Sword Coast Stratagems v34 (edit: 34.3) now available

[Bartimaeus]

2 hours ago, Guest guestMM said:

According to the wiki (https://baldursgate.fandom.com/wiki/Saadat) Sadaat has spellcasting of Chromatic Orb ,Magic Missile (×2), Melf's Acid Arrow ,Fireball ,Flame Arrow, which is pretty much equivalent to a PnP Rakshasa. Giving a 3rd level spellcaster level 6 spells is a pretty huge balance change and giving a 9HD creature level 6 spells and an 11 HD creature level 7 spells when Wizards can't even cast them until 12th and 14th level is pretty horribly inconsistent no matter what ruleset you are going by.

The wiki is inaccurate, in a sense. DavidW was careful with his words, and for good reason: Saadat is scripted to use Death Fog, a sixth level spell. His creature file may not have any sixth level spells memorized (there are multiple in his spellbook, but none actually memorized), but his script allows him to cast both Cloudkill and Death Fog even without them being memorized, which you can test out yourself in the vanilla game. If you check out the Ihtafeer page on the same wiki you just linked, it mentions them all of the Rakshasas casting Cloudkill and the fight being "surprisingly difficult".

Edited July 8, 2022 by Bartimaeus

[Guest guestMM]

I don't currently have enough space for a separate install, so I'll take your word on it. Still seems like a bad Idea to give level 9 enemies with normal weapon immunity and breach immunity protection from magical weapons. It seems inconsistent with the mission statement of monsters fighting by the rules not being granted illegal powers or abilities they didn't have before.  

This is also I assume just part of the general "smarter mages" component that gives Rakshasa these custom spellcasting abilities as I don't see a separate option where I can turn off their cheat custom abilities like I can for Fiends/Celestials.

[DavidW]

5 hours ago, Guest guestMM said:

According to the wiki (https://baldursgate.fandom.com/wiki/Saadat) Sadaat has spellcasting of Chromatic Orb ,Magic Missile (×2), Melf's Acid Arrow ,Fireball ,Flame Arrow, which is pretty much equivalent to a PnP Rakshasa.

Don't always trust the wiki.

If you look at the actual creature and its scripts (as you do later), Saadat can cast Death Fog, so he is at minimum L12. SCS makes him L13, I think as a legacy of early versions that looked at his actually-memorized spells (which in principle show up in game if you charm him). I'm not averse to the argument that he should be L12 rather than L13.

5 hours ago, Guest guestMM said:

If we are looking at RAKRUH01 specifically for our justification a single cast of death fog is pretty big difference in balance versus 2 casts of PFMW both in terms of spells known and its impact on a creature that is immune to breach by default.

1 cast vs 2 is L12 vs L13, and as I say, I can see the argument there. But on the broader issue: yes, it is way more sensible for a creature with strong immunities and L6 spells to learn PMW than to learn Death Fog, and yes, that makes the encounter way harder. But the whole point of SCS's Smarter Mages component is to (a) work out the level of spellcasting a creature has based on in-game cues and (b) to make actually-intelligent choices of spells and tactics within that constraint. If you don't want a component that does that, I can't really suggest anything except 'don't use this component'.

5 hours ago, Guest guestMM said:

it seemed more or less intentional that they were fighter/mage types rather than pure casters like liches

Agreed. And they are: they don't have speciality bonus spells and they like fighter-mage spells. But fighter-mages should absolutely use PMW.

5 hours ago, Guest guestMM said:

otherwise they should at least have their level increased so they aren't casting illegal spells according to BG2 rules

They do (or at any rate they should have). SCS infers a creature's caster level from their in-game displayed activity and sets their level accordingly. (But notice that vanilla BG2 does not do that: Saadat casts Death Fog despite being level 9.)

5 hours ago, Guest guestMM said:

As an aside, since you rebalance many spells in your mod, have you considered that in terms of balance it would make more sense if we got mantle at 6th level, improved mantle at 7th, PFMW at 8th, and Absolute Immunity at 9th? PFMW is one of those anomalous spells that forces the rest of the game to play around it, and it doesn't make sense that it should be lower level than mantle in the first place

There is a fairly good case for doing this if you were designing a spell system ab initio. But SCS doesn't aim to systematically rebuild the spell system, and the existing framework works. I actually like the idiosyncracies of the BG2 system, it leads to interesting emergent behavior. (It's also the case that the existing protection system is so deeply built into SCS's architecture that it would be a nightmare to change now.)

 

[DavidW]

3 hours ago, Guest guestMM said:

Still seems like a bad Idea to give level 9 enemies with normal weapon immunity and breach immunity protection from magical weapons. It seems inconsistent with the mission statement of monsters fighting by the rules not being granted illegal powers or abilities they didn't have before.  

I missed this. But they're not being granted 'illegal powers or abilities they didn't have before'. They have the ability to cast 6th level spells in the vanilla game. They're just making smarter choices of spells.

I should also have mentioned that if you want a lower-level Saadat, edit stratagems/mage/override/level.2da, and set 'trrak02' to whatever level you like. (You need to do this before installing Smarter Mages.)

[Guest guestMM]

Quote

They do (or at any rate they should have). SCS infers a creature's caster level from their in-game displayed activity and sets their level accordingly. (But notice that vanilla BG2 does not do that: Saadat casts Death Fog despite being level 9.)

This is incorrect. SCS only assigns higher level spellbooks to all the rakshasa in the game and does not adjust their level to match. If you install it Saadat will still be level 9 with a level 13 spellbook, 4/6 Rukhs will still be level 11 despite having level 13 spellbooks. Ihtafeer similarly has a level 15 spellbook despite still being coded as level 11 for hit dice and caster level.

