The mp3 of the podcast can be downloaded here, played in the embedded player below, or you can follow us on iTunes.
- Procedural content
- Lovely lovely permadeath
- Real-time vs Turn-based
- Loot, glorious loot
Games mentioned include:
- Dwarf Fortress
- Diablo
- The Binding of Isaac
- Desktop Dungeons
- Spelunky
- Legerdemain
- Triangle Wizard
- EarlSpork
- Auro
- The Archive - the space roguelike Andrew mentions around 1:04:40 in development by Surprised Man (not Suicidal Man!)
Join us next week for discussion of the SNES classic Shiren the Wanderer.
I have been enjoying this so far, just want to chime in on permadeath. While it is a key to Roguelikes its only half of what is needed. The other part is the randomly generated dungeon.
ReplyDeletePermadeath is boring in a static world as once you have done something, you've done it. Even roguelikes such as ADoM where the overland map is static has random dungeon levels. Without the random content permadeath is artificial difficulty and a cheap way to lengthen a game. Think of having a Desktop Dungeon's map that is always the same no matter how many times you play it. Eventually you would win but you replaced what was a fun game with a knowledge grind. Yes permadeath is basically the Thing with roguelikes but without random content it is a horrible mechanic.
Anyway once again great podcast, I have listened to all of them so far and really enjoyed them.
A few points:
ReplyDeleteI think that you are undervaluing story in games. It's difficult to have an engaging story in a game with permadeath, that's why the story aspects of RLs has lagged behind the rest of gaming, hovering in the NES-era "Find the princess" and "Banish the evil" story-lines. Folks like Joe Hewitt have tried to push this into the SNES-era with games like GearHead, but to do so, he has had to weaken Permadeath to Permafail. This is going to be a weakness in RLs until someone can find a way to procedurally generate a story as complex as Final Fantasy 4 (tossing a stake in the ground) involving heroic sacrifice, betrayal, redemption, and story twists.
Making an enjoyable real-time roguelike was Blizzard's big success in Diablo. So, most of what RL fans call "real-time roguelikes" would probably be better called Diablo-likes, the exceptions would be games like Toejam and Earl. It's the same way that Rogue named the genre for bringing D&D to the computer and hence we don't call Nethack a computer D&D-like.
Finally, Shiren have good UI? Really? Ever try to move or attack diagonally on the SNES version?
I don't think we'll ever have procedural stories that are any good. The best we can hope for is something like ToME4 where the concentration is on setting and environment than on a specific story.
ReplyDeleteAs I said, Gearhead makes a decent effort to procedurally generate a story that makes sense and is interesting. And it works ... at least half the time ... if you don't push it too hard. The trick is that you don't have to provide all of the story, just enough for the player to sink their teeth in and fill in the rest. Actually, getting feedback from the player as to what they think is going on probably isn't a bad way to bolster a procedurally generated story.
ReplyDeleteThe only 'roguelike' game that we missed that's up and coming is, of course, Monaco...
ReplyDeleteI think everyone is missing an important part of what OD&D and games based around its concepts have for stories. Originally there was no storyline nor thought out campaign. You played the game, stuff happened, and afterwards you fit it all together to make a story. There are whole stories based off of fighting things like dragons and in roguelikes and games in general you fight many of those things without even thinking about it. Those are the stories, you are all just missing them because people have been trained by current generation games to only accept stories that are spoon fed to them. You know that one time you had that tough fight and the monster almost won but then you remembered that one item and was able to win and then you went and told others about it? That is your procedural story right there. Its how real stories work, you do something impossible and then tell people about it. Have you ever read any of the good Dwarf Fortress stories? Boatmurdered is an amazing story that happened completely by random chance and yet is better then most current generation preplanned stories. What we need is not built in stories but a game that is deep like DF only as a roguelike instead of DF. Of course the problem is that saying deep like DF is like saying you need to make a swimming pool deep like the Mariana trench.
