(untitled)

What do you think of roguelikes, Lain?

Used to spend a lot of free time in an iOS port of NetHack. Lowest I've gotten was only about level 9-10 I think. There's one YASD that I remember very well. I was a vampire barbarian and was doing really well. Got to a personal best depth, and still going strong. I usually had problems with food managment and would often die of starvation, but this time, as a vampire, I could just eat the flesh of my enemies so didn't have to worry about securing rations. Just kill, drain, kill, drain. This was the game I learned what happens when you eat cockatrice meat.

I'm still playing the Doom Roguelike. So many challenge modes, so many easter eggs...

Nethack was the first video game I took seriously and sank over 100h into. My first ascension was one my proudest achievement at that point, and my last one with a tourist even more so. Afterwards I quit Nethack because each play takes so long, and the late game is a bit boring ; which is only fixed to an extent by Slash'em. Even now that I try to focus on genres with quicker gameplay, such as danmaku, some of my favourites could be considered at least roguelite. Faster Than Light, Binding of Isaac, Don't Starve, Dwarf Fortress. >>762 You know that you can eat corpses even as a non-vampire, right ? (Except for the poisonous ones of course, and the hallucinogenic ones, and the petrifying ones, and the rotten ones, and the ones that drain levels, and ... ) The whole "Taste terrible" thing has no effect. Stick to mammals' corpses if you don't want to read spoilers and still avoid nefarious effects.

>>764 Actually I totally forgot, and your post reminded me of this - it was actually a Slash'EM port, not just vanilla NetHack! >You know that you can eat corpses even as a non-vampire, right ? I knew you could eat them. I thought that for most races there'd be random negative effects. I didn't really manage to get a good hang of the mechanics, so if you're saying there was a way to tell which corpses were safe to eat, I never figured that part out, haha.

I enjoy roguelikes, ADoM, Nethack, and DoomRL are my regulars.

>>761 I really enjoy them! My all time favorite is DCSS (Dungeon Crawl Stone Soup). Currently (as of a few months ago, haven't checked recently) I've won with every species in the game at least once. It is a really fun game, although the learning curve can be rough. It definitely feels like one of the most "fair" roguelikes I've played, in that a good player can almost always (if not always) win, regardless of how bad their RNG is. If it counts, Cataclysm:DDA is also pretty fun, although I haven't put nearly as much time into it as I have DCSS. Also enjoy dwarf fortress, and Faster Than Light was fun for a while.

I played Cataclysm for a week or two, I don't much care for survival horror but its games last 20 minutes at most so it cuts down on the tediousness. If you enjoy a game that surprises you a fair amount for the first 30 hours of play, you'll like it.

>>777 >its games last 20 minutes at most I wish ! Once you get going there is very little than poses a threat, and there is no end-game to push you to get out of your comfort zone. I sank a few dozen hours in it to max all skills before realizing I was just bored. Now I try to help on the source code instead, it's more interesting.

I enjoy nethack, dwarf fortress, CDDA, and I really enjoy cogmind. I love the developer, and the game is totally worth supporting. DoomRL also tickles my pickle on occasion

Sil is the best.

Not to brag or anything, but today I ascended and made it into the top-ten score list, on dank.ninja, playing Nethack Fourk. One thing that I love about nethack is to be able to play through ssh on remote servers. The single most stealthy way to play at work : nothing installed on the local machine, and who is going to notice this ssh connection amongst the dozen of others ? And if anyone asks, you're just monitoring your home server.

I play DCSS off and on. It's a great game, but the recent philosophy of the dev direction has certainly left a sour taste in my mouth.

>>769 DCSS represent

>>761 They're excellent, of course! I play Pixel Dungeon on android. http://pixeldungeon.watabou.ru/ >distributed under GPLv3 what's not to love?

>>768 I have a love hate relationship with ADOM. I love the setting and world of that game but hate the grind and those stupid deaths after several hours of playing the game. I remember two deaths that made me furious. The first was a promising human farmer that died on the banshee level because it was the first time I got to the banshee level. The second was a dwarf weaponsmith who managed to kill vampire lords but died on a random encounter fighting a giant boar. Now I mainly stick to coffebrake roguelikes because of real life things I have to do but still enjoy them very much. DoomRL is perfect for this because I can play it very fast and don't have to remember that there is a white altar in the small cave on level 3 if I happen to save the game and leave it for a week. Other then that I enjoy Infra Arcana (similar to DoomRL but a litle slower paced, also set in the lovecraft setting) and Slimy Lich Mummy (kind of quirky game and not well polished, but a session are about 20-30 min top for me so that is perfect).

Has anyone here played Lifeweb, or the Colonial Marines mod for Space Station 13? Jupiter Hell is supposed to be coming out soon.

>>952 There's also a fork, called Shattered Pixel Dungeon, which is actively developed. It's available on f-droid as well.

I love they TBOI its my favorite

The only ones I really play with any consistency nowadays are Elona and DF, and occasionally DoomRL. I used to play a lot of ADOM and Nethack back during my first job when I could just bring PuTTY on a USB and jump on whatever servers had them available. On the roguelite side of things, I've been playing this one called Ziggurat, it's basically an FPS that draws inspiration from Heretic and Hexen for its designs. I've tried to make my own before, but the project turned out to be way too ambitious for me to even make a playable pre-alpha within a decent timeframe. Coding the basic framework was kinda fun, though. I recommend making at least a simple one, it's a good way to test your ingenuity.

