AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |
Back to Blog
Logicbots wiki4/20/2023 ![]() On the online servers, qw plays with an extra added delay so that it doesn't use too much server CPU. Historically its best 3-rune winrate was 15% DDFi^Makhleb, and, for 15 runes, about 1% with GrFi^TSO before the 0.28 hell rework.īranch order: D -> Lair:8 -> Orc:3 -> D:15 -> S:5 -> Vaults:4 -> Depths:5 -> S:5 -> Vaults:5 -> Zot ![]() See qw for a summary of results.Īs of 0.29-a, qw has a 0.36% winrate with GrBe with 1 win in 276 attempts. The first DCSS bot to ever achieve an uninterrupted and unassisted win (see '!lg qw won 2'). Qw: A fully automated lua bot written by elliptic, with some code borrowed from parabolic and xw. Unlike the approach here, it got to benefit from human expertise, and also it can only win with a very narrow set of characters. The bots from the NetHack Challenge also can't participate as they need a custom NetHack binary that outputs the game data in a machine parsable way whereas the tournament requires you to play on existing public servers.ĭCSS has been beaten by a handcrafted Lua bot, written by the (at the time) indisputably best player in the game. ![]() So a generic bot that is trained to win the game isn't best at getting those.Īnd for today, I think there is no bot that can win the game for the newer versions or any forks. ![]() That's mostly due to the tournament including lots of NetHack forks and the clan trophies being special side achievements that aren't necessarily needed for winning the game. The only time a bot clan was participating AFAIK was 2015.Īnd it didn't do badly (it did score 2 of the 5 clan trophies) but also not particularly well. This year, I don't think there are any obvious bots running. Right now, Junethack, the NetHack Cross-Variant Summer Tournament (aka the tournament that is abusing players for finding bugs in nethack forks tournament) is running. (Which is something that symbolic approaches also don't come anywhere close to doing, because they just cheat by the capability being given to them by hand-engineering rather than having to autonomously read, understand, and apply.) There are increasing experiments in making DRL agents exploit or initialize from pretrained language models ( come to mind) or reading manuals ( all the way back in 2012!), and of course, a DRL agent can learn a tremendous amount without actually doing any playing by offline and off-policy and imitation learning, but while it is exciting and things like Gato look like the future, there is a long way to go from feeding in a dump of the Nethack wiki which mentions offhandedly "you can do X" to an agent recognizing an opportunity for X in the wild and executing it. Then you have the extensive hand-engineering of expert knowledge which goes into ascension agents and the symbolic winners, above and beyond merely plugging in a Sokoban solver. Nethack literally has levels which are just Sokoban and which are important to solve GOFAI can push the boulders around perfectly, even though it would be unable to recognize a photograph of 'a boulder' or use the word 'boulder' in a story. Moravec's paradox again: humans find perception easy but things like Sokoban hard, while GOFAI approaches find Sokoban so trivial that it's common to use it (or Sudoku) as a toy problem introduction to constraint solving. (The real world is not made of a clean little grid of discrete high-level objects like 'dragon' or 'black jelly'.) Because Nethack and video games remove most of the reasons that GOFAI failed for example, everything hard about perception from pixels is removed when you have a little parsed text grid of cells presented to you or are hooked into the game engine and query objects directly. ![]()
0 Comments
Read More
Leave a Reply. |