ReplyDeleteYou must have had a pretty lazy DM, even 1st ed was designed to be played to a module or campaign ;)
ReplyDeleteThe problem with the player procedurally their own story for existing RLs is that the emergent gameplay required to make such stories interesting is extremely rare. Dwarf Fortress is a game designed to be rich in emergent gameplay and still for every thread like Boatmurdered there are dozens of boring, ignored ones. Even the lauded Boatmurdered is interesting around half the time.
Burzmali: The old school D&D revival is pretty much all about sandbox play - minimal to no preparation before running the game and letting the story emerge from play.
ReplyDeleteI think Boatmurdered is an example of what's wrong with trying to procedurally generate a story. There is no real plot, no character motivations, no logic, and the only reason people like it is because it's weird and funny. We're a long long way from having real random stories with characters that people can relate to. Of course as many would point out, that's possibly irrelevant, since games don't need stories at all.
ReplyDeleteAndrew: I hadn't heard about that, any links? I can't imagine that works too well for tabletop play, a DM/GM running a game by the seat of his pants comes off like bad improv 9 times out of 10, I've sat through enough sessions like that in college.
ReplyDeleteDarren: I guess that brings us back to the eternal Gamist/Narrativist/Simulationist debate. Though, I'd hate to think that the Temple of the Roguelike has a "No Narrativists Allowed" sign hanging over the entrance. I keep coming back to Gearhead, because Joe has done his best to incorporate the work of Propp as a method of fabricating a story. I think his work shows that such a game is possible.
I agree about Boatmurdered, it is more an example of good (at times) storytelling than it is a procedurally generated story.
I guess GearHead is a game we'll have to cover some time :) I'm certainly not against story in roguelikes, having made a story-driven game (Broken Bottle) and written about 30k words of lore for ToME4. I've always been tempted to try procedural narrative before, but can't help but think it'll fall at the most important point - interesting and believable characters.
ReplyDeleteDarren: "interesting and believable characters"
ReplyDeleteIt'd be kind of disappointing if RL developers can't even match Planetfall's Floyd, considering it only took Interactive fiction 8 years to reach that point. ;)
@Burzmali
ReplyDeleteThe focus of the old school D&D revival has been sandbox play for a while. You do need a decent DM to make it work but some of the seat of your pants type of stuff is negated because instead of making a story they focus on making the world. If the world is thought out then you don't need much because then when the players do something stuff will happen naturally. Also there is a big interest with Megadungeons where you have a giant (no, what you just thought of was small, I mean BIG) dungeon that can be the main area of adventure for the party to adventure in. There are some good blogs about the revival of old school D&D, a couple of them are:
http://grognardia.blogspot.com/ which is one of the biggest in the OSR
and
http://batintheattic.blogspot.com/ which not only has a number of good post but there is actually a specific set of posts on making your own fantasy sandbox (http://batintheattic.blogspot.com/2009/08/how-to-make-fantasy-sandbox.html)
Not to throw gas on the fire. But procedural generation is about game play mechanics. Story and what not are a different matter all together. Just my take.
ReplyDeleteBoatmurdered was awesome. But this is an example of the player placing a story on top of great game play. You can find similar stories of fearless captains defending the earth against hordes of alien invaders (space invaders) and the life of a nanite fighting to clear the arteries of a nearly dead heart attack victim while being assailed by the immune system (pacman).
ReplyDeleteNote that we are also talking about Dwarf Fortress here. The most complex game ever made. Or at least nearly so.
That said I would LOVE for a game to have a procedurally generated story. Absolutely love it.
>That said I would LOVE for a game to have a procedurally generated story. Absolutely love it.
ReplyDeleteSo, there are three options here. Either you have no idea about how difficult it is to write a good story, or you are talking about something far, FAR into the distant future. Probably after we're all dead. The last option is if you just don't care if the stories are crap, in which case, hell I can code you a crap story generator in 15 minutes.