>>906 >but the recent philosophy of the dev direction has certainly left a sour taste in my mouth Could you elaborate on this?

There great games. I mean there mostly free (at least the most classic roguelikes are free), they focus on gameplay and interesting mechanics (and not on flashy marketing, hype and graphics like most AAA games) and you can sink hundreds of hours and don't get board. It true that they are less accessible then other kind of games but it started to change (DoomRL would be a good example of this). >>1104 >I've tried to make my own before You are not the only one. I'm still planing to make something one day (I mean I have the programming skill to do this but I lack motivation). If your interested in development of roguelike there is a great podcast I recomend: www.roguelikeradio.com

I played the old DoomRL until I started jumping every time I saw a capital C JESUS CHRIST IT'S A CYBERDEMON GET IN THE CAR OH MY GOD THERE'S THREE MORE

>>1176 Always happy for an excuse to bitch about the crawl devs. Since about version 0.15, they've been consistently removing or nerfing powerful effects and interesting interactions. >tree form can't teleport - devs didn't like the "teleporting tree" incident, where some clever person used tree form + controlled blink to trivially clear one of the game's hardest levels >=tele is no longer evokable - it teleports you at random intervals toward random enemies anywhere on the map, and you used to also be able to activate it at will. obviously not a good item, but it could maybe save your ass if you're willing to deal with the many downsides. now it's not evokable at will and is nothing but a junk item to pad out the ID game >regen, mana regen, and gourmand jewelery have some bullsoykaf "charge up"/"attune" mechanic instead of taking effect the moment you put them on - devs for some reason hate the idea of players swapping gear depending on situation, so you can't swap in =regen when your health is low for example >no more haste, healing, or tele wands - straight-up nerf for all characters, these wands were the most useful in the game >no more haste spell - major nerf for hybrid fighter/caster builds, can't "haste up and then hit them with your sword" any more >no more free infinite sif channeling for undead/lichform characters >haste no longer stacks with oka's finesse - this used to give a total of 3x reduction to attack speed, but now you only get the 2x from finesse Short version is, half the fun of crawl was progressing from an XL1 wimp who's scared of gnolls to a nigh-unstoppable killing machine who can blow away demons by the hundred without even blinking (as long as you use your resources intelligently and don't get in over your head). But for a few years now, the crawl devs have been consistently nerfing everything strong so that the player experience is like kicking dead whales down the beach equally at every level.

>>1372 >no more free infinite sif channeling for undead/lichform characters I hadn't heard about that change. Never got to try a classic Sif/Lich build before, guess I never will now. >haste no longer stacks with oka's finesse - this used to give a total of 3x reduction to attack speed, but now you only get the 2x from finesse Also missed this change since I never play Oka. Again, one less reason for me to branch out and try a different diety. Thanks for the summary. I was pissed about the =tele not being evokable anymore too.

I like the idea of cogmind but I haven't put much effort into it since only windows

>>1373 >Never got to try a classic Sif/Lich build before, guess I never will now. Well, you could always load up 0.18, which is just before the version with the Sif rework. It's also prior to the removal of the haste spell and the good wands.

>>1372 Also the gave Mountain Dwarfs the axe.

>>1411 I think that was a good thing they removed them. I'm getting sick of the fantasy stereotype created by Tolkien where elfs are good tree-loving hippies, dwarfs are miners who drink a lot of beer and orc are evil savage creatures. They literally use those same sterotypes in ALL of the fantasy. By removing the the Mountain Dwarfs the developers made room for a new race, the Deep Dwarfs which have a distinct gameplay mechanic (they don't heal over time). You could argue that they could have added the Deep Dwarfs and leave the Mountain Dwarfs but that would cause the kitchen sink effect where you add things that aren't very distinct from each and other (NetHack had/(or still has, I don't know) the list of polearms copied from a D&D manaul and that list was like 20 types of polearms which worked exacly the same). Minotaurs could be seen as almost identical with the Mountain Dwarfs, so by removing the Dwarfs they didn't change the gameplay of the game and only flavor of the text displayed. Also if you can't live without the stereotypical dwarf in Crawl you can fork the game and change the name Minotaur to Mountain Dwarf.

>>822 I recently started playing Sil again since the Android port came out. Thief characters are a lot of fun.

>>761 Are there any notable sci-fi themed roguelikes?

>>769 Every species? Even the cat?

>>1418 DoomRL

I’ve been playing a Metroidvania roguelike lately named Rogue Legacy. It’s reallt fun.

>>1418 Caves of Qud is worth checking out

>>1463 Live and drink, wayfarer. >You pass by some brinestalk >You pass by some brinestalk >You pass by some watervine >Your skin itches...

DoomRL is very good. Not too big a fan of how much additional complications they added with all the classes, crafting systems and soykaf, but it's still, at its core, a very fast paced and fun roguelike which plays surprisingly like Doom. I wonder if Jupiter Hell, its spiritual successor, is going to suck.

>>1508 I like DoomRL a lot. It's too bad Zenimax sued them.

>>1526 The author just changed the name to DRL (from DoomRL) and released the source code, so nothing really happend. He is developing a spiritual successor called jupiter hell so you might be interested in that.

>>1527 It's the principle more than anything. I just think it's soykafty when entertainment companies use copyright to kill fan projects. They weren't going to sell fewer copies of Doom because somebody made roguelike based on it.

