What can a "smarted-up" CRPG look like?

It would certainly be possible. Depending on how you went about it, I agree it probably wouldn't be all that simple to do, though.

With the right setup, for a number of (most?) scenarios, you could simply alter some of the parameters to the AI routine for different levels of intelligence (for example, risk/reward weightings or search depth - this is how a number of chess games change [or at least, used to change - not sure if modern ones tend to do it the same way] the difficulty level). This would give you "smarter", or "dumber" choice making, but the process would be essentially the same - some of them just favor certain types of behaviour or look farther ahead than others.

Of course, this means you have to have an algorithm that will make noticeably different choices based on those variations. I would leave it as an open question as to if this would be sufficient for what you are looking for.

However, if what you are actually looking for is different methods of choosing rather than just making the individual choosing 'smarter', then you may have to have entirely different routines (or at least portions of the routines) for the various scenarios - with a leader, you are essentially changing from a set of individual choices into a centralized decision making with the individual choices made within that framework. You could probably fake this with some behavioural patterns, but I'd have to think about how you would do that. In any case, this means more additional algorithms or additional complexity to the existing algorithm.

In any case, you now have many more scenarios to validate - instead of just checking the AI algorithm under scenario X/Y/Z, you have those same sets of scenarios, but with, at minimum, all the different relevant INT levels (or at least the appropriate sampling of them) plus the diifferent situations (If you have different AI routines, you'd probably want to test their interaction).

Of course, this is off the top of my head, so there are probably a bunch of shortcuts you could take that aren't coming to mind for many of the problems here.

 

One of the "short cuts" Temaperacl is called Turn Based games. Essentially a separation of time scales.

 

In games like Chrono Trigger, Lost Oddysey, and many others that I have only heard about like Temple of Elemental Evil, enemies act differently depending on what other enemies are present (new combo attacks, buffing and healing each other, and sometimes limiting each other, so that you have to defeat enemies and do other stuff in the right order or the battle will be harder than it could have been).

As far as coming up with completely new strategies dependent on a number of complex factors... I agree with Captain Shrek that no gamedev has yet figured out good "vanilla" AI without making the computer a cheating bastard (ex. you can only move and attack in the cardinal directions but enemies can do that and use diagonals, too).

Also, misdressing some tanks in mage attire so that enemies will attack them instead of your disguised mages just sounds like tagging mage-specific equipment with an aggro mechanic. I wonder what this would do to game ballance or at least how rediculous a world would be if you can tell a strong warrior by some robes and an old staff.

 

Except that isn't a short cut - it's just a different environment. The exact same complications I listed apply to a TB game. TB allows more time to process the data and may change the available actions (your or opposed). It certainly does make some things easier, and will probably change the decision making algorithm you use, but fundamentally the same issues of validation / need to write multiple or more complicated routines still exist.

The only thing that I listed that I think TB would actually help with is the separation of high-level planning vs. low-level decision making since you dont' have to worry about decisions being made while the high-level planning is changing, but, to be honest, that doesn't seem like it would be the biggest issue there. I would think that properly formulating a method of describing the high-level plan (It could be something as simple as "Target/Force Required" points, but I don't know if that would be sufficient for what was being looked for) so that you could make decentralized decisions for that plan (Unless you put all the decision making centralized into the same place as the planning.)

 

I don't get it. You make some of the arguments that SUPPORT TB as a short cut yourself and your starting sentence is:

Alright.

I can give you some more.

RT essentially has a continuous and simultaneous action-reaction model. This means that, whatever is the smallest unit of time measurement there, the AI and the player has to respond without it instantaneously.

TB obviously gives the computer a completely markovian (ok, probably not COMPLETELY, but you get the idea) space of choices where it has only to solve the problem from the current state onwards.

Please do not get me wrong. This is NOT to say that TB is made more simple or anything because of such a construction. On the contrary, it allows the developer to determine a more flexible and challenging AI that can really try to defeat you with evolving strategies at each turn, instead of scripted AI required to provide challenge in a simultaneous actions battle, which you can typically break by taking unorthodox actions.

 

The main reason why we don't get decent non-cheating scripts in cRPGs is that developers is that they don't have enough familiarity with the scripting language and don't have enough time to make intricate scripts.

 

I've seen this discussed often over the years, but I have never read an article that explains it so I understand it.

