A couple of weeks ago,
xalolo linked me to Legerdemain, a free-as-in-beer fantasy game in the space where Roguelike games and what one might call "traditional" CFRPGs [I'm really feeling the Ultima here] intersect. The writing is actually rather compelling, the gameplay is in some ways excellent and really interesting, and in some ways frustrating in ways that feel like poor design to me.
We were chatting about game design elements and what we liked about it and what frustrated us, and I noticed that I'd started saying things like, "when I write my epic roguelike, I'll do this differently". I've started putting a fair amount of thought into actually attempting to write a beast like this.
Borrowed from RLs is the food clock, which does a nice job of driving you forward at least when you're in dungeons, softened by a reasonable availability of food in towns which is probably good for a game of this magnitude; you can definitely pause to rest if needed. I also found the food clock an excellent motivator in the opening sequence, where I definitely felt driven to progress rather than exploring every nook and cranny.
RL "permadeath" is modified here to the point that it's really more of an 80s-style "save at inn". Ok, that's exactly what it is. When you save in an inn, a backup of your save game is also made which is restored when you die. So no permadeath if you've made it to the first inn (this took me three tries). However, in an excellent bit of design, you can also save anywhere, functioning like a roguelike save that's deleted when you resume. You'll never have to keep playing to get to the next save point like some poorly-designed games I could mention.
It also implements the roguelike identification puzzle. In an interesting twist, many items' descriptions and names give clues to their function. Identification is also fairly readily available, which is probably good, as the game is ultimately more driven by plot than these elements. Especially as item descriptions are the same for all players and playings, so the ID puzzle is only interesting the first time you play. Which also makes it scummable with town saves if you really wanted to. Which makes it good that ID is available enough that there isn't really temptation to scum it. Basically, it's interesting while you're out exploring (where the game feels more like an RL), but not too tedious while in town (where the game feels more like a more "standard" CFRPG), which is a good balance. I think I'll borrow it for my game, but bring back the randomized descriptions so it's still interesting when you're exploring the second time you play.
The game does some really interesting things with status effects. They're numerous, they're [mostly] annoying but not fatal if you respond to them (that is, they're interesting), and they're [mostly] preventable fairly early on in the game if you're willing to allocate your limited resources (equipment slots, inventory slots, encumbrance, money, spell slots) appropriately. This is all great design.
You might have noticed those "mostlies" above. That's because of the hallucination status effect, and the fact that it occurs in traps and ranged monster attacks from pretty early in the game, and not exclusively in areas where you can learn to avoid the area or tread carefully with active divinations to protect yourself. I've not yet found an item which protects against it. In a rather clever way, it's frequently lethal, but you get to walk around for a while hoping it will wear off harmlessly before you die. If it decides to kill you, there's not really anything you can do about it, although there are steps you can take to improve your odds. Even though this isn't technically an instadeath, it feels enough like one to me that it falls into that category where I agree with the Crawl developers: instadeath isn't fun.
[Aside: Crawl (Dungeon Crawl Stone Soup) is a really excellent roguelike. I'm totally converted from the cult of Nethack, which kind of makes me sad, because many of the things I used to love about Nethack I'm now recognizing as flaws. But Crawl is a great game and I recommend it highly.]
Another thing that isn't fun which Legerdemain decided to implement is breeders. That is, monsters which without restriction create full-strength duplicates of themselves, fairly frequently. In an anti-farming move, they don't provide XP. Unfortunately, they're perfectly happy to flood a section of dungeon while you're exploring the area nearby [as the crow flies, in a twisty maze where you actually won't get there for a long time] or resting similarly nearby [you frequently want to rest in dungeons, which is fine, but breeders sometimes turn that into an annoying surprise]. Due to the joys of exponential growth, once you've got a good colony of these things going they tend to come back as quickly as you can kill them off. My game will probably lack breeders (I find them thematically silly unless they're summoners), or will put some sort of limits on their breeding ability so they don't get out of control. I may make an exception to allow "summoning swarms" as a late-game hazard, but this will constrain population explosion to a situation where you're expected to have developed solutions to that sort of problem and where you've put yourself in the dangerous situation (confronted the summoning enemy, not just resting where it's geographically near you).