Recently I have been playing Cardinal Quest and Cardinal Quest 2 on my moblie (there is a steam version but it's not free) and I have to say there are very fun. Both games are free to play (made by different people so the experience is different) and thay remind me of Diablo and Diablo 2 a lot. The first game is set in the dungeon the second is more open world and have more heros to choose form. The gameplay in both cases focus more on short term tactics then long term strategy. All of our skills have a cooldown counter and don't need mana to use, what more there is no hunger clock (but also no HP regeneration) so you will probably start every encunter with all of your skills ready to use. The only complain I have is that there quite easy to win (probably because they focus on tactics and not long term strategy). Overall I highly recomend to anyone to try them out. > Currently playing Cardinal Quest 2, ACT 3, Ranger class

I recommend the podcast Roguelike Radio for fans of the genre. Because the roguelike development community is so small, they've been able to interview pretty much everyone who's anyone within said community. (They even got Richard Bartle for one episode, even though he's only tangentially related to roguelikes.) Further, the developers tend to have academic backgrounds, so there's some very interesting, in-depth discussion of the underlying theories of roguelike development. It's obvious that the roguelike genre is one of the most uninteresting genres visually, but what may not be immediately obvious is that it's one of the most interesting genres theoretically. I think the developers may have more fun making the games than the players have playing them.

>>1971 Right you are, it's a great podcast. The best episodes IMO are the ones interviewing the roguelike creators. Also from time to time i discover a new roguelike I would like to play thanks to this podcast. Recently I read somewhere that the podcast has a shortage of people who are willing to edit the audio for them and that is the reason why they now produce only about 10 episodes each year (as opposed to 20 avg. in the years 2011-2016).

>>1971 Roguelikes are unique in that almost every fan has tried their hand at being a developer too. I can't think of any other genre than has such a dynamic, but it leads to a great community and lots of interesting little games to enjoy. This is probably the source of my main gripe with the recent appropriation of the term roguelike These newer games (roguelites, whatever you want to call them) are great, but it's not like they unearthed some long-forgotten design pearls; there has always been a vibrant roguelike community with a shared understanding of what exactly a game like rogue is. An understanding very different to that held by video game players today.

I've been a fan of roguelikes since I found NetHack, and now, when I hardly game at all, they are the only games I'll play. I might put in a run every now and then in DCSS or NetHack, or Brogue, and then not touch games for a month. I love em and they're convienient for that kind of play. DCSS is probably my favorite but >>1372 is right, and it started sucking after 0.15, that's why I still play 0.14.2, after 5 years of updates, and I'm not looking forward to the day when I update and it stops compiling. When that day comes I guess I'll finally get around to checking out the new versions of NetHack that have come out since 3.4.3

Anybody have any experience setting up or running a roguelike server like crawl.akrasiac. or nethack.alt, a lain roguelike server would be pretty tight. Also it's really weird living in a world where Nethack actually gets updates, one just came out 3 days ago. Gonna have to try it out, wish me luck.

I really don't like em, I much prefer a game with a start and an end, rougelikes feel to repetitive, addictive, and time-wasting

>>1976 This is what all games are, it's just the ability to disguise the loops. But respect, I know these things aren't for everybody, and don't hide the loop for a lot of people, especially since they're usually pretty bare when it comes to presentation.

>>1976 I'm not sure what you mean by a start and an end. Roguelikes have a clearly defined win state; get to the bottom of the dungeon and retrieve the mcguffin.

I like roguelikes. The randomized item identification early sub-game is a bit hit or miss with most though. I can neither confirm nor deny whether I'm developing a rouguelike, but I think one with elements of Lain would be great. There are certain technologies that have yet to be used to create roguelikes with (or even most games). I think we could add platform features that translate into new and interesting in-game mechanics. Sadly those capable in "bleeding edge" tech typically don't make roguelikes, and gamedevs are primarily inept in the areas of higher tech outside of rendering systems. For instance: Just mention "neural networks" in the same breath as "gamedev" and you'll be laughed out of any "serious" game developer's presence. And yet, advanced behavioral prediction AI used to deliver adverts could easily be re-purposed for NPCs (and hasn't). That's just one example. There are many such unexplored technological realms. Roguelikes happen to eschew "muh graphics" in favor of mechanical richness, and if taken to the logical conclusion, could provide a proving ground for certain new technologies that would leverage the GPU for something other than putting shiny bumps on top of shiny bumps. Of course my statements apply to some other genres, but few games have the complexity of a roguelike and can also lend themselves to turn-based gameplay.

>>1982 >The randomized item identification early sub-game is a bit hit or miss with most though. They did a whole episode on this for Roguelike Radio if you're interested. >That's just one example. I've been doing my own research for, um, my own purposes, and I agree. As far as I can tell, most game devs just thoughtlessly do what everyone else is doing. I keep coming around to the same idea: games could easily be better. AI in particular could be much better. Devs keep dodging the problem by saying that players don't like smart AI because it beats them too easily, but that's why you design the AI to maximize player fun, not to win every time. If you're interested in how AI could be improved, I recommend the book Behavioral Mathematics for Game AI. It's obviously more about applying more appropriate models than applying new technologies, but it all ties together, I think. What are some other technologies that are underutilized, in your opinion?