The way I would try to create emergent behavior, is create a component script system for each creature, composed of many small scripts, each script describing a very simple behavior. As needs or wants arise in each creature, one or more behaviors would take precedence over all the other behaviors and take control of the creature. The only issue that would cause problems, are situations when conflicting behaviors get control of the creature at the same time.

 

The best example of emergent gameplay is Dwarf Fortress. Off the top of my head at one point the carp were eating dwarves. That led to some hilarity. The carp were set to be predatory and they had a size bigger than the dwarves so BOOMP they ate them up and everyone laughed.

Problem is it really doesn't work for anything but a game similar to that or populous. You can't really apply it to an RPG and expect anything but idiocy, you need strict rules of some kind to make a game that makes any sense.

Artifical Intelligence is about as serious business as programming gets. Any game that just uses a few lame scripts to make the AI will always have terrible AI. Even if they were willing to put in the time to make it happen you need to have as much performance as possible at your command as as many software engineering options as possible or it's always going to be bad.

 

Fuck trying to reinvent the wheel. Try to remake something like Star Trail first.

It's like Kwans talking about putting a man on mars and shit. How about rediscovering your ability to goto the moon first eh?

The best crpg is one with turn based combat, party based char gen, and tons and tons of stats and skills.

 

Like PoR:RoMD?

 

No, not like that at all! It did not have tons of stats and skills. Everything was fixed...you couldn't even choose feats! No player choice...shit game.

 

I have no disagreement with your points on the advantages of Turn Based (except one - below). I think our difference was just a matter of terminology and assumptions. I was also starting off from an (unstated) base of non-scripted AI and, while I didn't have it specifically in mind, I usually think of things from a TB perspective.

By "short cut", I intended to mean something like occasional hard-coded pruning of certain types of branches or using emergent behaviour to imitate the allocation of resources (say, all the 'leader' does is assign weights or modifiers to the targets that the individual AIs care about instead of some sort of more formal communication) or even something like specific scripts at some points.

The decision to go TB or RT is more of a fundamental design decision though - in that sense, I don't view it as a "short cut", although it may be choosing a more favorable environment for the AI creation. In any case, the complications I listed in my first post still apply to a TB environment as much as they do to a RT environment.

Given that actions usually take a certain amount of time to perform or before they can be performed again and that humans don't react immediately, AIs could afford to operate on slightly stale information in a RT game - they don't need to react immediately - even if you give 200ms between stimulus and response, that still gives you 100ms cycle time. A huge amount of processing can be done within 100ms, even in a non-dedicated environment. Of course, each AI running at the same time cuts your available time slice.

 

An important difference between TB and RT AIs is also that in TB, you do your move and then must live with it until all other units have made their moves too, at least if you don't move all your units in sequence but player moves in-between. Whereas in RT you can constantly adjust your doings based on what's happening around you, making it easier to keep formation for example.

 

You can't fix what doesn't exist. That at least, you correctly pointed out.

Obsidian coming up with a decent AI in their games any time soon ? After :

• IWD = scripted "Banzai! Sus à l'enemi! Kill them all, the Lord will recognise His own!" triggers, ordering NPCs to get abused in pathfinding-bugged corridors
• NWN2 = open the door, cast fireball, close the door. What do you mean, AI ?

:hahaohwow.gif:

If Obsidian is as serious about it as they are with their QA standards, hiring summer IT interns to work for a couple of hours on some scripts off-sourcing such a job to some third world devs that happen to possess some notions of the English vocabulary, that sure won't happen tomorrow. I say this with all due respect to the Obsidian's writer and art teams, which are what makes their games otherwise good (and they know it).

Having spent a fair amount of time in the past digging through the code of the glorious Bethesduh's Radiant AI, Bioware's NWN game (+expansions) scripts, Obsidian's NWN2 (+expansions) and FNV scripts, I can safely say that AI has never been a serious concern to any of these big AAA players (and by extension, I assume any of their ilk). Said scripts resemble nothing more than a messy patchwork of more or less talented coders that consistently conflicted with each other over time, overwriting what someone else had done, consciously or by ignorance, removing features, introducing new bugs, etc.

I wouldn't trust any of these companies to give more than a flying fuck if someone brought up "Let's make a great AI for our games" as the topic of the day at one of their Dev meetings, now or in the future. They aren't even trying anymore.