Quote

But on the broader issue: yes, it is way more sensible for a creature with strong immunities and L6 spells to learn PMW than to learn Death Fog, and yes, that makes the encounter way harder. But the whole point of SCS's Smarter Mages component is to (a) work out the level of spellcasting a creature has based on in-game cues and (b) to make actually-intelligent choices of spells and tactics within that constraint. If you don't want a component that does that, I can't really suggest anything except 'don't use this component'.

There's a ton of issues here with this mindset, the first being of course, that this mod treats the same as every other mage, rather than as magical creatures. You can't manually turn off their new behavior while using the rest of SCS, unlike how you can with fiends/celestials. The second is that it forces degenerate behavior and cheesy tactics rather than making for interesting encounter design. Having a creature that attacks you with deadly AoE that it is immune to makes for a more interesting battle than one where you wait outside the building for a few minutes because you don't have 8th level spells yet. The third is that it's arguable whether it's "sensible" for a wizard to dedicate their entire spellbook to PFMW in the first place. It only really makes sense if your only enemy is a low level player party and only if your goal is to drag the fight out against low level players. Improved Haste would make them a better hunter. True sight would help them deal with invisible creatures (currently their AI just has them stand around for 5 rounds doing nothing if you go invisible, then they'll start by casting detect invis and glitterdust both of which are very easy to counter). Protection from Magical Energy would protect them from horrid wilting (which is usually how you deal with the spellhold/suldanessellar ones since anything else is an exercise in tedium). For a group of hunters waiting in a hut to ambush people, Death Fog even in an enclosed area adds a lot of danger to the fight. And nonR

Instead, we get stupid boring encounter design. Our best answer to this encounter is to just cast invisibility. Wait 4 rounds, pop their next PFMW, cast invis again, wait 4 more rounds. Then fight the encounter normally. It's stupid. It's boring.  There's no pile of AoE fog for us to be afraid of. They don't even like to use the AoE spells they have memorized until they go through all their single target spells, despite their immunities to them. There's no "wizard fencing." There's no counterplay. Instead we just fight degenerate tactics by abusing their bad AI and lazy spell selection because that's the only solution at the level you encounter this quest. The original encounter is more unique, more exciting, and more dangerous than the current SCS iteration.

I think there's a lot of room to explore enemy AI but SCS Rakshasa (particularly these 3) just come across as lazy.

[DavidW]

Thank you for the feedback.

[subtledoctor]

Stepping back from the charged rhetoric for a second, I do find compelling the idea that SCS has perhaps made some overly generous assumptions about rakshasa here. The fact that a script contains one 6th-level spell, which deviates from the written source material, could be inferred to mean the creature should act like a 12th-level caster (or, since they are functionally fighter/mages, a 12th/12th level F/M which is equivalent to a 13th-level creature) instead of their stated 9HD or 11HD. 

However, that is taking one deviation from the source as justification for adding several more deviations. It seems equally reasonable to interpret that scripted spell as a singular high-level special ability granted to a creature which otherwise remains 9HD or 11HD and has fairly limited spellcasting. There are oodles of creatures like this in the game, and in AD&D generally. 

I personally think this is a better interpretation of how those creatures are coded. They are after all holed up hiding from powerful genies; they have had time to fortify their position and could bring to bear a powerful magical ambush (Death Fog, Cloudkill) that they would not otherwise have access to and which is not in their specialty (rakshasa are traditionally illusionists). This could explain why those spells are in the script while the .CRE file hews more closely to the PnP source. But whether this more conservative take would make for a better encounter, gameplay-wise, is of course subject to debate. Mods being mods, I think the best way for something to be done is the way the mod author prefers it.

I will say, I think pointing to the ability to change creatures’ level via that .2da table is not an entirely reasonable solution. It’s nice that that table is there, but 1) it is very hard to find without specific direction, and 2) through no fault if your own, BG2 .CRE filenames are pretty opaque and I doubt my own ability to correctly guess which ones encompass the set of rakshasa. And I’m someone who has delved into the game files a lot more than most. A more typical mod-using player should not be expected to sort that out. Which is to say “this is the way I prefer it” is a more reasonable reaction than “you can change it if you want” (because most people in fact cannot, notwithstanding your admirable externalization of that table). 

Pie-in-the-sky: given the prominence of these encounters and the more radical changes that SCS makes to rakshasa compared to other creatures, it would be lovely to have a more generally accessible option to decide whether SCS should set up make rakshasa as high-level casters or low-level casters. Something reactive to a fine-grained difficulty option would be over the moon. But that would take work and time, and I don't think you owe that to anyone. 

(Neither here nor there: I’ve pondered modding rakshasa myself to try to make them into enemies who are dangerous and fun while stating focused on illusion magic instead of general BG2 fighter/mage tactics. But I’m leery of putting in the effort as most players will install SCS afterward and SCS would obliterate any such changes…)

Edited July 9, 2022 by subtledoctor

[Guest guestMM]

3 hours ago, subtledoctor said:

snip

Thanks, I think every point of this is a probably better worded version of what I was trying to say. I also think the flavor you give the encounter is an interesting interpretation. Maybe they had a wand of cloudkill or a death fog trap that couldn't easily be coded since they initially start out as a shopkeeper.

One thing I want to expand on is how one of my big concerns is that editing the .2da table is really not an elegant solution.

Since SCS edits their AI as well as their spellcasting, and editing their .2da tables just changes their spellbook to *another* SCS spellbook that looks nothing like their original ones. (They'll still blow through all their chromatic orbs before even attempting to AoE, for example). And if I want to make my own rakshasa mod, SCS overwrites that, and there is no toggle option like other magical creatures. I think the toggle options for beholders/fiends/vampires/etc are great since those aspects can be more controversial than some of the others.

It is also missing several rakshasa throughout the game, so it results in inconsistencies where some rakshasa are still coded as level 9 mages or level X fighter/mages, where others are coded as level X Rakshasa with level X+4 wizard spellbooks.

