Larian Studios Forums Baldur's Gate III Baldur's Gate III - General These skewed rolls are ticking me off

Its 1/400

Im quite sure it isnt.
This formula dont take under concideration that those numbers are supposed to be in specific order.

This seems like a chance that out of 4 rolls, you get those 4 numbers, nothing else.
Meaning 2 3 4 5, or 3 5 2 4 would be both acceptable.

Unless i forget something ... it has ben a while since i needed this.
//EDIT: Oh you said ^4 ... that changes things, i thought it was * ... i take it back.

I understand that the odds are small. But it does happen. Ive also had the opposite. Rolling 2 20's. Or twice the same number, etc.

Talking from a tabletop perspective. Generally dice are rolled at a far faster rate in bg3 then in actual pnp dnd (at least for my groups) and the player is making ALL the dice rolls. So if hes at the table and his friend rolls a few ones where you all had a good laugh about it. Now the player has ALL of those and it suddenly feels alot more personal.

I think its just confirmation bias tbh. We tend to overlook the positive rolls and only renember the (really) bad ones and when you get all the bad rolls of the party you start to take notice.

If it happened once I would just chalk it up to bad luck, but it happened multiple times in multiple combats. Also realize that RNG in computers is moonshine and should be treated as such.

Is it not part of the game? Otherwise this would be a totally different game. Role-playing, it looks like the bard may have the flu, it happens.

From what iv noticed rolling a Nat 1 gives you -5 on your roll so if you got more bonus then the new -5 you still pass the roll... Yet more reason to always use Guidance lol

Currently in game, rolling a natural 1 on any check automatically fails you without taking into consideration any bonuses or even the original DC - Larian have confirmed that this is their change and is working as they intend. All we can do is bombard them with regular feedback to tell them that natural 1s are not auto-fails on skill and ability checks, or saving throws, and should only apply to attack rolls (and the very specific death saving throw), and to ask them to change their mind about this.

You can do so on their bug report form, here: https://larian.com/support/baldur-s-gate-3#modal

In terms of the RNG, Larian have been shown to use an RNG that doesn't source its chaos variables well enough, or regularly enough, and so its roll pattern creates a visible sine wave of variance, where other contemporary games do not.

https://docs.google.com/spreadsheets/d/1tdyBoQNS_vwEGZGBgFRQex7b-Ma8S6P7zvEMK5wh9n4/edit?usp=sharing

This record was made in an earlier patch - You can see that while the RNG ensures that the balance of numbers comes up within the average bounds, the actual distribution creates a visible algorithm pattern that you can legitimately track in game.

I've not had the stamina to re-do the numbers since then, but it was before they introduced loaded dice as a 'fix' to the RNG complaints, so I've got little confidence that they've actually improved their base RNG. If you're not using loaded dice, you'll experience this - if you are, you'll experience an even more visible correction effect that will do you dirty as often as it helps you.

There's something seriously messed up with their RNG.

I don't know a good way to get the game to output from the combat log to demonstrate this though.

Thanks for linking to the data. I'd heard about the sine wave shape of BG3's RNG, but seeing it like this makes it much clearer. I now understand where that sinking feeling comes from when I miss three 80% shots in a row: I'm about to fail everything for the next round or two.

The best way probably involves a mod interacting directly with the RNG engine to automatically log rolls.

The dirty way is to get a PC's AC to 19 or 20, then attack them with a wizard (-1 strength) with a long sword (no proficiency). Every roll would miss except a natural 20, which gives you a bunch of consecutive rolls to record by hand from the combat log.

I don't play TT 5e, but page 242 of the DMG addresses the issue of saves and ability checks, discussing not only critical success/failure (extreme outcomes), but also partial success/failure, and marginal success/failure. The notion is to encourage DMs to consider a continuum of outcomes and complications. rather than always apply simple pass/fail absolutism, which is not always a satisfying experience for players.

So whatever Larian have decided to implement for ability checks and saves, it is not a "bug", but a development choice. You can still tell them you don't like it ( with or without coherent reasoning ), but it isn't wrong according to 5e DMG.

Personally, I haven't found it matters much. The game plays enjoyably well, regardless of how the die falls.

Doing it that way takes about an hour an a half to record 400 rolls. I've got some reading to do to analyze the data properly; it's here should anyone like to take a crack at it.

I did get a response from Larian that weighted dice are meant to be better for the PC.

So weighted dice should be considered a difficulty setting button, and unrelated to the RNG convo. I didn't explicitly realise that before.

I think pure RNG isent what we would want honestly. Because we will then get scenarios when we get chains of 1's. Or chains of rolls below 10 for example. I get it at the tabletop sometimes but the social aspect and the sillyness of it makes for some laughs with your friends. In a single player computer game you lack that though and you just get the bad luck of the dice.

