It's no secret that I am a huge fan of roguelike games. For those who don't know, and who are too lazy to read Wikipedia, these are games that feature procedural, randomized content; statistical-based character advancement; deeply complex item systems, often requiring the player to try to identify items before using them; player-like AI; complex, often unexpected world interactions; death systems that severely punish; and a brutal attitude towards unskilled play. Diablo is probably the best known roguelike in the West, although it's also one of the shallowest. If you have never played one, NetHack is the canonical place to start (although I don't like it much anymore), and Ancient Domains of Mystery is one my favorite free ones.
What may be a secret, and what certainly caught me off-guard, is the extent to which the genre has revived within the past year. If you had asked me at the start of 2007 what my favorite roguelikes were, I would've mentioned NetHack and ADOM, Rogue itself, and Shiren the Wanderer for the Super Famicom. These are also the only ones I invested more than a few hours in anyway; if you were to ask other people you might hear Omega and Angband, and a dozen variants on all of these, but nothing particularly new.
But in the past year we've seen Pokemon Mystery Dungeon (millions sold in the North America! A roguelike with sales in the millions!), Izuna: Legend of the Unemployed Ninja, Rogue Hearts Dungeon, much-enhanced remakes of Shiren and Baroque, Chocobo's Mystery Dungeon for Wii with a North American localization, Shiren 3 in Japan, and sequels to Pokemon Mystery Dungeon and Izuna. We've also seen games like S.T.A.L.K.E.R. (player-like AI and procedural weather), Hellgate: London (random levels and items), Spore (animation and creatures), and Dwarf Fortress (huge procedural worlds) that are not, strictly speaking, roguelikes, but borrow design lessons and elements from them.
Seriously, what happened? Roguelikes have never been totally dead, but they've always been a very niche genre. Outside of Chunsoft, there aren't many commercial ones, and those don't usually meet with much success. So why the sudden explosion?
First, to be blunt, Pokemon. Getting the Pokemon license introduced roguelikes to a whole new generation, just like the Dragon Quest license did over a decade ago. While PMD isn't a great roguelike it's a good one, made by developers that love the genre and are exploring it trying to find a way to make it more accessible to younger and less involved gamers. In general though, bad games don't kind the kind of critical reception PMD did, and so there's something else going on too.
As games get more expensive to make, procedural and emergent content look much more attractive. It's easier to design (though maybe harder to balance) 100 interacting items - giving, at first blush, 2100 different combations - than to design 10,000 items, which is nowhere near close to the depth provided by the 100. Maybe most of those combinations are boring, but then so are most of the 10,000.
As graphics technology gets better, procedural content also looks much more attractive. NetHack won't wow anyone with its display, but Baroque and Shiren 3 look impressive despite the fact they're fundamentally laid out on a grid and assembled automatically. Pouring a modern graphics budget into a roguelike could result in something that looks spectacular.
Dwarf Fortress does very interesting things with modern physics-based gameplay, something that many roguelikes have ignored. Physics is deterministic, but in sufficiently complex systems becomes chaotic - which is exactly what roguelikes thrive on.
Roguelikes also provide a solution to the perceived problem of cheating by information sharing, regardless of whether it's a single, multiplayer, or massively multiplayer game. Raph Koster of course totally misses the point when he discusses static information - even if you have all roguelike data (like a friend at work who has a giant table of Fei's Final Problem item costs for ID purposes), the game still provides an insane amount of unique and interesting situations, that require real strategy, each run. In this way you extend the life of the game, not by passing around stupid information tokens, but by making so many potential data that the player has to play the system and not the walkthrough.
There's huge unexplored swaths of design space here. Where are the roguelike Castleroids or puzzle/exploration-based games (Dungeon Hack had procedural key/door puzzles, which I don't think I've seen since)? Where is capture-the-flag on a procedural map? Where are the RPGs with shifting economies and genetically-bred enemies?
Well, they're coming. Not this year, but soon; and when they do, three decades of research in procedural content and exploration-based mechanics that have been all-but-ignored by mainstream games is going to explode.