Larian Studios Forums Baldur's Gate III Baldur's Gate III - Suggestions & Feedback Josh Sawyer about the DICE in a low level D&D + HP bloat is no solution

Dungeons and Dragons has always, in every iteration, struggled with making things challenging. Encounters are too often either effortless or impossible. I think the combat scenarios so far in the EA are pretty well done. I have found if I blunder into things that sometimes I will get my ass handed to me, but that there is always an approach which gives the player an advantage if they take the time to look for. I actually have no qualms with Larian on this topic. . .

However, I have to admit I have save scummed my way through some shit, mostly dialogues. Statistics should be meaningful but they simply aren't. I am sure anyone here who played XCOM has some RNG nightmare stories, but as much as I hated that rare miss at 100% it was the only game I have ever played where I would take throw away shots at 11% and still make them often enough that I have distinct memories of them. Of quite a few of them. And as rough as the RNG was, it rarely felt unfair (although there were times, let me tell you). There were instances where it was wildly off and some encounters I would be rocking would flip and I would find myself decimated. . .but most games with an RNG seem worse. It feel as though your odds are never better than 50-50. You start to just accept 65% will always miss despite the fact it is a two out of three chance to hit. 33% is 1 out of 3 but realistically it is more like 1 in 10. That is why I save scum, when I do save scum. If you are going to feed me percentiles that I am supposed to make my choices contingent upon, make them accurate or find another system to use. I understand there will be aberrations, but after awhile I feel like I am just being played for a sucker. 40% to pick a lock? If I lose four lock picks and still haven't gotten past it then fuck you. If I have four die in my pool to use on a re-roll and I can't break ten then fuck you. Whenever I save scum it is never about winning, at that point it is just me telling the game to fuck itself.

Before anyone starts saying I don't know how this shit works, I won't argue, I have had long and tedious arguments over this shit in the past. However. . . I play Horse and 10,000 with my favorite bar tender over his tips to pass the time. I play spades and/or Texas once a month religiously with some friends. I haven't hit Vegas in years, but a buddy and I used to go there twice a year and while we only played Blackjack, neither of us ever left worse than more or less even with what we brought (though we would also bring 5 grand or so to ride out those rough patches). All of which is to say I may not get statistical theory but I get percentages and I have seen some shit, but nothing in reality is as consistently bad as video games.

I used level 20 in both cases because 5E doesn't have epic level rules (yet). And it makes far more sense to compare same level to level.

Despite the naming of the tiers, the 3.5e and 5e characters still advanced in the same way. You still get level 9 spells at level 17, and the Monster CR are roughly equivalent in both editions (i.e. strongest Dragons are all around mid-20s CR).

3.5E's epic rules were 100% supplementary - not part of the Player's Handbook (which also went up to 20). It was also terribly balanced - as epic martials without insane min/max-ing via prestige classes from in supplementary material are a joke (they'll soundly lose to pre-epic casters). Nevermind that AC scaled so better than AB in 3.5 that martials will eventually not be able to even touch themselves. BTW, in 3.5e, epic fighters and wizards had the same attack bonus scaling (1/2) in epic levels.




I used pretty typical numbers for both editions (not super-munchkin, but respectable).

35 AB = +20 BAB, +5 Weapon, +8 STR (26*), +2 Feats (Weapon Focus, Greater Weapon Focus)
*16 Starting + 4 Levels + 6 Giant Strength Belt.

You have to remember Pathfinder/Kingmaker and its +8 every attribute items are an oddity when it comes to table top games. Items in the 3.5 DMG mostly only went up to +6. BTW, the specialization feats only add damage, not Attack Bonus.




I thought your point was that things in 3.5 were easier to hit and less dependant on RNG? Granted, Pathfinder is not 3.5e - it's a bit more higher power, and Kingmaker is even more so.

A 42 SR would mean a level 20 wizard with both Spell Penetration feats (+24 total) has a 15% chance of landing a spell that checks for SR. IIRC his SR in core difficulty is in the mid-30s, but that's still high enough to make magic use against him basically RNG.

It's funny that you bring it up because the Spawn of Rovagug in Kingmaker is EXACTLY the thing you're complaining about. HP Bloat (1.5k HP), and high Spell Resistance = making spell attacks useless.