Take some of the best authors in the world. They will spend YEARS writing a book, and after all that, it could still suck. They could spend months on a single scene.
So what you're advocating is a computer that's not only smart enough to write a story that's as good as a good author can do it (we can't do this yet), but do it ON THE FLY, during gameplay. This intelligence would not only have to be as smart as the great human authors, it would have to be WAY smarter in order to do that.
And by that stage we won't be playing with computers, computers will be playing with us :-/
ReplyDeleteKeith you are creating a false choice, perfection vs. crap. While an author might spend years writing a novel, most are written in under 6 months, they aren't all art but they're functional. That's all I'm looking for, a functional procedurally generated story. The benefit of interactive mediums is that the observer is part of the story, provide decent hooks and people will be write Boatmurdered for your game.
ReplyDeleteThis comment has been removed by a blog administrator.
ReplyDeleteProcedurally generating a story is easy. Most stories can be broken down into a series of small events, with rising tension. If you programmed enough events into a game that stayed within the context of dungeon crawling, picked a handful randomly and then swapped a couple so that the grander scaled ones were generally last, there's your story. The hardest part would be linking the events together, but you could probably get away with using just a generic meta-plot (find the 7 Golden Foos to defeat a great evil) if the smaller quest sequences were interesting. *Telling* that story narratively with any believeable characters or emotional attachment whatsoever would be near-impossible, but even an obvious bare-bones 'random events plot' would still make for very replayable and interesting gameplay.
ReplyDeleteFor example, say you rescue an NPC from a trap in a dungeon. The NPC asks you to help find their friend who was also lost down there, and you agree. In a normal game this would be a simple escort quest or temporary useful ally (depending on the NPC strength), but if you know the game has multiple resolutions - you find the friend safely / you find the friend captive by a powerful boss / you don't ever find the friend and the NPC sadly gives up / you find the friend but it was a trick and both suddenly attack you - then you're going to tread much more warily, especially if the game programmer has been clever enough to include hints as to which plotline you're on. (If it's a trap, maybe the NPC AI only bothers to protect itself instead of attacking monsters who are damaging you.) After the resolution either the NPC would join you permanently for a greater quest (good ending), they'd tell you a rumour and leave (meh ending), you'd a find map on their corpse (killed by monsters) or you'd find a letter from the their evil boss (bad ending), any of which would direct you towards the same next event sequence.
Programming these different events would obviously take QUITE some time, and different dungeons and challenges (sieged city, ambushes etc.) would need very good scaling algorithms to stay balanced whenever you encountered them. It would all feel a bit artificial once the player starting recognising the events and patterns ("Out of place corpses again? Sigh, better head to their natural locale for a necromancer plotline.") but by incorporating the story with game mechanics ("Noooo, my meatshield buddy betrayed me. ;_;" or "Gah! The orcish invasion of elf village X has cut off my teleport scroll supply! I'll get those darn orcs. \_/") you could still have real emotional responses. By keeping track of player actions you could even generate the next chunk of story specifically to have an impact: Fled from a tough unique in the first dungeon? Guess who gets picked and level-scaled up to act as the head baddy for the next invasion plot.
That's the kind of roguelike I'd love to play.
Lots of thoughts from this episode:
ReplyDeleteWhat makes permadeath a significant design choice rather than 'simply being able to lose the game' is that in roguelikes you usually spend much longer and put much more customisation into building up your character than, for example, your ship in an arcade shooter where you might grab 1 of 3 power-up types and be done in 20 minutes. This means starting over can be MUCH more frustrating (or winning much more rewarding), so it's not to everyone's tastes if they want to play a game more for relaxation than challenge. Funnily enough, this line of thinking makes coffeebreak roguelikes feel close to strategy/arcade games because of the reduced attachment (and thus risk/reward).
I hardly think roguelikes should be mixed with ANY genre. I can't even imagine how you'd add roguelike elements to a puzzle game like Zork, or a dating sim, or a 2d fighter ...