It's really little things that bother me. Like the above. Or the one item per map tile restriction (doesn't create interesting choices, is just annoying, unlike inventory slots). Or the fact that monsters don't have inventories, just random drops. Or that monsters without inventories can pick up items or steal your money (so the money that bandit just stole has vanished into the aether). Or the way level contents are forgotten the instant you leave the level. [Maybe we don't want to record that detailed game state everywhere, but a compromise would mean going up and down a staircase wouldn't instantaneously reset all the monsters, items, and traps]
To wrap this up for now, although I may have more to say about the game later, the user interface is riddled with frustrations. Again, I'm spoiled by Crawl, which is so user-friendly that it includes an auto-explore function [surprisingly, no, it doesn't ruin the fun, just removes some tedium]. I miss having a far-travel (and you do a lot of travelling in Legerdemain). I long for items identical to a consumable you've identified to also be identified, so I don't have to write this stuff down. The remembered conversations and key paragraphs are awesome, and the Halls of Memory are a nice touch, but why couldn't the latter provide a link to the former? Far look [Survey] is a UI feature because the viewport is small, not a character action; don't turn it off due to status effects. If you're going to have long-lasting effects that I'm likely to want to rest out [Strength drain, I'm looking at you], make sure they count so I can use rest-until-healed for them. If you want status effects to prevent resting (in the recovering stamina and health sense) and searching (because it's distracting), make sure I still have access to a command to wait it out. A command that gets interrupted if a monster comes into view; holding down the wait key or pressing it 300 times are both unacceptable. Nethack taught us that effects that make you forget the things you've IDed or forget map segments aren't interesting, they're just annoying and encourage people to "cheat", don't use them.
no subject
OMG, try Crawl! If only for its totally awesome and customizable UI which generally does the right thing out of the box and doesn't let you do stupid things like killing yourself in lava because of a finger slip.
Right, flaws in Nethack...
A feature of Roguelikes is that the true game is sort of a meta-game. That is, you send your disposable characters into the highly random dungeon, knowing they're likely to die, but trusting that from their failure you as a player will learn more about the game world and its interactions, or develop better tactics to deal with situations, or the like. That is, even when you lose, as you frequently do, you're getting better at the game so that over time you'll live longer and do better.
The problem with Nethack is what's usually referred to as the necessity to be thoroughly-spoiled on the game to win it. Which isn't strictly true, but . . . there are a number of effects available in the game which are a significant aid to victory which result from using resources in highly unintuitive ways to produce what turns out to be a highly desirable result. A number of them are rather hidden in a murky UI and referred to under the "the Devteam thinks of everything" moniker. The route to becoming really good at Nethack involves a pretty significant amount of "I wonder what happens if..." I do something really weird, or consulting spoiler documents. The Oracle provides hints to some of this eventually, but not all of it. Good at Nethack is much more about "I know the game inside out and can apply that knowledge to make the game pretty easy" than about "I'm good at solving the tactical problems with limited resources this kind of game presents".
Another problem with Nethack is that there's a really long segment of not particularly challenging game in the late game, where almost no threats are relevant but there's an awful lot of game to get through. To the point where people have developed strategies for making the late game faster not to make it safer but to make it less boring. Player boredom is said to be the #1 killer of late-game characters. If you've ever gone through a segment of a Squaresoft game holding down 'A' until the fight's over and wishing the stupid non-threats would stop attacking you so you can get on with the game, you know what I'm talking about.
I think it's mostly the "game of Lore" aspect, which I used to love, but now realize is somewhere between poorly implemented and poor design. Possibly the single most important interaction in the game is one you're not particularly likely to try for yourself and is incompletely (and I'm not sure helpfully) hinted at by exactly one entry in the true rumors [includes minor Oracle consultations] file (I just looked).
Also, as mentioned in the Guidebook, spamming the E-word is totally broken.