Guess what? It works just fine in Kingmaker, because the game doesn't follow PnP, and is designed in a way that you get insanely good equipment. 1500 HP and 40 HP regen isn't that much when your characters in Kingmaker can put out 600+ damage per round, and trivialize his AC and DR with Smite, etc.

My own feedback on this point is something I wanted to go into in more detail at some point, but for now:

Larian have made comments about how missing isn't fun. Sure, maybe, BUT, in 5e, missing regularly isn't actually a major problem, in my experience. You succeed more than you fail, generally speaking. (A base level character will have on average a +5 to hit, and so will hit an AC14 creature (the generous average for CR1 targets), slightly more than half the time. A level 20 character will reasonably have a +15 to hit, and will hit an AC22 creature (average AC for CR 25-30 creatures) notably more than half the time) bounded statistics serve the purpose of not required characters to be super-specialists and use all of their growth options on 'given necessities' just to have a chance at being effective against on-level targets, especially at higher levels, and it works. Super-specialists still have an advantage, of course - that level 20 archery-focused sharpshooter might have a +21 to hit, and will very reliably hit even Tiamat, but people who haven't sunk literally all of their growth into laser-focusing one thing will not feel like they have no chance to contribute either.)

Despite this, they've tried to 'address' this perceived problem by reducing AC and increasing hit point, and another poster elsewhere has already gone into great depths on why that's a terrible solution, which only succeeds in devaluing other aspects of play, and channeling players into specific courses of action over and over again, because they're the only things that are effective uses of their turns. In particular, they've increased the success/fail ratio artificially for ONE side of the offense equation, and completely greatly devalued the other - it's no easier to succeed when you're forcing saves, and when you're forcing saves for damage, that damage is devalued by the increased hit points... so with saves, you'll fail more, and achieve less when you do succeed...

The real issue is not with the game rules and mechanics, which Larian are hacking apart (badly) to try to fix their perceived issue. The issue is with the computer RNG that Larian are using.

No computer RNG is truly random; it's still a code based formula for deriving values. Some, however, are better than others, and if you spend enough time with one, you can get a feel for the way it churns its values out. The rng that Larian are using is not a good one. It's not 'bad luck' when it's consistent and reliably predictable - it's a problem in the code.

This isn't about missing on an advantage roll when you only need a 3 or higher to hit... this is about missing multiple times on advantage rolls when you only need 3 or higher to hit, specifically when you're targeting boss or low hp targets, and having this occur *Frequently* and *Consistently*. I've been watching my combat logs for the past dozen sessions and noting situations - and it has been far, far, FAR too consistent where cases of rolling under 5, even with advantage, upwards of eight times in a row (that is, four attack rolls with advantage), on creatures that are on very low hp, or are boss-like creatures. It's not 'bad luck' when it happens this reliably. This happens to players too (enemies trying to down you), though to a substantially less pronounced extend (maybe only 40% as frequently).

Larian needs to invest in a new RNG to run their system. That would, quite probably, solve a great many of the 'too frequent missing' problems... The other would be to actually revert their ridiculous house rules and run the actual 5e system, so that they aren't creating imbalanced gulfs without understanding that they're doing so. Please, Larian - get a new RNG to run your system, and *Trust* the base 5e rules and statistics to create an engaging combat.

Nor will have, a lv 17 char is "hero of the world", meanwhile, on 2e, Meredoth is the ruler of the Nocturnal sea realm of dread and is a lv 20 necromancer. You can have high level and local adventures.




And was fun non the less.

The point of a game is to be fun.




I was wrong. He has 34 SR only, so is about 50% chance of hitting with both feats.



Do you know what Rovagug is?

Rovagug is a God of Destruction which NOT EVEN ALL GODS COULD DEFEAT. Only trick and imprison. Their spawns has a tiny fraction of their power but are godlike non the less.

Here is a guy killing this "bloated" boss within 2 minutes on the hardest difficulty.



And PFKM would be better if was more faithful to P&P.

If magical gear wasen't that powerful and sneak attack wasen't broken OP. Keep in mind that is possible to kill him with phantasmal kill.