(and this part will sound less polite I guess but “this is the way I prefer it” seems to be the stance of SCS on a lot of things and sometimes I just wish it would admit that on the documentation/advertising. It represents itself as solely an AI mod that is completely different than Tactics/Anvil/etc, but it takes a *lot* of liberties with rebalancing things, adjusting levels, or giving creatures unique abilities. I think a lot of them are cool design decisions but they stray from pure AI more than someone installing it for the first time might expect). I obviously understand modding isn't a job and I'm not paying him, so any rudeness should be interpreted as a lack of elegance in my wording, rather than an entitled demand for change.

Though if someone is looking for a mod idea, I think it'd be interesting to see someone put the same effort into warriors try their hardest that SCS does to mages. There are a lot of ways a warrior could better equip or "prebuff" himself like how SCS mages do and could be interpreted as potion/scroll/amulet use or weapon breakage the same way SCS mages can spend time prebuffing with shield/mirror image/etc. 

[DavidW]

20 hours ago, subtledoctor said:

Stepping back from the charged rhetoric for a second, I do find compelling the idea that SCS has perhaps made some overly generous assumptions about rakshasa here. The fact that a script contains one 6th-level spell, which deviates from the written source material, could be inferred to mean the creature should act like a 12th-level caster (or, since they are functionally fighter/mages, a 12th/12th level F/M which is equivalent to a 13th-level creature) instead of their stated 9HD or 11HD. 

However, that is taking one deviation from the source as justification for adding several more deviations. It seems equally reasonable to interpret that scripted spell as a singular high-level special ability granted to a creature which otherwise remains 9HD or 11HD and has fairly limited spellcasting. There are oodles of creatures like this in the game, and in AD&D generally. 

OK, I've done a deep dive into this (certainly much deeper than SCS originally did back in the day, when its automated system just saw they had mage8d.bcs, said 'eh, they're mages', and got on with it.) TL;DR: I still think they're fighter-mages, although I agree that it's a bit messy.

Two general observations:

1) A general observation: BG2 is a huge game that notoriously had scope-creep issues, and a lot of its creature files and creature scripts were clearly written in a hurry. Level and known spells are really unreliable, often the result of a quick copy-paste from some other creature. Tweaks to spellcasting are often done on the fly by ForceSpell, and you can't infer that something is an innate power just because it's ForceSpelled.

2) A corollory of (1) is that spell choice and use in unmodded BG2 is extremely narrow. A core part of SCS is trying to broaden that and get spellcasters to memorize a more interestingly large range of spells. (That said, if something is an innate ability rather than a spell, of course I want to leave it alone at least insofar as SCS's AI components are concerned.)

Moving on to Rakshasas: There are only really three templates for Rakshasas in BG2.

(i) The most common (the baseline Rakshasa) casts spells of levels 1-3 through normal scripting, borrowing an off-the-shelf mage script; it also force-casts either Death Fog or Cloudkill. Sometimes you meet versions of this with more combat ability but the power system doesn't change. 

(ii) Then there is a supercharged version of the Rakshasa, usually but not always called 'Maharajah' (Ihtafeer is this, for instance). This is scripted to cast a bunch of spells: chain lightning, cloudkill, death fog, ice storm, minute meteors. It also has an off-the-shelf mage script but I think due to script interactions it doesn't actually do anything. (This is what I mean about creature scripting being hurried.)

(iii) Finally there is a group of rakshasas in one specific encounter in Suldanesselar who are scripted to cast a lot of very powerful spells: Mislead, Pro/MW, Stoneskin, Shadow Door, Sunfire, Symbol: Fear, etc, etc.

Now looking at PnP (which isn't decisive, but which SCS uses both for inspiration and for a fallible look at developer intent): in 2nd edition AD&D baseline rakshasas are granted a somewhat ad hoc number of spell levels (including a few priest levels), and higher-level rakshasas are distinguished from the baseline ones basically by getting more spells at higher levels. In 3rd edition D&D rakshasas have sorcerer class levels and more powerful ones have more levels. Note that in both versions (and in all versions of D&D I know) rakshasas have the ability to cast mage spells; they don't have innate spell-like abilities.

Putting all that together: I want to stick to the view that it is *at least* defensible to interpret BG2's Rakshasas as having mage levels and using mage spellbooks. You could try for another interpretation but I don't think it's nearly as natural; certainly it's not *more* natural.

20 hours ago, subtledoctor said:

I personally think this is a better interpretation of how those creatures are coded. They are after all holed up hiding from powerful genies; they have had time to fortify their position and could bring to bear a powerful magical ambush (Death Fog, Cloudkill) that they would not otherwise have access to and which is not in their specialty (rakshasa are traditionally illusionists). This could explain why those spells are in the script while the .CRE file hews more closely to the PnP source.

It's an interesting observation - but note that it's not just the Trademeet rakshasas who get it. More than half the rakshasas in BG2 have the identical scripting block.

20 hours ago, subtledoctor said:

I will say, I think pointing to the ability to change creatures’ level via that .2da table is not an entirely reasonable solution. It’s nice that that table is there, but 1) it is very hard to find without specific direction, and 2) through no fault if your own, BG2 .CRE filenames are pretty opaque and I doubt my own ability to correctly guess which ones encompass the set of rakshasa. And I’m someone who has delved into the game files a lot more than most. A more typical mod-using player should not be expected to sort that out. Which is to say “this is the way I prefer it” is a more reasonable reaction than “you can change it if you want” (because most people in fact cannot, notwithstanding your admirable externalization of that table).

It's not really intended as a solution. "This is the way I prefer it" - or perhaps better, "this is my preferred interpretation of how to understand the abilities creatures display in-game" - is always the answer, ultimately. The level-override table is externalized because code/data separation is good practice, not primarily to help players customize the game. That said, I always try to point out these options as 'consolation prizes' for people who disagree with my interpretations.