Yet, there are games that were developped by relatively small teams that had decent AIs. Games, some Japanese games for one (and I am not talking about FF8 here), and quite some Western ones as well, that focus on tactics and strategy more so than on flashy 5 seconds-long "combo-breaker" animations, or spraying gore and zombie brain mip-mapped textures all over stylish Halo <insert latest iteration here> armors.

If change happens, gaming companies will have to look back to the past, and analyze what the games above (and teams) did right. Before you ask, I am not going to unilaterally establish the definitive list of such games, because such an effort will stall even before it starts (it's the Kodex after all), and it should be pretty obvious to everyone that a game that has more AI (custom and competently made) user mods than I have fingers on my both hands is probably not among them.

Gaming companies will need to convince themselves (or be convinced, but can they be convinced by Facebook sheeples that drench their wide-open wallets with drool in front of horse-armor DLCs ?) to accept hiring capable professionals for the job, because developping AIs (and incidently Quality Assurance) is NOT as trivial as an after-thought that one squeezes into a week or two, completed just before release date.

I am definitely not understimating the complexity of such a noble endeavor (and I will probably post more about it later), but going back to the message that motivated my question, what exactly is not "that simple" about what I underlined to a competent developper ?

 

Well with modern graphics there is this issue: In the olden turnbased games there were few animations, so combat and turns proceeded briskly. But with modern graphics you did have to wait for your turn a bit.

BUT, handled intelligently and with fun encounters this should not matter much, if at all. TOEE was pulse pounding, I never grew bored in combat. (the game outside of combat was another matter though...)

 

Often I wish there was just a button to speed up all simulation by 2x... If your computer can't handle it, just don't use it.

But I guess many devs won't want to include such a thing since it'd be admitting their game has boring skippable parts.

 

Fire Emblem. Front Mission. Jagged Alliance 2 (modded). Tactics Ogre. Final Fantasy Tactics. Der Langrisser. Wizardry 8. (EDIT: ToEE was a disappointment for me, but YMMV of course). Games with the kind of AI that you describe already exist, you're right . Some of them were developed by very small teams/companies. Proficient player-developped AIs exist for several other games. Not high-class AIs like in some chessgames, but definitely good enough to be worked with for a starter.

I present you with a typical RPG battle (EDIT: spoilerized) to be waged between :

• a group of adventurers (player controlled)
• a group of enemies (NPCs)

Show Spoiler Hide Spoiler

If enemies have a decent leader, at least of human intelligence level :

• A-0/ they can just rush in (I guess)
• A-1/ they can get a rousing speech to raise to their morale (bonus)
• A-2/ they can put more elaborate tactics into use
• A-3/ they can call in for reinforcements
• A-4/ they can send out scouts to prepare ambushes
• A-5/ they can attempt to parley, and betray the group of adventurers
• A-6/ the leader can step forwards, to settle the combat over a duel
• A-7/ the leader can attempt to rally routing troops
• A-8/ they may hide, to collectively evaluate the chances of their survival before attacking
• A-9/ they may decide not to fight to death
• ...

If enemies are without leader, but at least of human intelligence level :

• B1 = A-8/ they may hide, to evaluate the chances of their own survival before attacking
• B2 = A-9/ they may individually flee mid-battle if the tides aren't in their favor
• ...

If enemies are with leader, and are of animal intelligence level :

• C1 = A-6/ the leader of the pack can step forward, to challenge the adventurers
• C2 = A-9/ if the leader is defeated, the rest of the pack flees
• ...

If enemies have no decent leader, but have War inscribed in their genes :

• D1 = A-0/ they can rush into the group of adventurers, axes raised, orc-WAAAAAGH style
• ...

If enemies are usually solitary predators, of animal intelligence level :

• E1 = A-4/ they will lay low in a bush/tree/web, waiting to pounce on the adventurers
• ...

I could go on for pages.

What exactly makes coding such a differenciation "not simple" ? Once you have class A, the hardest part, granted, class B and beyond should come with no trouble.

Computers should be cheating, and they will continue to cheat, till everyone gets computers at their home with the horsepower needed to predict player moves simply from their playing style and past actions in the game.

I agree with you that they shouldn't cheat as in allowing hostile NPCs to constantly (ab)use excessive in-game powers (using "diagonals") or features (hp bloat, buffs) that the characters under the control of the player can't. That is not a challenging AI, that is a frustrating AI, that still plagues alot of games as of today.