The mega bosses in PoE2 is the worst boss encounter design ever....

I found PoE as a very lackluster version of BG 1/2. Pathfinder Kingmaker in other hands. I loved so much. Have 523 hours on it.

This is my number one problem with by-the-rules D&D, and ESPECIALLY especially with 5e. I call it the "Tyranny of the d20". It's been a problem always, but it became less of a problem in 4th edition, and WAY more of a problem in 5th. (Note that I am mainly talking about skill checks here, but also applicable to attacks.)


Here's what else Sawyer had to say in that Twitter thread:

The difference between a character with Proficiency and without is 2, a measly 10% on the d20. With rogue expertise (for example) it's doubled, but that's not usually where these checks come up. A good or bad attribute can easily overshadow that.

So it's not just that it's random, it's that the range of the die and the distribution of results is wider than the typical range of bonuses for starting characters, which means that the die result is more consequential than how you build your character.

Feels bad man.

Cf. Disco Elysium, which uses 2d6 (bell curve, not flat distribution) for its skill checks and adds to it a range of values that are comparable to (if slightly lower than) D&D's.

The dice are generally less important than how you built the character. Feels better man.



This is what I'm talking about. The d20 is SO swingy. It can be a 1 just as easily as a 20, and either of those just as often as a result in the middle. You can roll very low several times in a row, and that isn't a rare occurrence. And in 5e, you don't get to add much to your roll. At first level in 5e, you can have a maximum of +5 to a skill check, or +7 IF you are one specific class. And you're usually rolling against a number between 10 and 20, most often (for anything meaningful) around 15. If you're a "specialist" in that skill, and not a Rogue, you're still only looking at a 55% chance to succeed. A 65% chance if you're a hyper-specialized Rogue. There's simply no way to be better at that skill. Any skill. When you can only add +5, and you are rolling the ENORMOUS swingy range of 1 to 20, your character's own capability means very little compared to the pure random chance of the die.

In 3.5, you could have up to a +13 to a given skill at level one, if you specialize in it, and that's without being any certain class. This allowed you to be really good at something, if you wanted to be, and thus made the roll of the die much less important in comparison to your actual character's build.

In 4e, you could have up to a +17 (!) to a given skill at level one, if you went all-in. Now, you really could be actually a specialist in something, and the die roll could hardly screw you over at all.

5e, let's dial it all the way back to +5. Now, when people have to make skill checks, it's about who gets lucky, not about which character is good at something. DM says, "everybody roll Stealth", and oh look, the max-investment Rogue rolls a 3, while the chucklehead with an 8 Dex and no proficiency rolls an 18. Isn't it hilarious? The supposed expert sucks at this, and the inept clown is actually really good! What a funny anecdote! If that happened ONCE, it would be a funny story, "remember that one time when...?" But in 5e, it can happen over and over and over again, because the result of the d20 roll HUGELY outweighs the way a character is actually designed/concepted/built.

How many times in your games has it been the 8 Int Barbarian who has somehow come up with the clutch Arcana knowledge, while the Wizard is scratching their head, because "lol crazy die rolls, am I right?" The swingyness of the d20 is too great. The possible bonuses for characters needs to be higher, in order to compensate for it. And while I'm using skill checks as my example here, it also does apply to attack rolls as well. 3.5 and 4e gave PCs more options for becoming extra accurate, whereas in 5e, you're pretty much stuck with the same shitty bonus to hit that everyone else has. "Cool, I guess we're all at +5, then? Yeah, looks like it's +5s all around."

Like Sawyer says above, this FEELS BAD. It feels bad to build a character who is supposed to be specialized in Persuasion and Deception, and helpless at Religion and History, yet find that you fail those social checks over and over simply because of the Tyranny of the d20, and yet somehow you're making all the knowledge checks because "lol, these dice man". At that point you don't even have a character concept. Who your character is and what they're good at is just randomly determined on a case-by-case basis, depending on the capricious whim of the dice.


Lurking in the shadows, Darkstalker Shadowcloak whispered, "I can get past those guards unseen. I've spent years mastering the arts of stealth. I am a ghost in the night, I will slip past them as easily as..."