21 hours ago, subtledoctor said:

Pie-in-the-sky: given the prominence of these encounters and the more radical changes that SCS makes to rakshasa compared to other creatures, it would be lovely to have a more generally accessible option to decide whether SCS should set up make rakshasa as high-level casters or low-level casters. Something reactive to a fine-grained difficulty option would be over the moon. But that would take work and time, and I don't think you owe that to anyone. 

It's worth considering. (It doesn't exist at present because I interpret rakshasas as fighter/mages and so they fall under the scope of 'Smarter Mages', as opposed to, say, demons, which have a completely different interpretation.) The problem is more technical than anything else - it might be a pain to disentangle rakshasas from the rest of Smarter Mages.

21 hours ago, subtledoctor said:

Neither here nor there: I’ve pondered modding rakshasa myself to try to make them into enemies who are dangerous and fun while stating focused on illusion magic instead of general BG2 fighter/mage tactics. But I’m leery of putting in the effort as most players will install SCS afterward and SCS would obliterate any such changes…)

SCS detects mages by script. If your mod removes all the standard combat scripts (mage8d, raksha02, etc) from a rakshasa and uses bespoke scripts, SCS will skip it. (Note that it ignores the rakshasas in the Black Pits II.)

[DavidW]

17 hours ago, Guest guestMM said:

They'll still blow through all their chromatic orbs before even attempting to AoE, for example

That's not anything specific to rakshasas, and isn't a priority thing. The standard SCS spell priority prefers (most) area effect to Chromatic Orb, so if you have Rakshasas who have area effect memorized and are casting CO instead, they can't find a good target for area effect. That may be a bug but I'll need a much more specific report to check: let me know exactly which AoE spells in the spellbook (checked from NI) aren't being used. Area effect is very fiddly in IE scripting in general because of friendly fire: Rakshasas are supposed to be exempt from that but it might not be working fully.

 

17 hours ago, Guest guestMM said:

And if I want to make my own rakshasa mod, SCS overwrites that, and there is no toggle option like other magical creatures.

See my comment to SubtleDoctor, above - you can get around this very straightforwardly.

17 hours ago, Guest guestMM said:

It is also missing several rakshasa throughout the game, so it results in inconsistencies where some rakshasa are still coded as level 9 mages or level X fighter/mages, where others are coded as level X Rakshasa with level X+4 wizard spellbooks.

If you've seen specific missing cases, please do report them. (But note that these inconsistencies are all over unmodded BG2 in any case, so it's not as if I'm breaking something antecedently working).

17 hours ago, Guest guestMM said:

“this is the way I prefer it” seems to be the stance of SCS on a lot of things and sometimes I just wish it would admit that on the documentation/advertising. It represents itself as solely an AI mod that is completely different than Tactics/Anvil/etc, but it takes a *lot* of liberties with rebalancing things, adjusting levels, or giving creatures unique abilities.

There are two issues here. I want to stick to the claim that SCS's AI components do what they say: i.e., try to interpret what abilities the creature and then use those abilities more intelligently. I don't think I add unique abilities in the 'improved AI' section of the mod except in a few places that are explicitly called out in the documentation (fiends, most notably). BUT if you start actually trying to implement that policy, you find that the original game is absolutely full of inconsistencies (L8 mages casting L9 spells, say) so that even a 'smarter' mod has to make lots of judgement calls. I'm always happy to get feedback or suggestions on those judgment calls but ultimately it's not a democratic process: if I've listened to feedback, thought about it, but still disagree, I'm not going to change it. In that sense, it is indeed 'this is the way I prefer it', though (as I said to SD above) 'this is my preferred interpretation' is closer to how I think about it.

Of course there is a *much* more minimal interpretation of 'smarter AI' that just takes *exactly* the set of abilities and spells a creature displays in-game and just scripts them better. SCS is not that mod and I think writing it would be a frustrating experience given how slapped together much of the vanilla game's allocations of spells and scripts are. I think SCS's documentation is perfectly explicit that 'smarter choices of spells' is part of its conception of 'smarter'.

[Guest guest]

  

5 hours ago, DavidW said:

That's not anything specific to rakshasas, and isn't a priority thing. The standard SCS spell priority prefers (most) area effect to Chromatic Orb, so if you have Rakshasas who have area effect memorized and are casting CO instead, they can't find a good target for area effect. That may be a bug but I'll need a much more specific report to check: let me know exactly which AoE spells in the spellbook (checked from NI) aren't being used. Area effect is very fiddly in IE scripting in general because of friendly fire: Rakshasas are supposed to be exempt from that but it might not be working fully.

That specific encounter. I'm sitting there in the room invisible. I use summon monsters. It starts by sitting around wasting rounds casting any spell protections not covered by contingencies/prebuffing while trying to melee them, then it starts casting chromatic orb/remove magic/magic missiles on them, then when those run out it will finally go to cone of cold, then when it runs out of casts of that it will finally use sunfire. It never used ice storm or death fog at all, and it never threatened anything invisible by using AoE damage like a tactically minded player would do to deal with thieves (much quicker to kill them with a sunfire than bothering to burn through double-invis). Their AI would be completely unprepared to deal with an SCS inspired thief who goes invis stab invis.

Quote

If you've seen specific missing cases, please do report them. (But note that these inconsistencies are all over unmodded BG2 in any case, so it's not as if I'm breaking something antecedently working).