>>1982 >For instance: Just mention "neural networks" in the same breath as "gamedev" and you'll be laughed out of any "serious" game developer's presence. And yet, advanced behavioral prediction AI used to deliver adverts could easily be re-purposed for NPCs (and hasn't). That's just one example. There are many such unexplored technological realms. I can't say I agree with you that putting "neural networks" in games is the way to go. Making such a system is time consuming and that time could be used somewhere else. Also lets look on the average time you interact with a NPC. If it's a monster then you just try to kill it, and lets say you will kill it in 10 moves after you see it. There isn't a lot of room to show of complex intelligent behavior in that time that you can't make with a simple state machine (and the state machine will be simpler to make). It's not even said that a neural network AI in a game would be better then a simple system because more often then not video games have a clearly defined problem to solve (kill the player, and to do that I need to attack the player) so it's easy to make rules for such a system to make it look intelligent. Neural networks on the other hand shine in areas where the problem (or rather the solution to the problem) isn't clearly defined: is this a picture of a dog, is this a picure of a man wearing a fedora hat, hay self driving car can you take me to the bus station, ... and so on. Let's look at a example. In Diablo 2 you had a monster called the Fallen and Fallen Shaman. The Shaman would try to stay at the back row and throw fireballs at you resurrecting the dead while the Fallen would try to swarm you but would retreat if you killed one of them nearby (as there ware cowards). After a few sec. they would again try to attack the player. This is a very simple system (you could probably program this in a few hours) that it tries to simulate intelligent behavior and if you ask me it makes a good job. Sure we could make a neural network that acts the same but it won't take you a few hours to make. Could it be better (as more intelligent)? Maybe, but Fallens and Fallens Shamans ware pushovers that died quickly so you don't have a lot of time to show that nice neural network logic to the player because all monsters will be dead in a few seconds. You should now add that it's just one of many types of monsters in a game and it would be nice that they acted in a different way the the Fallen and Fallen Shaman so now you need to make a few other neural networks (each take time to develop) that will be used for other monsters. I would just like to add that I don't think that Diablo 2 had some good AI, but I think the example fits perfectly in the discussion (simple logic can be seen as complex behaviors).

>>1988 I don't think you're seeing the big picture. >If it's a monster[...] You may have a point, but your entire post after this point talks about monsters. Let's say we're not just talking about monsters. You're making the mistake of assuming that one AI has to correspond to one NPC. I would point you to the Director from Left 4 Dead 2, which was used to coordinate enemy waves in order to keep an optimal amount of pressure on the player, and the squad manager from FEAR. As AI gets better, it can act more and more like a gamemaster, among other things. >[...]you don't have a lot of time to show that nice neural network logic to the player because all monsters will be dead in a few seconds. Ok, but what if the network persists across incarnations of the same enemy type? One incarnation learns something useful and dies, but the next incarnation retains that "knowledge." I'm not even arguing that we necessarily need neural networks, but I think we can do better than state machines. It just seems like the same tired doctrines and perspectives are being applied in the field of game AI. I don't even think planning AI has been sufficiently employed, to say nothing of more modern technologies. (To head off any confusion, I'll inform you that I'm not the guy you responded to.)

>>1988 As an unprofessional but experienced gamedev i agree with this. Neural nets may seem at first glance like a great way to make an ai. However they are overkill for most games, say as fps and rts. They certainly could be used for games where very advanced gameplay mechanics are involved, but i know of no genres where this really would be useful. In games like fps and rts cheap algorithms can do miracles, more advanced behavior would be hard to implement, the most obvious choice of use would be for terrain/cover, but that requires a significant framework to account for such factors.

>>1990 >As an[...]experienced gamedev[...] I don't want to personally attack you, but you seem to be trying to build credibility by making this claim, so I'd like to ask: are you a good gamedev? Do you have experience with making games people like? With breaking new ground effectively? Experience doesn't mean as much in this field as it would in a field like baking. If you've been "gamedeving by numbers," then the value of your experience is seriously questionable. This is a creative field. It seems like you're saying that we don't need new technologies to effectively keep doing what we've been doing, which is true but obviously beside the point. We shouldn't be asking "Do we need this technology?" but instead "Now that we have this technology, how can we use it?"

