. I don't know if that's indeed what's going on. I should have in mind that the attacks without the advantage also had seemingly lower hit rate than they were supposed to, but the amount of attacks without advantage was just 18, so I can't draw definitive conclusions from that. Another plausible explanation would be that the proficiency bonus is applied to the 1-st roll, and it compares the 1-st roll, with the bonus to the 2-nd roll, without the bonus. But if this was the case, than the critical rate should've been closer to 10% than it was in my test.

And I'd like to add, that while 180 attacks may not have been a sufficient amount to get a more accurate statistics, I had noticed this problem BEFORE conducting this test. I had noticed that I miss more often than I'm supposed to and this is why I started this thread, thinking that the advantage bonus wasn't working.

[id795078477]

Originally posted by WereElf:

Well, I did a little experiment on the Cambions in the tutorial area, and here are the results:
40% chance to hit (no advantage): 18 attacks - 4 hits, 1 crit, 11 misses, 2 critical misses.
Or 27,8% hit rate, with 5,55% crit rate and 11,11% critical miss rate.
51% chance to hit (30% with advantage): 88 attacks - 34 hits, 5 crits, 48 misses, 1 critical miss.
Or 44,3% hit rate (which is close to the 43,75% hit chance of a 25% hit attack with an advantage), with 5,68% crit rate and 1,13% critical miss rate.
64% chance to hit (40% with advantage): 74 attacks - 38 hits, 4 crits, 30 misses, 2 critical misses.
Or 56,75% hit rate (which is close to the 57,75% hit chance of a 35% hit attack with an advantage), with 5,4% crit rate and 2,7% critical miss rate.
So the conclusion from this limited test is:
- advantage does seem to work to at least some extent.
- the actual hit chances seem to be lower than they are supposed to be (at no point of time did I exceed the expected hit rate, after the 1-st 6 hits).
- the critical rate doesn't seem to be affected by the advantage, as it should. In theory, attacks with advantage should have around 10% crit rate. While in my test I got 9 crits out of 162 attacks with advantage, or 5,55% crit rate.
- since the advantage hit rate seemed to be about as high as if the enemy had +1 armor class, and the crit rate wasn't as high as it was supposed to be, I can speculate, that the extra roll is a 1d18 instead of 1d20. But this is just a speculation, that could explain these numbers. I don't know if that's indeed what's going on. I should have in mind that the attacks without the advantage also had seemingly lower hit rate than they were supposed to, but the amount of attacks without advantage was just 18, so I can't draw definitive conclusions from that. Another plausible explanation would be that the proficiency bonus is applied to the 1-st roll, and it compares the 1-st roll, with the bonus to the 2-nd roll, without the bonus. But if this was the case, than the critical rate should've been closer to 10% than it was in my test.

And I'd like to add, that while 180 attacks may not have been a sufficient amount to get a more accurate statistics, I had noticed this problem BEFORE conducting this test. I had noticed that I miss more often than I'm supposed to and this is why I started this thread, thinking that the advantage bonus wasn't working.

https://en.wikipedia.org/wiki/Statistical_significance

EDIT: That's NOT to say you're wrong. It's just that your observation is not (yet) conclusive. Given enough other players noticing similar patterns it may have better grounds.

[Hobocop]

You'd need to run several hundred tests (ideally, around 1000 or more) to have a statistically significant sample. A few dozen isn't enough to determine much of anything.

There was a post by someone months ago on these forums who did 1,000 test rolls in the tutorial area against the same imp in patch 3 and recorded every result. The conclusion was to be expected; a relatively even distribution of results with about half being 10+ and the other half being below 10.

[WereElf]

I wasted more time on this test than I was willing to. If you're up for it, you can do it too and post your results here. The more people do it - the more accurate statistics we'll get.
But please, keep different track records for hits with advantage and ones without it. And also note when a hit is critical and a critical miss, so we can have more useful information, and if there is indeed a problem, we could use the data to determine what could it be.

[dolby]

Originally posted by WereElf:

Well, I did a little experiment on the Cambions in the tutorial area, and here are the results:
40% chance to hit (no advantage): 18 attacks - 4 hits, 1 crit, 11 misses, 2 critical misses.
Or 27,8% hit rate, with 5,55% crit rate and 11,11% critical miss rate.
51% chance to hit (30% with advantage): 88 attacks - 34 hits, 5 crits, 48 misses, 1 critical miss.
Or 44,3% hit rate (which is close to the 43,75% hit chance of a 25% hit attack with an advantage), with 5,68% crit rate and 1,13% critical miss rate.
64% chance to hit (40% with advantage): 74 attacks - 38 hits, 4 crits, 30 misses, 2 critical misses.
Or 56,75% hit rate (which is close to the 57,75% hit chance of a 35% hit attack with an advantage), with 5,4% crit rate and 2,7% critical miss rate.
So the conclusion from this limited test is:
- advantage does seem to work to at least some extent.
- the actual hit chances seem to be lower than they are supposed to be (at no point of time did I exceed the expected hit rate, after the 1-st 6 hits).
- the critical rate doesn't seem to be affected by the advantage, as it should. In theory, attacks with advantage should have around 10% crit rate. While in my test I got 9 crits out of 162 attacks with advantage, or 5,55% crit rate.
- since the advantage hit rate seemed to be about as high as if the enemy had +1 armor class, and the crit rate wasn't as high as it was supposed to be, I can speculate, that the extra roll is a 1d18 instead of 1d20. But this is just a speculation, that could explain these numbers. I don't know if that's indeed what's going on. I should have in mind that the attacks without the advantage also had seemingly lower hit rate than they were supposed to, but the amount of attacks without advantage was just 18, so I can't draw definitive conclusions from that. Another plausible explanation would be that the proficiency bonus is applied to the 1-st roll, and it compares the 1-st roll, with the bonus to the 2-nd roll, without the bonus. But if this was the case, than the critical rate should've been closer to 10% than it was in my test.