It is a system issue where anything with Class [Rakshasa] is given X+4 wizard spells while any Race [Rakshasa] with player class levels like fighter or mage are given a legal spellbook. It applies mostly to ones added by EE, but there's others as well (Actually looking deeper at it, it seems Rakshasa are given pretty arbitrary levels here. Some level 9 ones are listed as level 13 in level.2da but still only have level 3 spells memorized, others are given +0, +2, or even +4 levels in their spellbook somewhat randomly in accordance with the level.2da file, but the 2da file still includes systemically based on class rather than race). Also a weird oddity is the Ruhk Transmuter who has abjuration magic because he is coded as a [Rakshasa] despite being a Transmuter by name. For some reason he is also given a level 13 wizard spellbook despite being level 18 (but still has comet memorized which I'm going to guess is the override file not handling HLAs correctly). This seems kinda silly when he by default has a much larger spellbook than the level 11 Adratha/Ihtafeer who is boosted to a level 15 spellbook and given level 7 magic she never knew in the base game.  I assume he was "nerfed" as some sort of balance decision and she was boosted because SCS assumes the Rakshasa class is wizard +4 rather than going by the spells known by each one in general.

 

Re: Interpretations: I think there's a pretty broad spectrum you can have between "exact spellbook" and "Improved Fiends/Celestials." *Most* of it is pretty well-documented and modular.  In this instance I think looking at a specific scripted ability of one encounter and extrapolating it to a broader casting boost to an entire race of creatures is a bit of a stretch. On a more nuanced level I think there are some instances where we can ask "Is there a reason literally every spell in this creature/mage's spellbook is evocation?" (which would be the case for Ruhks/Ihtafeer in the base game, but not the transmuter). If we want to get pickier, we can ask "where did all these mages find/buy all these scrolls?" which may be a bit much, but I *do* kinda think there's a philosophical reason that we can buy say, freedom and symbol stun in the adventurer's mart straight out of Irenicus's Dungeon, but Horrid Wilting, Time Stop, and Spell Trigger are much rarer (I don't expect SCS to ban spell trigger, but I mention it as an example because in Forgotten Realms Lore there was exactly *one* mage who knew the spell). Obviously we're going to houserule it so the PC and his party can learn it in our fancy new computer game, but some spells are rarer than others and hitting level X doesn't mean you can necessarily find any scroll you want right away or at all. Sorcerers obviously break this logic, but I'm not even sure they existed in 2E, there's no DM to tell us "no," and I can't think of any sorcerer NPCs in the original game off the top of my head anyway. IIRC they were just an advertisement for the upcoming 3E in the first place.

There are also a lot of balance changes enabled by default in the install made seemingly because the AI does not know how to handle the original behaviors particularly well or instances where the AI really likes newly buffed spells/abilities despite them not particularly being useful for the player. I do think there is a general issue with the mod where mages can sometimes *overly* prioritize on defense to the point where they lack the ability to present a threat to your party, but that's mainly a consequence of using blanket personalities rather than hand-picking the spells of every mage and applying custom scripts for encounter design, and even then vanilla BG2 probably had similar issues more often overall than SCS did. Sometimes they can also end up feeling a little monotonous by the end of the game, and if I were a better and more ambitious modder, I might attempt something like a spell randomizer, though that's more of a shower thought than a serious suggestion. It probably wouldn't play well with the AI anyway.

 

But my issues with the mod itself are pretty minor outside of raks. The premise of my frustration in terms of interpretation/representation was less about the actual detailed documentation, or the ability to pick modular components, or the design decisions themselves, but the chain of logic: "SCS is fair and only AI! There's even a 14 paragraph intro saying how it is more fair than every other AI mod (slight exaggeration here)" - > "Smarter Dragon/Beholders/Celestials/Improved encounters/whatever are part of SCS so they must also be fair despite the fine print" that isn't explicitly said but is an implicit assumption of a good chunk of community using the core/mage/priest AI as a motte and bailey defense where everyone talks about SCS as if its just the core components of the install when it comes to fairness, but in any discussions otherwise its assumed that you're talking about the mod as a whole. When we talk about SCS it plays fair and is AI only, but every exception is just a single exception, despite the exceptions probably making up more than half of the mod on a full install.

For reference: I *do* use all the improved encounters mods and most optional components (yes dragons, no fiends/beholders) when I play SCS, and I think most of them are fun. I just think it's really silly how we have discussions where we try to pretend we aren't giving creatures special powers like Tactics/IA/whatever saying silly things like creatures don't use powers available to the player and aren't given "uninterruptible spells, improved casting time, free Alacrity, or the like" when if we page down a few times there's even a component specifically to give instant, uninterruptible casting to dragons. I don't really think it's an issue with the mod itself as much as the culture surrounding it.

[DavidW]

Picking up a few bits of this:

1 hour ago, Guest guest said:

That specific encounter. I'm sitting there in the room invisible. I use summon monsters. It starts by sitting around wasting rounds casting any spell protections not covered by contingencies/prebuffing while trying to melee them, then it starts casting chromatic orb/remove magic/magic missiles on them, then when those run out it will finally go to cone of cold, then when it runs out of casts of that it will finally use sunfire. It never used ice storm or death fog at all, and it never threatened anything invisible by using AoE damage like a tactically minded player would do to deal with thieves (much quicker to kill them with a sunfire than bothering to burn through double-invis). Their AI would be completely unprepared to deal with an SCS inspired thief who goes invis stab invis.

That's nothing to do with rakshasas in particular, it's a general (and legit) criticism of SCS AI. How to handle summons when the player is invisible is fairly complex and this looks like a place where it could be improved.

1 hour ago, Guest guest said:

Re: Interpretations: I think there's a pretty broad spectrum you can have between "exact spellbook" and "Improved Fiends/Celestials." *Most* of it is pretty well-documented and modular.  In this instance I think looking at a specific scripted ability of one encounter and extrapolating it to a broader casting boost to an entire race of creatures is a bit of a stretch.

I accept there are judgement calls here; that said, if we're talking about Death Fog/Cloudkill, it's not 'one encounter', it's literally every base-level Rakshasa in the game.