>>1989 >but your entire post after this point talks about monsters. Fair point, but this is a thread about roguelikes and monsters are a mayor part of those games. We also could look at friendly NPC. The merchants and actors who give quests are even less interesting as they just wait till the player ask them something or tries to buy/steal a item (you don't really want to see the merchant pack his store and leave to find a city with more customers). Making a NPC that travels the world in the quest to find the player just to give him a quest is also a bad idea (most of the traveling of this NPC would be made behind sceens, and the player wouldn't know about this). The most promising NPC that would benefit from a better AI would be in this case a companion that travels with the player and gives him aid. But in this case you might want to have some control over your minion (at least change the tactics or order him to move someware or attack someone). That limits his decisions a lot. Other then that you want him to attack enemis, don't let himself killed to easly and heal you if it's capable of this. Again not a lot of room for advance intelligence just some general rools the actor should fallow. A game that does that well enough in my opinion was Fallout 2 (not a roguelike but still), and they rather didn't used anything fancy. You could talk to your companions and tell them that they should watch out not to hit you, or when they are allowed to use healing. Other then that they acted as they "thought" fit. But ok, lets look on different genres as roguelikes tend to move out of there dungeons (pun intended) and part of the mechanics from roguelikes show up in different game genres. The Director from Left 4 Dead is a nice example but it also doesn't seams to need the full force of neural networks. If player is getting his butt kicked the tone done the waves of zombies if his murdering the zombies without a sweat then send more next time. Hell we could remember how much troubling were each type of zombie (how much damage was done to the player or something) and choose are tactics accordingly. The thing that would be difficult to do with a algorithm is finding the places where zobies hide (like the one that used his tongue pull you, sorry I don't remember his name, I played L4D like 10 years ago), but those parts are easly done by the developers (they can create muliple spawn points and the system can choose randomly from them). I never played FEAR (shame on me I know) so it's hard for me to coment on that. >As AI gets better, it can act more and more like a gamemaster, among other things I was once thinking about a system that would give you items or put monsters on the level to dynamically set the challenge level but then I understood that I didn't like the idea. The problem with that system is that it's very easy to make a system that holds your hand and thats not what I really like (give me Demon's/Dark Souls any time). >Ok, but what if the network persists across incarnations of the same enemy type? So the AI would learn? Sounds good on paper but I'm sceptic with that. Lets assume the AI would be persistent between games. The problem is that the system should not only get better but also worse(to oscillate near your skill level). If we would only allow it to get better it could find a optimal strategy to kill the player each time. On the other hand (assuming we let the AI get worse) the player could learn to train the AI in a way that the game would become simple for him (start 10 games, die in a stupid way, that leads the AI to take it easy on the player, and then use the real strategy to win). Also it could lead to a situation where the only strategy left for the player are very boring (and it won't change because the AI oscillate it's strategy very closely). If the AI wouldn't be persistent between games then maybe we wouldn't have this problem but would the game lenght be enough to let the AI create a interesting strategy? Overall there are a lot of IF and BUTS. As for me I don't think implementing a very complex AI system makes sens as this isn't transferable to player having fun (well except maybe for strategy game). The game with great AI might be good or bad but if the AI is good enough then I would prefere the devs to use there time to polish other parts of the game (as with everything it's a trade off betwean what you wan't in a game and what you have time to implement). The subject of AI in game in general seams interesting and if you have some other thoughts then please let me know.

>>1992 >they just wait till the player ask them something or tries to buy/steal a item This is the traditional way but not the only way. In fact, it's probably only traditional because of technological limitations that don't necessarily exist today. >you don't really want to see the merchant pack his store and leave to find a city with more customers You might not want this as an isolated case, but if the rest of the economy is as robust as that, things become interesting. >Making a NPC that travels the world in the quest to find the player just to give him a quest is also a bad idea You're only thinking in terms of old paradigms. Even the concept of a quest can change with time. If the underlying systems are sufficiently advanced, "quests" may develop organically. Roguelikes are great proving grounds for these kinds of systems. Of course, we're beginning to touch on a different cornerstone of roguelike design: procedural generation. >The most promising NPC that would benefit from a better AI would be in this case a companion One of the only games to implement machine learning was Black & White. It was a pretty rudimentary implementation, but it's something to look at if you're interested. >I never played FEAR (shame on me I know) It's not shame on you, but it is a shame. It was the first game to use planning AI and, in spite of its age, it still has some of the best-regarded AI in a game, in terms of player experience. People have be labored the point that it isn't as smart as it appears, but that's missing the point that, whatever tricks were employed to create the effect, it feels very smart in a way that game AI almost never does. Off-topic, it's also an achievement in terms of game aesthetics. The lighting, effects, etc. are still impressive. It's really not one to skip, as it hammers home the point that player experience is king, not to mention that it's an uncommonly good example of game elements working in concert. The principles behind FEAR apply to more than just FPS's. >it's very easy to make a system that holds your hand It all depends on the parameters you're using. It's easy to make any system in a game work poorly, but if it would do something that you want done, you just have to tweak it until it works for you. >the player could learn to train the AI in a way that the game would become simple for him To begin with, I'd like to say that the player can exploit traditional AI and often does. Certain countermeasures can be employed. As a rudimentary and hasty idea, you could only have it respond to the player's best runs as measured by points, time spent alive, or other metrics. I don't mean it rudely when I say that we have brains so we can solve problems. The first problems that come to mind should be met with focused thought. Anyway, there's a certain amount of "gaming spirit" that the player should have. As a roguelike-specific example, save scumming is a known trick, but you're not supposed to use it. It's easy to break a glass if you try, but it still holds liquids admirably when you don't. I wish we could get past the idea that the player is an asshole and thus everything must be player-proof. >it could lead to a situation where the only strategy left for the player are very boring This is a good point. You're right. That's a common hazard of traditional AI design as well, though. >As for me I don't think implementing a very complex AI system makes sens as this isn't transferable to player having fun If you do it wrong, it isn't. If you do it right, well, things may change. >as with everything it's a trade off betwean what you wan't in a game and what you have time to implement I just think we need more passion projects. Not having to keep a timetable, not having to worry about profit, and not having to expect a general audience: all of these would theoretically serve to open up the possibilities, creatively and mechanically speaking. Look at the complexity of Dwarf Fortress. That's the result of focused passion over a long time. If you spend a lifetime crafting a great novel, you're a genius. Spend a lifetime hammering out a game, and you're an idiot, apparently. I think there's a general lack of purely artistic motivation within game development. Maybe I'm wrong. >if you have some other thoughts then please let me know. I recommend looking at Raph Koster and Richard Bartle. These two are specifically virtual world design experts, but they have a lot to say about game design in many fields (virtual world design is pretty all-encompassing, after all). Bartle in particular mentions a few times in his Designing Virtual Worlds that developers mostly just copy old ideas with no understanding of why those ideas developed in the first place. He's surprisingly forward-thinking for the fact that he's as old as he is. Both of them touch on AI and virtual economies, which you mentioned at the beginning of your post. Here's Koster on NPC design: https://www.raphkoster.com/2006/06/09/why-dont-our-npcs/. That blog is a treasure trove, if you like what you read.