"Hold up there a second, buddy," Bumblefoot Mugbreaker interrupted. "I'm pretty sure you actually just have about a 10% better chance of doing it than I do, and I've got this club foot here."

Quietly, at least 10% more quietly than others spoke, a voice from the darkness muttered, "Fuck this game."



Applicability to Baldur's Gate 3? Uh, I dunno. They have to do it the 5e way, I don't think it's something they can really fix. Unless they use some cheater RNG that doesn't actually model a d20.

Your posts are so good. Thank you for this. HP Bloat is a MYTH. It's not really a thing. And to whatever extent it IS a thing (which is very small), it's still not a RULES CHANGE. It's DM prerogative to adjust creature stats, or to make NPC monsters with class levels. The published modules are full of cases where monsters have non-MM standard stats, ESPECIALLY hit points. But there's barely any "bloat" to speak of in Baldur's Gate 3, anyway.

I'd agree that calling BG3 HP standards "bloated" is very much hyperbole at the moment.

But I would call them inflated. And so far we have only just seen the low end of of the level scale, so to speak. How inflated will the HP be when we are level 10 or 12? I do think there is cause to think about this now.

I’d say that the bloat is less of an issue compared to the mess with the AC at the moment. That combined with the advantage system turns utility spells created in a way that you can hit more lose their values. I have already posted a list of spells which were directly affected by this adaptations and not to mention the direct impact in some classes advantage dependent.
The math is quite simple: a +2 over a basis of 55% worth more than a +2 over a basis of 75% although the flat increase is 10% for both scenarios.

If the dm decides to change the challenge rating of a battle, modifying only the AC instead of modifying both AC and attribute modifiers will snowball to through the entire game balance. In the end, what concern me the most is the spell efficiency which is clearly one of the most important pillars of the game.
Buffing cantrips and nerfing spells slots isn’t the best solution to a caster joe feels more useful.

According to me there is another problem that leads to many frustration and totally unbalance the game : the lack of possibilities to increase our % to hit.



That's right, but advantages are the "god mode" of D&D and the AI will never be able to create the best conditions to play very well with advantages/disadvantages.

You cannot balance a game if your only mechanics to increase/reduce the % to hit, on a scale from 1 to 10 are either the 1 (where nothing happens) or the 10 (which is overpowered and equal to a level 9 spell).
That's probably why the game is way too easy for many, and way too difficult for the others... Because there's nearly nothing in the middle.

2 - 9 are a great place for custom rules even if 1 has to be a part of the game and if it's fine and fun to be able to use 10 in a few specific conditions.

- Doubled proficiencies bonus in certain conditions could be a thing. (that's not more than +20% to hit if you're level 8-12, a little bit less than an average advantage).
- New actions, bonus actions or reactions could lead to many new things... Let's say a +3 to attack roll if someone use the "half custom rule" to use"help" as a bonus action to distract your target.
- +1 / +2 / +3 arrows or melee weapons could be another one.
- Higher ground could also lead to another x2 proficiency bonus (or a modifier x1.5, I don't know) if you're higher. This could be at the end of the game something "easy" to increase our % to hit but it could be "a part" of the mechanics, not THE things to do.

This would lead to a more complex strategy game... In which, why not you could even have higher bonuses when you perfectly known the game, the synergies betwxeen companions and the mechanics of the game...
I guess now it's time for you to try the tactician mode...

BG3 could play with many rules and parameter of D&D instead of giving us the god mode as the more common ways to increase our % to hit.
This would probably lead to more tactical choices for the players, more ways to increase our % to hit, less AC reduction and less HP bloat, so less issues with D&D...

% to hit should definitely be an important part of the difficulty of a D&D video game.

At the moment you just have to find how to enable your super-over-power so the D20 is not any obstacle anymore.
Not sure that's the best approach if you call your game a strategy TB game.

This pretty much sums up where 5e failed, and what they should correct for the next edition of D&D.

Make the skills matter more.

Higher modifiers, rolling 3d6 instead of 1d20, there are ways.

If BG3 can't do that, what they could do is use thresholds rather than the d20 roll. Good old Take 10. Starting skill levels range from -1 to +5 and those differences need to matter much more. Only roll d20 when under stressful situations when it makes sense even the most skilled character can fail because of circumstances you can't control, or the least skilled can succeed because of adrenaline.