1 hour ago, Guest guest said:

On a more nuanced level I think there are some instances where we can ask "Is there a reason literally every spell in this creature/mage's spellbook is evocation?" (which would be the case for Ruhks/Ihtafeer in the base game, but not the transmuter).

Unmodded BG2 spell books are usually very unreliable as an indication of the resources actually used by a creature, and very often are just cut-and-pasted from a small number of templates. More powerful rakshasas, in particular, largely ignore their spellbooks and force-cast from a script.

That said, I do agree that vanilla-game rakshasas throw around indiscriminate amounts of AoE (to which they are immune) a lot and that SCS doesn't replicate that. I think that's a good thematic suggestion that I'll consider (although the time cost is not trivial).

1 hour ago, Guest guest said:

Actually looking deeper at it, it seems Rakshasa are given pretty arbitrary levels here. Some level 9 ones are listed as level 13 in level.2da but still only have level 3 spells memorized

Again, spellbooks are a bad indicator here. SURAKW1-SURAKW4, for instance (I suspect that's the example you have in mind?), only have spells memorized at levels 1-3 - but in the unmodded game they completely ignore their spellbook and force-cast spells at levels 6, 7 and 8. Setting them to level 13 is a bit undergenerous if anything - their displayed abilities make them at least L16. I don't specifically recall why I did that: probably a judgement call that it was too much of a difficulty spike.

1 hour ago, Guest guest said:

Also a weird oddity is the Ruhk Transmuter who has abjuration magic because he is coded as a [Rakshasa] despite being a Transmuter by name. For some reason he is also given a level 13 wizard spellbook despite being level 18 (but still has comet memorized which I'm going to guess is the override file not handling HLAs correctly). This seems kinda silly when he by default has a much larger spellbook than the level 11 Adratha/Ihtafeer who is boosted to a level 15 spellbook and given level 7 magic she never knew in the base game.

Again, the spellbook is a very bad indicator of the actually-displayed abilities. If you look at what the Ruhk Transmuter does in-game, his scripting will only cast Death Fog and Cloudkill - that fat spellbook doesn't get used at all. (I tend nonetheless to pay a *bit* of attention to large spellbooks and official level, which is why the Ruhk Transmuter isn't just treated as a baseline transmuter - judgement calls again.

The Ruhk having Comet is a straight bug. At a guess, my HLA-assignment algorithm is looking at the original level (18) and not the adjusted level.

A Ruhk *transmuter* having *abjuration* spells is more of a scope limitation. SCS only works out how to stat four speciality classes (Conjurer, Necromancer, Invoker, Enchanter), and in addition can't handle speciality fighter-mages. Ideally I'd have a genuine bespoke transmuter script and spell allocation for the Ruhk but it would be a lot of work and it's never been a priority over other projects (and my day job!)

As for Ihtafeer, IIRC SCS's original algorithm looked at her spellbook and saw that she knew an 8th level spell, and so went with L17; when rebalancing I made a judgment call that it was a bit high given she's not actually scripted to use her 8th level spell, and bumped it down to L15. These are inevitably quite fast judgement calls (this is a hobby for me) and if you wanted to argue for a lower level still I would see the case (though I think the minimum is L13 given that unmodded she can cast up to 2 L6 spells). 

2 hours ago, Guest guest said:

There are also a lot of balance changes enabled by default in the install made seemingly because the AI does not know how to handle the original behaviors particularly well

Yes, absolutely, though it's more that limitations of the IE scripting language mean that it's not really possible (or at least: I don't see how it's possible) to handle the original behaviors particularly well. The readme is pretty open about this, I think.

2 hours ago, Guest guest said:

If we want to get pickier, we can ask "where did all these mages find/buy all these scrolls?"

My assumption has always been that since most NPC mages have had years or decades (for liches, centuries!) to fill their spellbooks, they're less constrained than the player, who only has weeks or months. Certainly that could be challenged, but at some level I don't think adding a restriction like that would make things more interesting. I suppose in a sense I simulate it by using quite a wide range of spells even when some are clearly less good than others (I don't think the Bigby's Hand spells are really competitive, for instance, but I use them because variety is good). There's less variety with defensive spells (not none, though) partly on technical grounds: the code for sensible handling of mage defenses is already hideously complicated to keep track of.

If we're talking about what counts as 'playing fair', though, I don't see any evidence that the original game respects a principle like this: practically every appropriately-levelled enemy wizard can cast Spell Sequencer, for instance.

2 hours ago, Guest guest said:

I do think there is a general issue with the mod where mages can sometimes *overly* prioritize on defense to the point where they lack the ability to present a threat to your party

It's a perennial problem to balance offense and defense here. Get it wrong in one direction and mages can't actually do any damage; get it wrong in the other direction and they die like flies. The optimal balance depends on the particular tactical situation (which you're right, automated code isn't very sensitive to) but also on the player's level, party mix, and preferred playstyle (which obviously I can't anticipate). I've tweaked this back and forth many times; it's been a while since I've personally had time for an extended playthrough so it's possible the overall mix has got out of sync.

 

2 hours ago, Guest guest said:

I just think it's really silly how we have discussions where we try to pretend we aren't giving creatures special powers like Tactics/IA/whatever saying silly things like creatures don't use powers available to the player and aren't given "uninterruptible spells, improved casting time, free Alacrity, or the like" when if we page down a few times there's even a component specifically to give instant, uninterruptible casting to dragons.

In fairness, what the readme actually says is:

 I've made a fairly sustained effort to ensure that SCS opponents fight by the rules and don't use powers denied to the player (other than obvious things like dragons' breath weapons, liches' ability to see through invisibility, etc!) This is probably most notable for spellcasters - SCS mages don't randomly get uninterruptible spells, improved casting time, free Alacrity, or the like. (In fact, I've removed these powers from creatures in the vanilla game occasionally, though I haven't done so systematically.) However, "fair" is sometimes in the eye of the beholder, so please do read the component descriptions if this is important to you!

ln particular, I'm claiming that *mages* don't get uninterruptible spells and improved casting time. As you say, it would be silly to say that nothing does.