>>1988 >If it's a monster then you just try to kill it, and lets say you will kill it in 10 moves after you see it. There isn't a lot of room to show of complex intelligent behavior in that time that you can't make with a simple state machine (and the state machine will be simpler to make). It's not even said that a neural network AI in a game would be better then a simple system because more often then not video games have a clearly defined problem to solve (kill the player, and to do that I need to attack the player) so it's easy to make rules for such a system to make it look intelligent. Starting with a base neural network, each entity has a few bits of its neural net randomized. A scoring system ranks enemies of that species, choosing the best N enemies to act as the base for new enemies. No state machine literally evolve to adapt to player and/or environmental behavior, but genetic programming and neural networks both can. Player created map / environment. AI more advanced than state machines can populate it & learn to survive there, develop ecosystems, etc. Not enough time? OK, how about your leader board high score playthrough also updates the global enemy AI database? When the problem is survival, and the environment dynamic (as roguelikes can be) there is no clearly defined solution. AI was just one example of new technologies I'm playing with. As for the rest of the comment: You're so far inside the box I'm afraid you'll never think your way outside of it. Yes, you can program simple systems, and a script kiddie can bot them in an afternoon. So what? Protip: If you can't think up any new game mechanics then I don't value your input. Any idiot can clone some existing game logic, some is foundational but if all of it is, you're just re-skinning someone else's game. Any fool can shoot down a straw man of someone else's idea. I'm not going to waste time posting here. I prefer to shut up and code & let the output speak for itself. However, my initial results have been promising, and I thought that I'd mention that there sure is a lot of area to explore that traditional gamedevs are just too ignorant and inexperienced about to venture into. >>1990 Thanks for proving my point. Just mention "Neural Networks" and close minded simpletons who haven't even implemented them will start shouting about how bad it is. Muh "Neural networks" are just one form of machine learning. There are many cognitive models beyond [weighted] state machines or decision trees that can be explored. The problem is: Gamedevs aren't pros as actual AI. They don't even really know what there is out there besides "muh neural net" or "muh deep learning". I won't namedrop other tech, it's a waste of time... seek and ye shall find. Maybe go play with Noble Ape Sim for a while, its ape-script is pretty amazing. http://www.nobleape.com/sim/ Not using that codebase at all, but it was something fun I played more than a decade ago that had some ideas still not seriously explored by many gamedevs.

>>1991 >I don't want to personally attack you, but you seem to be trying to build credibility by making this claim, so I'd like to ask: are you a good gamedev? Im not a good gamedev to be honest, worse than my peers. I wasnt meaning to build crediblity of some sort based on that, merely point out i have experience. >We shouldn't be asking "Do we need this technology?" but instead "Now that we have this technology, how can we use it?" I disagree with that statement, the technology is only a means to an end, there is no point in using it unless it can be used properly. I just have a hard time imagining a game that is so sophisticated that warrants a neural net ai.