Mario is not game about decisions. It's about reactions and timing. Random levels are less important for real-time skill-based games because even if you are replaying the game and making the exact same moves it still tests your ability to time your button presses and whatnot. If there's a pit between you and the goal then you know on your first playthrough exactly what you do on your fifth; you need to run up to it and hit jump right before falling in. There's only so many variants so shuffling the order of tricky jumps with random levels doesn't really add anything.
^ Spelunky is different, of course, because you do have some limited resources to decide when and where to use. And there are other minor cases in eg. FPSs where a small part of the challenge is tactical as you need to stay alert for unknown enemies and clear areas safely. There, random enemy placement suffices.
Nethack is, if anything, an example of how NOT to do random loot. See: your inevitably identical ascension kit, random game-breaking wands of wishing, and gnomes picking up wands of death.
It's worth noting that the difficulty of recognising new enemies isn't entirely exclusive to ASCII. A player will recognise "that's a wolf" or "that's a dangerous coloured wolf" with graphics, but they're still not going to know how strong that wolf actually is in *this* game before fighting it. Roguelikes actually have an advantage here when they bother to include a useful description system (This wolf is [strong] for your current level. It moves [fast]. It can attack for [moderate] damage through your armour, or [howl for attention]. Etc.) straight up instead of monster memory or fancy literature quotes.
Another reason ASCII can be better than graphics particularly with tile-based games is that it allows the player's imagination to fill in the gaps. If you show the player a picture of an orc then they're mind is always just going to see that simple picture, but if you present the information with an abstract an indicator as an 'o' then running into a room full of them will cause the player to imagine their own scary and detailed scene rather than just seeing a square block of copy/pasted orcs. It's the same reason that a good book can be just as if not more engrossing than a flashy movie. There are ways graphics can achieve a similar impact, but they're time-consuming and difficult given the roguelike interface.
I've played one Pokemon Mystery Dungeon game and thought it was fun as a casual roguelike. There wasn't permadeath but there was still a harsh item penalty for dying and you couldn't just reload, so it was a milder version of the same overall feeling. I disliked a few of the design decisions but they were more to do with it feeling less like a Pokemon game than a roguelike one. The difficulty (bonus dungeons aside) was aimed at new players but still required some tactics, so it's a good gateway roguelike for people unfamiliar or struggling with the genre.
Procedurally generating a story is easy. Most stories can be broken down into a series of small events, with rising tension. If you programmed enough events into a game that stayed within the context of dungeon crawling, picked a handful randomly and then swapped a couple so that the grander scaled ones were generally last, there's your story. The hardest part would be linking the events together, but you could probably get away with using just a generic meta-plot (find the 7 Golden Foos to defeat a great evil) if the smaller quest sequences were interesting. *Telling* that story narratively with any believeable characters or emotional attachment whatsoever would be near-impossible, but even an obvious bare-bones 'random events plot' would still make for very replayable and interesting gameplay.
ReplyDeleteFor example, say you rescue an NPC from a trap in a dungeon. The NPC asks you to help find their friend who was also lost down there, and you agree. In a normal game this would be a simple escort quest or temporary useful ally (depending on the NPC strength), but if you know the game has multiple resolutions - you find the friend safely / you find the friend captive by a powerful boss / you don't ever find the friend and the NPC sadly gives up / you find the friend but it was a trick and both suddenly attack you - then you're going to tread much more warily, especially if the game programmer has been clever enough to include hints as to which plotline you're on. (If it's a trap, maybe the NPC AI only bothers to protect itself instead of attacking monsters who are damaging you.) After the resolution either the NPC would join you permanently for a greater quest (good ending), they'd tell you a rumour and leave (meh ending), you'd a find map on their corpse (killed by monsters) or you'd find a letter from the their evil boss (bad ending), any of which would direct you towards the same next event sequence.