They should cheat by detaining the extensive knowledge of the game beforehand, and making the best use of it, accordingly to their processing power and what the game designer/developer will have "taught" them. They should take advantage of every feature that the game has to offer, and (especially) combinations thereof. Such combinations should be complex, at least not immediately obvious to the player. In that way, the player would need to think while they play, learn, analyze the behavior patterns of the AI, and come up with equally complex alternatives to counter them.

Provided a fair starting ground (a functional UI, documentation, tutorial, whatever allows a reasonable learning curve), if the player isn't able or willing to learn/improve during the game, they should of course be punished by the victory of the AI. In the other case, the player wins.

cRPGs have a lot to learn from wargames, namely prioritizing threats and victory objectives.

It makes sense for an AI to attack whatever presents itself as the immediate source of highest threat, because it makes sense for a human intelligence as well. If two people attack, one with a visible gun and another with a visible bat, you will probably aim for the one with the gun first. Unless you had a wall to hide behind nearby, in which case you would probably first go for the guy that rushes forward with the bat. Unless neither had visible weapons, at which point you'll still prioritize otherwise.

Coding such an AI, that is able to prioritize threats/objectives in a rational fashion, according to objective factors (like visible weapons), is very possible nowadays, at least in a relatively satisfactory manner in wargames.

One could argue that cRPGs aren't conceptually as simple as wargames. Some D&D characters can potentially have hundreds of different races, feats, skills, stats, items, etc. Coding AIs for such NPCs is a *nightmare*, and probably not within the boundaries of a project like a video-game that is supposed to be developped in 2 years from scratch.

My point however is that eventually, an archer remains an archer, someone that fires deadly arrows from an ideally elevated spot. They are weak in melee (since they can no longer use the weapon they're the most trained with), which is why they need to either remain at distance, or be protected from foes. I have yet to see an AI in any commercial cRPG that attempts to protect its archers. The same kind of bare-bones analysis could prevail for mages, rogues, priests, warriors, etc.

Finally, no one expects, or asks, for human-like AIs in video-games, but a (simple) wargame-like AI, working on the premises of a slightly more evolved rock/paper/scissor scheme, that accounts for various contextual factors (presence of a leader or not, terrain, intelligence, factions, weather, personal reputations, etc.) is largely within the league of modern cRPGs. If we need to get rid of all the pompous D&D regalia to get this kind of AIs in cRPGs, then I am all for it. Let it happen, and start the whole genre from scratch again. Our games will be so much better, and this time, really challenging, and damn f*cking *enjoyable* for being so challenging.

 

Features don't hang in vacuum, they are not added for their own sake.

What would perma-death add? Do you mean it as iron-man system, or simply no-resurrection? Would your game be "easy" enough, to avoid randomly punishing players with having to restart (or reload if it's the latter)? If so would it also be hard enough to make anything beyond bare survival something you have to work for and not necessarily succeed*?
Would softer means, like potentially unlimited, but not game breaking reload punishment (autofail on non-essential quests due to unrelated circumstances, legendary items going missing due to competing adventurers who then sail overseas and become unavailable, etc?) be implementable and adequate alternative? Can you delay many desirable and undesirable consequences to softly force the player to commit to their choices and mistakes?

What purpose would forcing player to draw his own maps have, other than tedium, annoyance and murder of many trees? Does your game contain any features, devilish spatial tricks or other things that would break or be broken by functional automap? If not, then what's the fucking point of forcing player to do something a dumb algorithm can accomplish? Will you also implement ass wiping minigame as an element of your ultra hardcore survival system?

Is this stuff introduced in meaningful, convincing manner and is it in itself meaningful and convincing element of your world?
Is your bar for intellectual challenge set so low that removing automatic chronicling of mere facts, obvious shit and stuff outright spelled out to you remove challenge those plot- and setting-related elements provide? In other words, is your game aimed at retards?

Actually I don't really see anything wrong in either system. Static character system may have obvious advantages when the timescale is relatively small, as it does away with some balancing problems that often lead to some seriously retarded solutions.

Actually you had most of this stuff in Morrowind in some form, despite the game being light on lasting C&C.

Challenge is when the search space is too big to simply bruteforce right input chains and rule space is too big to simply blunder into good enough algorithm for generating them in responding to feedback.

M:

You can take your "Alas, Pervasive Moronism" shit and stick it up your ass.

You know what?

I can actually agree.
TB is a good way out if you already have a really good AI that has computational complexity as its only drawback.