Beyond that, all I can say is that SCS's readme is extremely detailed and lays out to a player exactly what they're getting from the mod, and that I stand by the accuracy of the overview comments. As I have been trying to stress throughout this discussion, there are a lot of judgement calls involved in interpreting the relevant design questions here - that's what I mean in the readme quote about 'fair' sometimes being in the eye of the beholder.

[Guest guest]

First I want to say thanks for engaging with me and taking my feedback seriously here. I'm going to drop the discussion of what SCS is/isn't since its not really productive and was meant as an aside in the first place and focus on the more meaty stuff.

Quote

Again, spellbooks are a bad indicator here. SURAKW1-SURAKW4, for instance (I suspect that's the example you have in mind?), only have spells memorized at levels 1-3 - but in the unmodded game they completely ignore their spellbook and force-cast spells at levels 6, 7 and 8. Setting them to level 13 is a bit undergenerous if anything - their displayed abilities make them at least L16. I don't specifically recall why I did that: probably a judgement call that it was too much of a difficulty spike.

Yeah. I'm looking at an SCS spellbook here where a few odd ones are with small spellbooks despite the 2da.  A force cast would explain that, and I didn't realize SCS used force-scripts as well, but without referring to the 2da or scripted abilities the obvious answer to me would be "a level 9 rakshasa should not have more spells than a level 11 Ruhk Rakshasa, who should have less spells than a level 17 Maharajah," but I don't know your personal reasoning.  In terms of the base game I'm not entirely sure the scripted spells they all use either since double-checking that would require me doing a fresh install, but Death Fog and Cloudkill were the two previously mentioned, both being evocation. I'm not sure if they use spells other than that, but I do remember them being blasty evoker types overall, and I still think the theming would be more exciting than the current interpretation, but that could just be my personal preference where I like the spice bloody, hectic fights bring. I appreciate the weird variety like bigby's fist on cowled enforcers, even if I haven't seen them use it yet.

TRRAK03 is the most obvious example of one of the missing ones in the 2da and I don't know if he has any scripted abilities, but open him up and you see class:mage with an appropriate level spellbook. I still not particularly happy with the current implementation of rakshasa overall as they're one of the less interesting parts of any run, but at least I'm learning how weird the game could be with scripting.

Re: Evokers/transmuters/etc - I guess regardless of the specifics of Rakshasa and if they had more scripted spells I'm unaware of, I still think there is a certain blanket distinction to me where "X knows level 5 spells from school X" is different from "X knows level 5 spells".  I don't think 100% fidelity to spellbooks is the right answer, but to me I think my ideal interpretation is one where their highest level spells and most of their spells in general are one school, that school would still be their first pick if they only have access to one or two spells of that level, and that their book keeps the same general theming. I realize others may not agree with me here.

Quote

My assumption has always been that since most NPC mages have had years or decades (for liches, centuries!) to fill their spellbooks, they're less constrained than the player, who only has weeks or months. Certainly that could be challenged, but at some level I don't think adding a restriction like that would make things more interesting. I suppose in a sense I simulate it by using quite a wide range of spells even when some are clearly less good than others (I don't think the Bigby's Hand spells are really competitive, for instance, but I use them because variety is good). There's less variety with defensive spells (not none, though) partly on technical grounds: the code for sensible handling of mage defenses is already hideously complicated to keep track of.

If we're talking about what counts as 'playing fair', though, I don't see any evidence that the original game respects a principle like this: practically every appropriately-levelled enemy wizard can cast Spell Sequencer, for instance.

In terms of playing fair, I think I already said that question was a bit of a stretch. but it brings up some more interesting questions. 

Regarding Sequencer, It's obvious BG2 houseruled it to be more widely known, but was meant was more of a thematic example of an in-universe rare spell.  People often refer to wizards as an S-tier class, and that's really just because we know spells in the game. If we looked at vanilla IWD where there was a much more random and sparse scroll selection, we got a different answer. I think thematically it makes complete sense for Irenicus to have every spell known to man, or Cowled Wizards to have a very large shared library, but other mages probably go through similar struggles to the ones we do. They have more time, but researching spells is still costly and expensive. Not really asking this to be changed (I do think *some* spells should maybe be less common, but that's a personal preference mostly), but I bring it up because it highlights the big difference between why mage enemies are so much powerful in SCS while warriors typically feel like lackeys, minions or pushovers.

Pulling a simple example out here, a 16th level Amnian Centurion has a set of full plate armor, a non-magical spear, and a nonmagical helmet. A single level 8 scroll in a 16th level cowled wizard's (speaking of, Cowled Wizards have no problem with AoE and will mow down their bodyguards with lightning bolts in order to hit my party at a good angle) spellbook has roughly the same value of his entire kit. Acknowledging most warriors don't even get that much, one of the existential flaws in terms of balance is that we can throw 50k gold at a wizard to make him an S tier opponent, but if we give the same amount of wealth to every warrior would quickly turn the game's economy degenerate quicker than it already does. I think BG1 does a much better job of creating scary martial enemies than BG2 does, and one of the factors is that they were willing to throw things like powerful magic arrows on Blacktalons, kobold commandos, and Hobgoblin Elites. Two of those three have separate arrows they drop from the ones they actually use on us. To a certain extent, Drow in BG2 do this as well with disintegrating loot.