Programming these different events would obviously take QUITE some time, and different dungeons and challenges (sieged city, ambushes etc.) would need very good scaling algorithms to stay balanced whenever you encountered them. It would all feel a bit artificial once the player starting recognising the events and patterns ("Out of place corpses again? Sigh, better head to their natural locale for a necromancer plotline.") but by incorporating the story with game mechanics ("Noooo, my meatshield buddy betrayed me. ;_;" or "Gah! The orcish invasion of elf village X has cut off my teleport scroll supply! I'll get those darn orcs. \_/") you could still have real emotional responses. By keeping track of player actions you could even generate the next chunk of story specifically to have an impact: Fled from a tough unique in the first dungeon? Guess who gets picked and level-scaled up to act as the head baddy for the next invasion plot.
That's the kind of roguelike I'd love to play.
Lots of other thoughts from this episode:
ReplyDeleteWhat makes permadeath a significant design choice rather than 'simply being able to lose the game' is that in roguelikes you usually spend much longer and put much more customisation into building up your character than, for example, your ship in an arcade shooter where you might grab 1 of 3 power-up types and be done in 20 minutes. This means starting over can be MUCH more frustrating (or winning much more rewarding), so it's not to everyone's tastes if they want to play a game more for relaxation than challenge. Funnily enough, this line of thinking makes coffeebreak roguelikes feel close to strategy/arcade games because of the reduced attachment (and thus risk/reward).
I hardly think roguelikes should be mixed with ANY genre. I can't even imagine how you'd add roguelike elements to a puzzle game like Zork, or a dating sim, or a 2d fighter ...
Mario is not game about decisions. It's about reactions and timing. Random levels are less important for real-time skill-based games because even if you are replaying the game and making the exact same moves it still tests your ability to time your button presses and whatnot. If there's a pit between you and the goal then you know on your first playthrough exactly what you do on your fifth; you need to run up to it and hit jump right before falling in. There's only so many variants so shuffling the order of tricky jumps with random levels doesn't really add anything.
^ Spelunky is different, of course, because you do have some limited resources to decide when and where to use. And there are other minor cases in eg. FPSs where a small part of the challenge is tactical as you need to stay alert for unknown enemies and clear areas safely. There, random enemy placement suffices.
Nethack is, if anything, an example of how NOT to do random loot. See: your inevitably identical ascension kit, random game-breaking wands of wishing, and gnomes picking up wands of death.
It's worth noting that the difficulty of recognising new enemies isn't entirely exclusive to ASCII. A player will recognise "that's a wolf" or "that's a dangerous coloured wolf" with graphics, but they're still not going to know how strong that wolf actually is in *this* game before fighting it. Roguelikes actually have an advantage here when they bother to include a useful description system (This wolf is [strong] for your current level. It moves [fast]. It can attack for [moderate] damage through your armour, or [howl for attention]. Etc.) straight up instead of monster memory or fancy literature quotes.
Another reason ASCII can be better than graphics particularly with tile-based games is that it allows the player's imagination to fill in the gaps. If you show the player a picture of an orc then they're mind is always just going to see that simple picture, but if you present the information with an abstract an indicator as an 'o' then running into a room full of them will cause the player to imagine their own scary and detailed scene rather than just seeing a square block of copy/pasted orcs. It's the same reason that a good book can be just as if not more engrossing than a flashy movie. There are ways graphics can achieve a similar impact, but they're time-consuming and difficult given the roguelike interface.
I've played one Pokemon Mystery Dungeon game and thought it was fun as a casual roguelike. There wasn't permadeath but there was still a harsh item penalty for dying and you couldn't just reload, so it was a milder version of the same overall feeling. I disliked a few of the design decisions but they were more to do with it feeling less like a Pokemon game than a roguelike one. The difficulty (bonus dungeons aside) was aimed at new players but still required some tactics, so it's a good gateway roguelike for people unfamiliar or struggling with the genre.