However there are other, imperfect, but good enough ways out.
For example you can process immediate response AI in real time, but let complex tactics processing be deferred. It's not like IRL you have flawless instantaneous communications, or individual combatants always performing like well-oiled cogs in a perfect machine, you don't need to process that every tick. Your player probably won't be altering their tactics more often than once in a few full seconds, so your AI doesn't need to be much faster.


*)

Fucking this.

 

DraQ : I love your insights. Unfortunately, I am unequipped to competently answer them. Answering personally, I haven't really played any complex games myself, so these example questions are concepts of what veteran players want from what I have observed on the Codex, crpgaddict , and elsewhere. Recently on CA's blog I read that he was really pissed when Ultima 7 switched to dialogue options instead of text parser because he then no longer had to pay attention to what anyone was saying. He played the whole game through with the obvious correct answers appearing right under the NPCs' speech. In that interview I linked to earlier, CA talked about how wierd it would be when (in a few years) he no longer has to take notes or pay attention when making maps because a quest log and automap/navmarkers take care of that for him. Other people who love games like Magic Candle (for which CA wrote 'The Book of Flame and Wax' for all his notes he made in order to solve the game) and Chaos Strikes Back (mapping and probably at least some note-taking) probably think features like these were a big loss of required intellectual effort, too.

As far as permadeath (yes, I mean iron-man, though no-res does fit a story a bit better in some ways), even in a static story game, we had a thread maybe a month or two ago written by a guy (who became very prosperous with BROFIST!s) who thought Fallout should be played iron-man without exception (he may have thought the same for all games, but I can't be bothered to see if he did make that claim or not). Apparently, this guy could beat the game this way regularly, utilizing tactics to effectively avoid FO's nasty crit-hit kills that seem to come at random. I don't subscribe to full-on permadeath, but I loved games that have lives and hidden 1-up items. I figured that if there were a CRPG whose load game menu included profiles that kept track of lives independently of your save files, then we could have a game that both allowed one to save-scum if they so chose, but once a life is lost (tracked in the profile element of the interface) only hex-editing/moding could get it back (note that in this scenario lives are not a save/re-load currency like money or experience is in some games). I was also thinking that the entire game wouldn't have to be lost, maybe you could just start at the last programmer-designated checkpoint (in the Zelda series you restart at the entrance you came through of whatever area you saved in) or choose to reload your last save with the beginning amount of lives and all potential hidden 1-up areas reset (this last part would apply in both situations). I am sorry if all this sounds like non-sense to you, but my experience has shown this to be a very good tried and true enforced-challenge model that I believe can be made to fit a CRPG.

I wasn't contemplating a fixed character system, but suggesting that you be able to level up whenever you want. Basically, I was stipulating about how to KILL! grinding (I should add that money shouldn't be grindable in this scenario. Maybe, you can only use items you find in chests, standard consumables, services you provide, and items from quests can be traded in shops for other item(s) or store credit, so you have to shop around to see who will give you the best prices) and put focus on making good leveling choices (such as when you have more to level after playing a while obtaining better equipment, abilities, and skills than what you can choose in chargen).

Again, thanks for fleshing out the discussion. What are your thoughts on how a "smarted-up" (vs. dumbed-down) modern CRPG can be made?

 

Karmapowered

I am serious about it. Make a thread on cRPG AI. We will have a field day there. I have a lot to talk about it and no one talk with on this issue. Just give me a couple of days. I am really REALLY busy right now .

 

Captain Shrek

Will do, and no hurry. It's NY's eve soon'ish for everyone.

I have refrained from walloftexting monopolizing this thread, because I realize that it isn't just about (better) AIs in cRPGs, and because gathering/reading input from other people (there is some good stuff in here) is much more rewarding after a while than a mere monologue.

I just wish that people would stop the general non-sense about challenge/smarts/incline/decline in cRPGs. When I read two posts back that people should go back to tag-based driven dialogs (how could anyone have really enjoyed that in DF/MW ?), draw their own maps on paper, and deal with U7-style inventories, in order to play smart(er) games, it makes my inner Baby Jesus weep

 

There won't be good AI in games until game devs start taking their shit seriously and start programing AI libraries that are higher level and coded against standardized rules in library form.
You see chess engines do this just fine, including plug-n-play and competitions (with a protocol; aka: a ruleset and a procedure for interactions, with precondition and post conditions).