In my eyes one of the existential and not easily fixable weaknesses of SCS is that it can sometimes create a lopsided experience in terms of challenge between martial and magical classes; a big reason is the 50,000g wizard versus the 600g warrior, and I think it would be fair to assume that warriors can usually acquire some sort of wealth with similar ease to their wizard friends. SCS plays around a bit with potions where a few will pop a potion of heroism or something, but 3 potions of extra healing mid-fight doesn't do as much as prebuffing with half a mages spellbook. I think if the we're designing encounters to be fully prepared and expecting to face *us* (mages rely on anti-player tactics rather than monster-hunting), a warrior with 50,000g can be a very deadly opponent, both in terms of prep and in terms of mid-fight. Arrows of dispelling and detonation would be scary for players and things like potions of invulnerability or magic shielding (or even magic blocking) could drastically change how a PC vs fighter encounter might go down. (CC/forget doesn't work anymore, haste/fortitude/regen/etc means both more deadly and harder to kill). To prevent loot treadmills and apply them fast enough the potions could be interpreted the same way mage prebuffs are (rather than current behavior where they will use an action to drink their first potion of XXX) and the arrows could made be unlootable like the hobgoblin ones (they broke when you killed them, perhaps). There's other ways to interpret things like oiled or poisoned weapons.

A very basic example might be all 5th level warriors get to prebuff with a potion of defense and a potion of hill giant strength, 10th get invulnerability and stone giant, 15th level get regeneration, heroism and fire giant as well,  20th get magic shielding and cloud giant added, etc. (i understand the issues with giant strength, but very rough idea here, and there's ways around it).

The level of abstraction one can make is depending on willingness and level of departure from the original game, and I don't know the right level of balance, but I think if we're following the line of logic that enemy wizards are very prepared, that it probably isn't the warrior's first day on the job either. Of course I realize that a lot of it is probably outside the scope of SCS in terms of both extra effort and mission creep, and I'm not entirely sure what level of gear would be balance-appropriate and fun, but I think it's a good mental exercise to remember that "Mages are S tier, Fighters are D tier" is less a truism of any version of D&D or even Baldur's Gate and is more because we can get lazy and the DM is willing to tell the fighter no but will be lenient enough to give the mage any spell/feat he wants as long as he can find it in the right supplement.

(tl;dr a minor hope of mine is to see the improved AI for humanoid martials slightly expanded upon, I think at least *some* of it falls within the design philosophy of SCS, and would not mind at all if you took the ideas if you ever do another passthrough on martial content.)

[DavidW]

Just as a quick comment (I don't right now have time to write a longer one): I think the really decisive reason mages are (often) so much stronger than martial types in SCS is that in unmodded BG2 mages just don't fight very well because their scripting can't handle good spell use, so Bioware made them very high level to compensate. SCS actually tries to dial that down in a few really extreme cases, but not systematically. You run into L20-L25 mages much earlier and more often than L20-L25 fighters.

The other issue, of course, is that if you kill a mage you can't steal his memorized spells, whereas if you kill a fighter you can steal his stuff. So increasing fighter loot creates an arms race if you're not careful. Potions in SCS already have this consequence to some extent. (Incidentally, I don't think potion prebuff would change much given the way the engine works: you can swig a potion without forfeiting a round of attacks.)

Thanks for your feedback in any case; I'll try to reply more fully when I have a chance.

[Guest guest]

Quote

I think the really decisive reason mages are (often) so much stronger than martial types in SCS is that in unmodded BG2 mages just don't fight very well because their scripting can't handle good spell use, so Bioware made them very high level to compensate. SCS actually tries to dial that down in a few really extreme cases, but not systematically. You run into L20-L25 mages much earlier and more often than L20-L25 fighters.

  I very much disagree that its solely a function of level. A single planetar could probably 1v1 any humanoid fighter in the game, and the buffed Conjure Animals that SCS has is significantly more threatening than an equivalent xp fighter as well. Yet we can solo the game with a fighter if we're clever, patient, and use our resources wisely. 

Quote

The other issue, of course, is that if you kill a mage you can't steal his memorized spells, whereas if you kill a fighter you can steal his stuff. So increasing fighter loot creates an arms race if you're not careful. Potions in SCS already have this consequence to some extent. (Incidentally, I don't think potion prebuff would change much given the way the engine works: you can swig a potion without forfeiting a round of attacks.)

I directly address the first point in my previous post. There's 2 things that prebuffs would do. 1st that it bypasses the loot treadmill since they'd be relying on a scripted action rather than an item we're giving the player.

The second is the same difference between insane prebuffing mages and the most basic version where they don't prebuff at all. You can easily shut down an unbuffed enemy in 1 round since they don't have any protections at all. Mages aren't challenging because they know PFMW or spellshield or any single particular spell. If they only had PFMW, all you need is to spam magic missile. If all they had was protection from magical energy, you could kill them with a single backstab on round one. They're challenging because they start the fight up with a dozen buffs at the same time. You could give them all the AI in the world and without prebuffing and contingencies mages move down several tiers.

Prebuffing through a full array of potions means a warrior could theoretically come out of the gate hasted with +30% hp, 30% lower thac0, Full Plate equivalent AC, 24 str, 21 dex, 18 con, 2hp/round regeneration, immunity to charm/fear/etc, a -20 bonus to all saving throws, and 50% to all magic and elemental damage, 50% magic resistance, or even complete immunity to most elemental damages and another -10AC bonus to crushing. (And depending on how liberal you want to be, you could stack potions or even add green scrolls or common charged items as prebuff effects). I'm not sure if that sounds completely overbearing or if it's underwhelming compared to the arrays of buffs a wizard can get by max level, but all of that is less expensive than the average wizard's spellbook, and we have to remember that item use and wealth by level are class abilities too. Arguably you could give mages about 75% of those potions, but that's just by using wealth that they could otherwise use on their spellbook. How far we stretch the idea goes up to personal interpretation and gameplay taste, but like I said, DMs too often let the mage have everything while telling the fighter "you can only autoattack."