>>1994 >No state machine literally evolve to adapt to player and/or environmental behavior, but genetic programming and neural networks both can. And can you give examples in what way would this system evolve? What would it learn? Because how I see it you may give a player counters that tracks how much he uses magic/range combat/melee combat and have 3 state machine for each creature to use countertactics against that playstyle (you could even look on the player weapon in hand). If we are talking about environment you could also add recognizing the environment in the state machine that would modify the tactics. For me it sounds a litle like you don't know what you want the system to do so you want to let the computer make it for you. >Player created map / environment. AI more advanced than state machines can populate it & learn to survive there, develop ecosystems, etc. Sounds like your more interested in making a simulation then a game. I'm not saying that there can't be a game and a simulation in one but: >The most important component of being a roguelike game is actually being a game. > - Jeff Lait in rec.games.roguelike.development >When the problem is survival, and the environment dynamic (as roguelikes can be) there is no clearly defined solution. There is because you know (as a developer) what the environment is allowed to do. If you won't allowe it to rain fish the game won't disobey you so it have a bigger chance to kill the player. Unless you also plan the environment to be generated by some super AI (maybe you want to make a AI that makes games instead, those kind of programs ware made in the past: https://www.rockpapershotgun.com/2018/07/04/angelina-is-an-ai-that-makes-games/ ). >You're so far inside the box I'm afraid you'll never think your way outside of it. Maybe your inside a box yourself friend. You discard the idea that you don't need a complex AI to make the game engaging and complex. I for the other hand am trying to have a civil conversation without attacking anyone. >I'm not going to waste time posting here. I prefer to shut up and code OK >However, my initial results have been promising, and I thought that I'd mention that there sure is a lot of area to explore that traditional gamedevs are just too ignorant and inexperienced about to venture into. Sounds good, can't wait till you show the world your creation. >>1993 >you don't really want to see the merchant pack his store and leave to find a city with more customers >You might not want this as an isolated case, but if the rest of the economy is as robust as that, things become interesting. If the economy/trading is a mayor part of the game then it might be interesting, if we wan't to make a hack&slash roguelike were you enter a dungeon and kill monsters then it could be frustrating that you have to travel to a different town just to buy your stuff. >You're only thinking in terms of old paradigms. Even the concept of a quest can change with time. If the underlying systems are sufficiently advanced, "quests" may develop organically. Roguelikes are great proving grounds for these kinds of systems. Of course, we're beginning to touch on a different cornerstone of roguelike design: procedural generation. If the system would be as advance as to make better quests then "kill X", "find Y" then yes that would be great but the technology isn't there yet. You can't blame game devs for not trying to implement those systems in there games as it's not really there goal to do this. There goal is to make a game. You probably would need some company that creates the tech and then sell it to game making ompanis (like the unreal engine is being sold). The problem with that it's a big investment. >One of the only games to implement machine learning was Black & White. It was a pretty rudimentary implementation, but it's something to look at if you're interested. I played it (the first part) but in my opison the systems didn't worked well. I had moments where your familiar did somthing stupid, and no amount of punishment seemed to make it stop (probably a bug of some sorts). Also I didn't finished the game because I encountered some other bug that prevented me from playing further. I presume that BW2 had a better AI then BW1 but never played that. >To begin with, I'd like to say that the player can exploit traditional AI and often does... Yes, but in this case your creating somthing as a meta game. You encurage players to play bad so they weaken the system. So for a big part of interaction with a game you don't really play the game, you play the meta game. When you exploit the tradional AI you still play the real game. I'm not sure if I'm clear on this or not? So in this case you have to decide if your OK with making such a game or not (a lot of players could be push away from something like that). >it could lead to a situation where the only strategy left for the player are very boring >This is a good point. You're right. That's a common hazard of traditional AI design as well, though. Well but in a traditional AI it's easier to test it. If a AI is comlex then it's harder to test and decide if there are such dangers in the game. If the AI learns then it's even more complex to test it (you can't say for sure in what it will evolve). >I just think we need more passion projects. Not having to keep a timetable... And I agree with you 100%. I would love to see more of those kind of games, but reality I'm afraid is different. >Spend a lifetime hammering out a game, and you're an idiot, apparently. Not for me. If the end product is good then your also a master mind. >I recommend looking at Raph Koster and Richard Bartle... I didn't had time to look at it yet but i will in my free time, thanks.

>>1996 >If the system would be as advance as to make better quests then "kill X", "find Y" then yes that would be great but the technology isn't there yet. Even handmade quests frequently have this kind of simplicity. I think pretty much any quest that follows a template can be generated without involving any fancy new technology (like neural networks). Now would be a good time to mention that my actual technical background is very limited, so I'll have to speak abstractly, but I imagine a Mad-Libs-style generator that has an understanding of the properties of the "words" it's using. You might look at Dwarf Fortress' Adventure Mode; I think it has something like this. Anyway, even a "find this" quest can be made interesting if the path to the mcguffin is sufficiently complex. Remember, one random generator can access other random generators further down the chain in order to flesh things out. It's a problem for procedural generation like any other. For more about the principles and applications of procedural generation, there's the book Procedural Generation in Game Design, which was written almost exclusively by individuals from the roguelike community it's on libgen. Here's a question for the thread: why don't we see more procedurally-generated monsters? The idea came to me when I was reading the fantasy web serial Mother of Learning and the main character was exploring a dungeon. The main danger for him was the fact that every new creature he encountered had unknown abilities. Capturing that kind of danger in a roguelike seems fun, in my opinion.

>>2000 >Even handmade quests frequently have this kind of simplicity. Yes, but only the core of the quest is as simple as that and we ware talking about making a complex system for creating the questes by the computer. If this complex system would just make simple quests like "kill X" then why do you need a complex system to do that? We can automaticly generate such quests with a simple systems: >pick a random location >pick a random monster >pick a reason for a quest (the monster kill the wife of NPC or stole his keys...) >all done you have a simple quest Given a big list of locations/monsters/reasons for a quest we could generate hundreds of quests. There is a problem with that "quest system". It will generate quests that are boring and very similar. Now lets look on a quest made by hand but not in a roguelike (I will get back to this why not in a roguelike in a second): >a NPC asks you to kill a goblin as it's stealing food from his house >the NPC points you to a local cave >when you get to the cave you see traces of blood and it looks that the cave was populated by more then one goblin (a tribe, but somthing/someone killed them) >You start to dig deeper in to this and find out that the towns man didn't like that the local goblin tribe ware catching fish in the local lake (as there ware scared that the fish would run out) so thay attacked the tribe and killed them. One of them got away and they can't seam to track him down. >You as a hero are better at tracking and you find the goblin, he is wounded and is unable to catch fish until he recovers. >You find out that the goblin has a magical item that helped him escape and avoid being track down by the towns man. >Now you can kill the goblin and take his magical item (still the goblin could put up a fight) >OR Heal the goblin >OR confront the towns man and convince them to leave the goblin alone (maybe some of them would attack you as it would to them that you are some hippie goblin lover) in return the goblin could give you the magical item after he recover. So basicly from a "kill X" monster quest it turn into a complex quest/story that is interesting to play out, have plausible reasons to exist and it's all thanks to adding "road bumps" in the story (you don't find the goblin in the cave, the story of the towns man don't add up at first...). Now this would work great in a CRPG or a Pen&Paper RPG but not so much in a roguelike (assuming it would be made by hand). The reason is that you tend to replay the same segments in a roguelike over and over again so if such a "complex" quest would show up in the game it soon would lose all it's charm (you would know the outcome and it would become a chore). Now here is where we would need a really complex system to generate those quests. We would need something that adds those "road bumps" in the quest but it couldn't repeat it self and the obstacles in the story would need to fit in. I believe that sooner or later those kind of systems will be created. I remember that not long ago I read about a AI creating a story that scored a lot of points in a literary competition so if this is possible then maybe a complex quest system also is possible (not necessary made with neural networks). >why don't we see more procedurally-generated monsters? In a sens that the stats and abilities are random or in a sens that the computer created a 3 headed hydra-pig with wings? The first is done I think often (you have random stats, random life, random items the monsters possess, some games even have random abilities assigned to them like extra strong/fast and such). The second idea with a pig was I think made in No Man's Sky (not played it, but read about it). I read it didn't work out well as the monsters generated ware very similar to each and other (this one has a different head, this one has no tail and so on). The ASCII representation in this case could be helpfull as you would have to create only the random description for each creature but you may end up with no setting at all (as the RNG would create space orcs right next to rat that the hero needs to kill in his first quest). It could however be interesing. You got me thinking. I once read about the vision off some AI profesor that in the future computers and people would assist each and other to create better products. So for instance a person would ask a computer to generate a car body and the computer would generate a few skeleton projects and show it to a person. He then would pick one, maybe modify it and then ask the computer to do something with it again. The process would run a few times. Now we could do something similar with monsters or item describtions (a magical item could have some story writen in it like in Baldurs Gate items). Ask the procedural generation algorithm to create a new race of creature, you pick one modify it a litle and ask the computer to give the race some properties to make it stronger/weaker. This could lead to some interesting new settings. The game would ofcoursebe be released with static monster types but the creation of those new races would be random during development. I thinks ECS (Entity Components System) would be usful during this process (look up a talk from Thomas Bishop, creator of ADOM on the ECS).