You don't see every C++ or Java coder trying to replace boost or collections, why does every dev have to recreate the conditions and effects of; say, cleave, or the bonus against attacks from flanking, or even to integrate complicated pathfinding with complicated AI (blocking, funnelling)?
AI should either think in totally deterministic generic-micro fashion (if you want to annihilate the player, ie; with planing out the actions to leave the least amount of reaction time to the player), or very high level indeed, higher than what is currently possible because no one wants to take working pathfinding code and special case it to add obstacles or create ambushes.
Both strategies require DATA, and guess what, 98% of dev time is wasted with DATA processing and mismatched tools because different data interpretations.

BTW; we may be seeing soon a return of the parser. Voice recognition, courtesy of google and their massive datasets, is vastly improved on the last few years - just try it on the Android if you have it; ask '133+400' or '7 kilometers in meters'. Yes, you'd have to license their ultra-dupa neural network, and it probably wouldn't run too good, but should be plenty fast enough to run on a desktop on a game 'dialog mode', especially as cores increase, since their algorithms are all massively parallel.
Wouldn't it be cool to say 'attack that guy with the mace' and have your AI 'companion' attack the nearest visible male with the mace? Or the conversation key word system of the early ultimas with a microphone and no visible user text at all?

 

There are two main problems with perma-death:

First, not only can you not afford to have a game with shitty design, the design has to be virtually perfect. One poorly designed encounter where you get stuck in a situation where you can't get out of alive through no fault of your own can ruin the game. How many RPG developers do you know of that are up to the task?

Second is that it imposes design limitations. Say you have a proper monocle massive open world RPG with a ton of character options. There's no level scaling because that would be popamole, so you don't always know if the next area you walk into will be way over your head. Say you have to pick between three quests:

a) A man throws a party in his house. Suddenly he hears loud screaming and growls from the cellar, so him and the other party goes barricade the door with people in the basement banging on it and screaming as something tears them apart. He runs out looking for someone to investigate and if necessary kill whatever monster lurks there.
b) A nearby forest is rumored to be haunted. This hasn't usually had much of an effect but recently children have begun to wander into it and haven't come back. The mayor wants someone to go into the forest, find out what is going on, and hopefully bring back the children.
c) Every night the adult inhabitants of a small village disappear from their beds and reappear in the morning with dirty clothes, cuts and bruises on their feet and arms, and feeling exhausted. They have hired several adventurers in the past, but every time the adventurers would disappear with the villagers but unlike the villagers never returned in the morning.

Some characters could gather information using their skills:

IE:

a) A ranger could use his tracking skills to determine if some manner of beast wandered into the cellar from the outside. If not he could attempt to retrieve one of the corpses by the door and study the wounds (in daylight, to avoid potential threats from vampires or werewolves). If he determines the slaughter was caused by an ancient vampire he has no hope of beating in his current condition he can run over to a nearby temple and inform them of the creature, and at the very least seal the door so it can't get out until the ranger is stronger and can deal with it.
b) A proper necromancer with useful non-combat abilities like detecting undead and talking to corpses could get a sense of what kind of presence the forest holds. He slowly wanders in able to feel any incoming undead long before they can sense him. He finds a corpse who informs him that he was a monk in a monastery when a powerful ghostly wizard appeared out of nowhere and slaughtered his brethren. He ran away and almost made it but the wizard sent a quick undead scout after him who cut him down from behind and left his body to rot. The necromancer wanders a little further in only to detect a massive zombie army being built up, with several extremely powerful zombie samurai and spellcasters backing them up in addition to the ghost wizard. The ghost wizard senses him and sends some undead after him, but the necromancer is far away enough that he runs like a bitch and avoids being skewered.
c) You're a wizard who is well versed in space-time and mind magic. You take the job, and inform the villagers that you will return in a few days to deal with their problem, but secretly use your magic to turn invisible and erase your presence to magical detection cannot find you. You also have a mark spell set up back at your wizard tower in case things get bad. You sit down in the village and wait. At midnight a powerful green light begins to emanate from a nearby cave that had been concealed during the day with a powerful illusion. All the adult villagers rise from their beds and walk out forming a line to travel towards the cave as if hypnotized. You try to wake up the children but they all stay stubbornly asleep. So you follow the adults to the cave. Inside you find another powerful wizard who was using the villagers as laborers to dig him a path to what he believes to be the ruins of an ancient city where secrets and artifacts could be found. The enemy is a war wizard well versed in divination. He senses you, strips away your invisibility and is about to conjure a massive wave of lava and melt you away but you have the sense to teleport away in time. You now have a good idea of what you have to face and how strong you need to be to beat him.