And I'd like to add, that while 180 attacks may not have been a sufficient amount to get a more accurate statistics, I had noticed this problem BEFORE conducting this test. I had noticed that I miss more often than I'm supposed to and this is why I started this thread, thinking that the advantage bonus wasn't working.

I'm no expert on randomness and my memory is crap. Someone can give you a better example or anything i found numbers boring in school... but still...
Is “1111” random? Or “01001001010010111”? It’s hard to tell, right? Why? Because the moment you start to define the parameters of randomness you lose your randomness...
And we sux at grasping it, i know i do anyway, i reckon best way would be to do a visual representation and even then.

As seen above it is nearly impossible or at least i have hard time knowing and to accurately see if that sequence of numbers is random. Generating a sequence with a million “‘0 and 1’s” can be very random, but not very desirable. It is easy to make correlations based on what logically looks random, but does it make sense, is that really random?

Anyway, you would need bigger sample size easiest way i reckon would be to do the flip coin test. at 50% hit chance for 100 tries and see how it goes. At 50% you should get around - ish 54 hits and around 46 misses and vice versa, something close that at this sample size.

Due to small sample size you need a huge session numbers to get to 50%...At 100 samples you would need at least 500 sessions the more the better. so thats 500x100 attacks... good luck with that.:))

50 sessions of 100 (5000 attacks) or less is never sufficient, best would be 2000 sessions+ Minimal gain occurs above 2000 i think.. Even at 500+ sessions the 50% 50% split would be rare per session around 7-8% if i remember correctly. So good luck proving that.
ow and you will need luck getting them to metter to anyone but yourself. Convincing people on forum is hard, even more so if they are fans...

i'm not sure why you are trying to do this anyway, but each to his own i guess. I would think casual players don't care about this, at least i don't... Don't know about the rest.

[guard65]

Not that I care that you miss for whatever reason in EA but I do not know where you get your math from .

If advantage is you roll twice and take the better of the two then 1 on a 20d is 5%. Or by rule of numbers your chances are 0.05 x 0.05 equaling 0.0025 = 0.25% = 1/400th odds. Just saying.

[raymond]

guard65 - just to clear up your misreading, the OP stated they had _two_ misses with advantage in a row, so that'd be a sequence of 4 rolls of 1 on a d20

[WereElf]

Originally posted by dolby:

Anyway, you would need bigger sample size easiest way i reckon would be to do the flip coin test. at 50% hit chance for 100 tries and see how it goes. At 50% you should get around - ish 54 hits and around 46 misses and vice versa, something close that at this sample size.

I'm fully aware how maths, statistics and randomness work. The post here is not about testing randomness. It's about testing if the advantage/disadvantage in BG3 works properly. And after conducting my test, it became more about "are the hit chances in BG3 what they claim to be?"
What you are saying is like:
Me - my car doesn't seem to work properly.
You - do you know what a car is?

I acknowledge the fact that my sample size may not be big enough to be accurate, but I also know that when you are working with chances, the success rate should be gravitating around the calculated chance, rather than around a number that's apparently lower. As I already said: at no point of time did my result get near the expected success rate after the 1-st 6 hits. Perhaps it would have started doing that after another 1000 hits, but I don't have time to do that. And that's why I'm asking others to do the same test and post their results here.

[guard65]

Hmm, so you mean, 0.000000390625 = 1/2,560,000 odds for 4 natural 1's.

Dice is this game needs complete log transparency on all rolls so you can see what their randomizer has generated and every modifier in the final to a value end. This includes both rolls used in advantage and disadvantage. It is annoying trusting them before they have proven their math or themselves - lol.

None the less if unknown modifier #9 gives you -5 to hit and you have +3 from abilities then perhaps a 1, 2, or 3 is a critical miss since the end results are a 1 or less in each case verse rolling a natural 1. I suspect hidden modifiers or bad equations on their part. Then again we do not see their pile of weighted dice value applied ether (which i disable). Only transparency will show the true cause and I hope they added a log toggle for show it all (debug).