The same applies for the enemeis you are fighting. When I dm and I get a chain of nat 20's il tone the results down. While I typically dont hold my punches im also not out for a TPK. If the stars align and my players get poor rolls and I get good rolls it becomes a slaughterfest with not much that the players can do to help it. I can tell you right now that true RNG simply isent always fun. I think Larian's system reflects that.

Mind you, ive seen the occasional bad luck streak but nothing to the degree OP is mentioning. Did others experience it as well?

Used to get occasional weird number runs; except it was not weird at all, since rolling a single die is an uncorrelated unique event, with exactly the same chance of each number on the die, every single time.

I don't think the periodicity in the RNG Larian use is as important as some would claim. It certainly would not cause such extremes as getting a "1" on ten consecutive rolls. A "true" RNG could cetainly do that ( as happens in real life TT ), but probably not an RNG that shows minor patterns.

Yeah, probability is only that, probability. Not certainty.

Would you consider Larian implementing auto-fail and auto-succeed for ability checks and saving throws to be "considering a continuum of outcomes and complications. rather than always apply simple pass/fail absolutism" ... Because I would not.

Pass/Fail absolutism seems to be entirely what they build all of their decision trees and dialogues around, currently, despite their claims to the opposite. Categorically, you're faced with a situation, and you can "do nothing" or "Not try" and see, guaranteed what the fail state outcome is... or you can "try to do something", and either succeed... or fail, and see the same failure outcome as if you hadn't tried. That's it. That's the entirety of their outcome mapping. They are not making-failure-fun, currently.


That's what I said, yes. They have confirmed as much directly.

Depends how you look at it. D&D mechanics are generally an exercise in simple linear arithmetic; a collection of + and - values, summed with a die roll, compared with a taget value determining pass/fail. To me, that is "simple pass/fail absolutism".

Anything diverging from that is a nuanced (non-absolute) result. So, advantage/disadvantage, or critical pass/fail, are both nuanced outcomes that diverge from the linear norm.

I am not aware of any base D&D mechanics that deal with partial or marginal success. It would be possible, but tedious, to detail a selection of outcomes for each occasion where a check ( combat and non-combat ) is needed, but over the years it has been easier for TSR/WotC to just give wooly advice ( as in the DMG p242 mentioned above ), and make it the responsibility of the DM.

So, in answer to your quesion; no, I would not say Larian have considered a continuum of outcomes and complications. They have just copied the WotC 5e ideas and implemented only the easy non-linear behaviours ( advantage/disadvantage and critical ), without any actual variance in the outcomes.

My point ( as much as I had one ) was that I don't see Larian's implementation as being outside the scope or intent of 5e, as documented, even if it is not as elaborate as it could be. You, on the other hand, seem to be suggesting that their implementation must be changed, even though you did not explain why this is so.


Yes, unfortunately, that is largely true, both of this game, and almost any other current game. Most game dialog systems are quite crude, usually allowing a few "flavour" interactions based on character attributes, or perhaps easier/bypassed checks.
Relatively few games have many interactions of world-altering significance, because an interaction-based world-state is very difficult to manage unless the story-line is quite linear.

As with many other games, BG3 seems ( at present ) to have a single major game-path choice, a selection of faction-approval choices, and a selection of incidental choices to gain advantages.

This is not particularly exceptional in EA, although we can set up quite a variety of future content for Acts 2/3. I am sometimes optimistic that Larian are knitting together a truly labyrinthine plot, but realism and experience suggest that is unlikely; partly because the development cost and difficulty would be high, and partly because a lot of the audience seem to care more about combat, and don't want the story to be complex.

If I told all people who has bought early access to BG3 to flip a coin 10 times. We would all TOGETHER get very close to 50% heads and 50% tails. I think we might be enough people that we get cases where some of you get all tails and some of you get all heads, the problem is that one person getting all tails and one person getting all heads still result in the total being 50% head or tails.

The fewer coin flips each of us do, the higher the chance of getting results where we get away from 1 tail for every 1 head. Flipping just one coin and it is impossible to get both after all.

This is creating a problem for some modern games that has concentrated the number of random rolls, like XCOM and all the missed 99% and even 100% shots.

Unfortunately BG3 being based on D&D is the perfect storm of all that. We have a SINGLE person doing all rolls, we have a low number of rolls and possible a high number of people able to easily compare data on the internet.

Let's face facts here, you are outright lying if you tell me it's impossible for someone to roll 10 heads or 10 tails in a row. Judging someone rolls based on the perfect math random distribution is another way to lie and threat people badly, because no persons reality is equal to perfect random distribution. It's like finding the average family, there is no family that is average, because no one has 2.5 children.

This makes it sound like you're in favour of keeping the absolute succeed/fail binary outcomes that Larian are using nigh exclusively, but you just want the way we arrive at those binary outcomes to be made more complicated... To be clear, I suspect that's not what you're saying, but that's how it reads and, if it is, I'm not a fan, personally. I'd rather have actual nuanced outcomes.