My two cents:
I completely agree with Sawyer. I hate the base concept of the D20 system, and especially that it so heavily depends on RNG. That is why, after having played D&D through four editions for more than 25 years, I have now come to the conclusion that I thoroughly dislike the D&D system/mechanics (but I care deeply about BG3 because I love the original BG games and also love the FR setting). So I am firmly in the camp of supporting deviations from the D&D rules and mechanics wherever and as much as necessary to make the game fun as a video game. So very glad Sawyer rejected using the D&D SRD for PoE, resulting in a fresh, new system that I absolutely love and consider to be way superior to D&D. More developers should be willing to take the intellectual plunge into trying to create new and better cRPG systems instead of just keeping on using D&D.

Three things:

- Is rolling 1d20 annoying in early level 5e - yes, but then again I can't blame Larian for using this mechanic.

- Adding modifiers for missing can be a solution, but I think it is much more fundamental task than tweaking monsters. Plus, might be confusing for players too, whereas rolling 1d20 + prof + skill is fairly straightforward.

- I don't think HP bloat is a thing in BG3, mostly complaints talk about goblins. And as others mentioned the Goblins don't have wildly different HP/AC compared to other monsters at their CR.

Well, what's being discussed here doesn't really seem to be an issue coming strictly from Larian's implementation, as much as a core issue with the ruleset.
And honestly there's no really any meaningful way around its limit except saying "fuck that, let's come up with our better homebrew variant or custom system replacing this one entirely".

The alternative is cheating on the dice rolls, which is even cheaper and patronizing as a solution, frankly.

Every system has its drawbacks just keep that in mind. If you then weigh things towards skills rather than the D20 then encounters become harder to make.

Sure, you could do thresholds in BG3 but then you're going to have to adjust all the skills and make everything that's save or die have threshold partials. Which is fine since it's a computer game come to think about it.

Some creatures just aren't meant to scale, if you replaced them goblins with cultists with PC levels you would see a ton of people just fuck straight off -- is that a fault of 5e?

I wasn't talking about the combat at all, so not sure where "creatures" even come into play here.

This was a discussion about skill checks and how little impact on chances proficiencies and abilities seem to have in this 5th edition.

Quick side note:


Expertise exists in the core rules and there are numerous ways to get it on a variety of things, not exclusive to rogue. If Larian would actually use the 5e rules as published, we'd have a fair amount of that.



The Help action as a bonus action (granting advantage) also exists in the rules and is available to a few different archetypes of different classes. If Larian would actually use the 5e rules as published, we would eventually have several methods of that available.



Seems like there's already plenty of these kicking about for a 1-4 group of adventurers so far?



Larian's high-ground darling really needs to go away and never come back. Implementing proper cover rules would simulate bonuses for high ground where having said high ground would actually provide a benefit, and it would also do more and add more to the game as a whole, in many situations, while being a relatively simple and mechanically straightforward concept to understand and identify.

That aside,

I'm finding myself wondering what sorts of DMs you other folks have had in the past, to leave you with view points like this...

- A level one character who is proficient, or even expert, in a particular skill is better, on the whole, in that area than someone who isn't, but they're not *That* great at it yet. No, they should not be a *lot* better at those skills at level one; they're level one. They've just got a good start ahead of most average others. At level one, if you've got a good stat to back up your good skill, you'll be a +6; at level one, you generally shouldn't be facing any checks much above a 12, at least not often. If you are facing a 15 or above at that level, then you're trying something that's actually quite difficult for someone of your limited experience, and yeah, you can expect to not nail it half the time. Welcome to level one? You're just starting out? Don't expect to be fantastic even at the things you're training yourself towards at this point? Expect to be competent at them, dare I even say, proficient at them, and better than your contemporaries... but not a master, or everyone's go-to gal. In a training room with 20 "my first lock" puzzles (DC 10), the level one wizard, who understand the concept of lock picking, but doesn't know a cinching pin from hair clip when presented with the tools, might be able to fumble his way through ten of them, or so. Meanwhile, a level one rogue who has learned how to use their tools properly, and is nimble of finger (18 dex, thief tool expertise, +8 bonus), might expect to open nineteen of the twenty locks in the same time... probably all twenty, but mistakes can happen. 5e's bounded statistics and the d20 are working just fine.