>>2001 I think you've summed up some of the problems with quest generation pretty well, and I don't have much to say about it, so I'll just start by saying that I appreciate your input. You might want to look at the game Ultima Ratio Regum, which was intended to be a game about unraveling a plot to rewrite history, and the particulars would be generated anew with each game; consequently, the original vision of the game included both the generation of whole cultures and the generation of complex conspiracies, without bothering to include the usual roguelike game mechanics. In other words, it was to be a different sort of game, through and through, and it made some impressive strides, mechanically speaking. Sadly, it looks like it will never be completed, as the creator ran into too many personal difficulties, including serious illness and depression. >No Man's Sky You reminded me of this video: https://youtube.com/watch?v=vupCgB8H9Og It's obviously humorous, but I think it says as much about the pitfalls of procedural generation as I could in a few paragraphs. >rat that the hero needs to kill in his first quest What would you say to the idea that roguelikes could be improved by doing away with the early game? Suppose a certain assortment of powerful equipment was randomly or semi-randomly selected for you at the start and you were immediately put up against powerful monsters. I don't think this would improve every game, but I'd like to see a few games built around this concept.

A Windows update just rendered my DoomRoguelike unplayable, all the graphics and colours are messed up, I tried reinstalling, and rolling the update back, FUARK!

>>2001 >I once read about the vision off some AI profesor that in the future computers and people would assist each and other to create better products. So for instance a person would ask a computer to generate a car body and the computer would generate a few skeleton projects and show it to a person. He then would pick one, maybe modify it and then ask the computer to do something with it again. The process would run a few times. You usually imagine something like this, where the program's output is checked for quality by a human. Here's a video of a guy talking about Path of Exile's map generator: https://youtu.be/pM_5S55jUzk?t=1841. The clever approach taken here is that the human decides on some details and the program fills in the blanks. It's a simple idea, but I hadn't considered it before.

>>777 The freezing element is like kicking dead whales down the beach cock.

>>2006 While I also find the freezing obnoxious, I think my biggest issue with Cataclysm is that it puts little pressure on the player. Older roguelikes have hunger clocks and such to motivate the player to find trouble as a matter of survival. If you play it smart in Cataclysm, you get a bunch of canned food and sit in a basement without ever coming out. You essentially have to choose to find trouble for no good reason in order to enjoy it. I guess it's really a sort of, well, cataclysmic event simulator, though, so it's really just showing that, realistically, surviving under those conditions would probably be a pretty boring affair. I also have no idea what that lainon is talking about. Cataclysm games last far longer than 20 minutes, in my experience.

>>2002 Sorry for not reply earlier, I had a lot of work. >What would you say to the idea that roguelikes could be improved by doing away with the early game? Suppose a certain assortment of powerful equipment was randomly or semi-randomly selected for you at the start and you were immediately put up against powerful monsters. You don't need to give powerful equipment at the start you just need to give them interesting items to have some other strategies then bump your enemies to death. Also they should be identified or given 2 of each type (if the player won't know what that item does it takes away some early game tactics, as you will need to first identify the item (and that could lead to wasting the item)). Infra Arcana gives you at the start of the game metal spikes that can be used jam doors (making it possible to run from monsters), any if I remember correctly dynamite (very powerful but limited). On one side the early game in most roguelikes is interesting as each step can be your last on the other side it's boring as you don't have a lot of tactics to your disposal. So I would say give more items (consumables that have some interesting uses) at the start of the game would be the best solution but you have to watch out not to make the player to strong.