As for rules regarding partial successes and failures and marginal outcomes, they do indeed exist and are in the DMG. Most DMs use them to some small extent, in various places, in places outside of direct combat. It's not actually that complicated, either - it's pretty straight forward.

It mostly comes up around ability checks specifically. In particular, in group play, graded outcomes and variable results are an important part of managing characters, when different party members have wildly different capabilities, but there's no good reason why several of them shouldn't make a roll. This involves modulating individual DCs and having variable graded outcomes for different characters. This is what prevents the 'bad dm' situation of "the barbarian recognises wall of force and knows that it can't be dispelled, but needs to be disintegrated, while the Wizard-who-cast-WoF-yesterday does not", which is a trap that Larian are currently falling into in several places.


It is outside the intent of 5e, directly, because it's not a feature in the core rules; it used to be a feature of earlier editions, but it is not a feature of 5e, and that is not an arbitrary choice. It's quite deliberate. The ruleset is build in many ways that take that as an understanding - that outside of attack rolls and death saves, 1s and 20s are not absolutes. Bounded accuracy plays a part in this, but there are other elements and features of the ruleset that were built with that understanding in mind.

So yes, I want them to Fix this rule change that they have made, which was a feature that was deliberately redacted from the current edition ruleset because of the way said ruleset was made. I know many people will reject the notion, but it's a system balance issue.

The development of variable outcomes doesn't need to be world-shattering; even something as simple as "I'm not going to do this, and I'm going to go report you to the guards now"/"I'm not going to do this, and leave me alone"/"I'll do this, but don't ever talk to me again"/"I'll do this, but [there will be a penalty]"/"I can do that, if you make it worth my while"/"I can do that, sure. Good luck.", where the 'world shaping' outcomes are still that something happens or it doesn't, but the persinal-level outcomes are more varied and have more outcome states... even that would be more that Larian are delivering at the moment.

==

To Miravlix - Probability is one thing, and most of us have an understanding of it. The issue is that in video games and other computer-based programs, we don't have actual random probability; we have simulated randomness, based on algorithms which, in the more sophisticated cases, regularly re-seed themselves and source chaos variables from external points. the problem arises when you use a more simplistic algorithm, and it becomes visible - when it becomes visible and to a certain extent predictable, that's a problem for a game based on those random number generations. Larian's algorithm for their random number generation is too primitive, and the flow of it DOES become visible, and to a certain extent predictable, all too easily, and that's a problem - that's what the data table I linked above is showing.

Indeed, if a game is going to use a random element, I'd rather that it is used to select between a continuum of outcomes. In much of D&D the continuum of outcomes is often just pass/fail. Critical success/fail moves the continuum of outcomes from 2 possibilities to 4, partial success/fail adds 2 more oucomes and marginal success/fail adds another 2. I generally prefer more outcomes, even if they are as restricted as critial hits and fumbles

But D&D doesn't do outcomes well, and never has. Most players of the "white box" edition that I knew disliked separation of hit and damage rolls, and the complete lack of correlation between the two. It would have been simple enough to use a single attack roll to scale the damage an attack might cause, but over 40 year's later, D&D still uses the same poor system, bar the reversal of the armor classes for players that had difficulty with subtraction.


Which D&D rules of the many available Larian decide will make a good videogame is up to them. You can certainly express a personal preference, but that is different from their choice being wrong.

There is no "intent" to 5e. It is designed around what TT players said they liked ( this is a good thing ), along with WotC having a better understanding of how to follow social trends that increase their market size. Both factors they seem to have totally screwed up with 4e.

One thing 5e is definitely not, is balanced. Chess is balanced, Checkers is balanced, Go is balanced, any game where the players have the same capabilities and conditions is balanced. RPGs are not balanced, and would be dreadful if they were. If anything, a better description of 5e is that it is more bland, because your character choices matter less, with limited meaningful specialization compared with earlier versions.



Then we seem to be playing different games ( or you always play the game the sme way, perhaps ) as I have found that there are many ways of approaching most of BG3's content, and many (apparently) different outcomes.

Larian could be less inclined to turn everything into a fight, on occasion, but on the whole there are many more choices available than an "average" RPG.

This is getting off topic, so I'm spoilering it out, and I'm probably not going to pursue an argument any further.


The one section I'll leave out of the tagging is this:


Expressing my opinion and calling for a change is all I have been doing.

I'm not sure why you've taken issue with me doing this, over anyone else doing the same.

I specifically called out confirmation that this was not a bug and was Larian's intention, and that the only thing people could do if they disproved was to give that feedback and ask for it to be rectified to 5e rules, if they felt that way. It is, in my opinion, a wrong decision; I want them to fix it. It is a homebrew rule that, in my opinion, they should not have made. That is quite directly what I have been saying each post from the first one you chose to respond to - that it is not a bug, that it is Larian's deliberate homebrew, intended by them, and that if you want it changed to line up with actual 5e rules all you can do is submit that feedback to them.