The scenarios are manageable for the above characters, even if they can't finish them at the time, since their abilities allow them to play fairly safely. But what if your character can't deal with those particular situations:

a) You're a war wizard versed in... pretty much nothing else because you like pew pewing enemies. You have no clue what's there and don't care. You walk into the cellar in the middle of the night like a boss ready to fireball whatever's there in the face. Whatever's there is an ancient vampire, and not the pussy 40 days of night shit. The Anne Rice nearly indestructible hyperspeed type who gets behind you in the blink of an eye and rips out your heart before you even realize what's there. You swear, press F9 but nothing happens because the game has perma-death.
b) You're a dumbass warrior. You don't care much for children's stories so you walk into the forest looking for heads to chop off. You run into a few straggling zombies and dispatch them without any effort. Then suddenly the mist thickens so you can't see two feet in front of you. The sound of moaning comes from all directions and suddenly you're surrounded by a dozen zombies. You kill a few of them but they keep coming. You try to run but you have no clue where to go, the minimap is useless in the fog, and since this is an intellectually stimulating game there's no quest compass to help you, which is just as well because your character can't even use a normal compass. The zombies are getting thicker and thicker and your swinging speed is slowing down. Then it gets worse: the undead samurai show up. You can't see them until they're upon you but they can see you. One comes from the front and blocks your tired sword arm easily while the second comes from behind and takes out your hamstrings. You drop to your knees as a third one chops off your sword arm. All 4 of them impale you then leave your corpse to the zombies to devour. You think fuck it, I'll just savescum my way out of this, but you press Esc and the Load Game button is blacked out.
c) You're a thief. You hear the adults disappear during the night and think hey, easy looting. But when the green light comes it turns out your feeble mind isn't exactly built to resist powerful domination spells and unlike the previous wizard, you didn't hide your presence so the spell had no trouble locking onto you as well. You don't die, but you have to spend the next 6 hours watching your character dig. Then you die unable to fight back as the wizard figures you're not a local and too much of a liability.

The only way for characters not built to detect/evade the above problems is for you to avoid any potentially interesting quest and just go for the obvious, dumb grindy ones until you're high enough level that you can take on everything. Then instead of having challenging but doable battles you go between safe grinding, missions that are now too easy because you're too badass, and only the odd high level quest/dungeon that poses a challenge. Then there's alternative: change quests so every class can do them (or at least run away):

No running zombies to cut down slow characters. No quick moving zombie samurai. They all walk slowly because if the player rolled a paraplegic commoner dwarf who can't afford rocket boosters on his wheelchair he still has to be able to run away from them. No mist or ambushes to build up tension/atmosphere. No supersonic vampires. Instead of lava walls the wizard casts slow moving energy bolts you can just dodge by stepping out of the way. His mind control spell mysteriously avoids working on you because you're the chosen one. Or you can level scale everything.

No but I would like cartography to be a skill. Maybe have no maps until your character learns it too, though it doesn't have to go that far. Ranger/explorer type characters could invest in additional mapping options, tracking, etc. Abstract it if you have to (IE: you now have 50 tracking, you can tell how strong the monsters in a dungeon are by hovering the mouse cursor over its icon on your map).

Journal should record the quest name and requested task, all conversation as well as all information obtained by skill checks (IE: in quest a above the ranger should record what he gets from studying the corpse). The journal should also have multiple sorting options, a favorite option and the ability to write your own notes and attach them to any quest data you want, as well as add marks and notes on your map.

Completing quests should be their own reward primarily. Material rewards should be at the designer's discretion. Some quests should fail and the world should change accordingly. IE: if you don't deal with quest b) above the ghost wizard eventually overtakes several towns (which then become uninhabited ghost towns) and a bigger central government is forced to send its own military units to go deal with him. Depending on circumstances your character's reputation could be tarred by your failure to deal with it and be scorned/derided/mocked (IE: drunkards start humming the Ghost Busters theme when you walk into a tavern).

 

I remember enjoying, the U7 inventory, not "dealing with it".

 

Why do so many, Codexers, abuse commas?

(brofist though)