Hey, you know what... I just ran that experiment five times, using physical dice... the 'rogue' got all 20 locks four times, and got nineteen one time, with a single natural 1. The 'wizard' got 9 locks, followed by 13, 8, 7 and 14 locks on the fifth set - he rolled slightly above average expectations overall. But I don't know what else to say, other than that the system does what it's intended to do... this all feels pretty correct and satisfying to me.

If Bumblefoot Mugbreaker needs a 16 to sneak past the guards, and has 8 Dex and no proficiency, then he's going to fail most of the time, but sometimes (4/20, 20%) he'll luck out and not tip them off, provided his club foot isn't giving him disadvantage on stealth checks to move silently... it probably should be... but if we figured that in, his success rate drops to almost nothing.
If Darkstalker Shadowcloak has been training himself in the arts of stealth for years... then he's probably not a level one character... but that aside, if he IS level one, then he's going to be decently dexy and is apparently a self-proclaimed expert. Let's cut it right back, the bare minimum for edgelording, and call it a 16... then he will succeed more often than not (12/20, 60%). It's a notable margin. It's not an insignificant margin, in fact. It's three times the chance, at level one. It's not 10%, certainly.

The sheer magnitude of skill numbers in older editions were the exact opposite of fun - they meant that, as with all contested arms races, right off the bat, the DCs for things would shoot so high that no-one except super-specialists would have any hope at all of succeeding. It discouraged people from trying anything outside of their very narrow, specific wheelhoues, and it forced players into wedging themselves into said wheelhouses, to the complete exclusion of almost everything else... because if they *Didn't*, then *No-one* would be able to do it at all. Let's NEVER go back to that, thanks.

In an average level party (let's say 9th level; the experts are really starting to seem like genuine experts at a world-class level by this point), that group stealth check goes something like this: The rogue rolls a 3, and has a final check of 16 (presuming no magic items). The clumsy one with 8 dex *Needs* to roll 17 or over (an excellent attempt and a slim chance at best) just in order to do as well as the rogue when she *screws up*. That sounds right to me; where's the problem?

How often has the 8 Int barb come up with super specific arcana knowledge while the wizard is left scratching her head? Well, for my games... Never. Here's why:

- Skill checks are supposed to represent situations where there is a realistic chance of failure or success. IS there a realistic chance of the barbarian with 8 int and a backstory that involves growing up getting hit with rocks might know the lore needed while the wizard doesn't? No? Then the barbarian doesn't roll arcana here. That's how it's meant to work. Players and DMs who regularly request skill checks over ridiculous things where it defies sense for at least one of the outcomes to occur need to take a step back and ask why they're doing a check there, and then not.

- An extra one or two makes a lot of difference, especially with attack rolls, in 5e. My variant human ranger started her adventuring career off with a +9 to her ranged attack rolls... compared to the warlock's +6 and the cleric's +5, it made a positively phenomenal difference over those first 3 levels. She virtually never missed, was happy to shoot with disadvantage, and it was so much of a sure shot thing that it felt completely broken and over-powered for the system. The small numbers are worth a lot; it's a good thing that most people's attack bonus, whatever their method, all generally start within one or two of each other, at least for those character who intend to be using that attack bonus most of the time.

If it 'feels bad' for you, to only be a *Little* better than everyone else at your chosen specialisation at *Level 1*, that's not the fault of the game, or the d20... that problem lies between chair and character sheet. If you are repeatedly and consistently failing checks that you have a high bonus for, and regularly succeeding in checks that you don't think you ought to have been able to succeed at... that's a DM problem, not a problem of the system or the d20 (maybe the d20, if it's the same one and providing such a predictable result - maybe float it and check).

No, definitely not. The high ground adds a substantial expansion to the combat playing field and so substantially increases the diversity of tactical considerations and options inside every combat encounter.

That cannot be substituted by the boring old cover system which would also demand complete rework of the environments in the whole game - for what? Dramatic increase of convenient boxes and barrels to hid behind?
The high ground advantages and disadvantages need adjustments, nothing else.