The New Computer Chess World Champion | Hacker News Reader

The New Computer Chess World Champion

rjlipton.wordpress.com

78 pointsbdr22 comments

I think the real story is how close and competitive the open source Stockfish engine is.

That was certainly the storry earlier this year, when stockfish won. Now I wonder what kinds of improvements Komodo has made to get here. Stockfish is certainly improving all the time through the fish test project. (and you can contribute too!)

Better SMP scalability mostly. Stockfish still has a definite advantage over Komodo on fewer CPU cores, and shorter time controls. But it scales poorly above 8 cores, while Komodo scales almost linearly (the TCEC machine had 16 cores).

I wonder what the game depth would be for significantly more complex games like Starcraft 2 or League of Legends. Given the way those games track statistics we should be able to get a definitive answer.

Do you recon we can figure it out ourselves? What is the depth of each starcraft league?

The game depth for those games is going to be enormously large. Chess is played with 32 figures on a 64-tile board, and that's enough to give us serious problems. Games like SC or LoL have way, way more options -- the set of all possible game states is of such a size that I don't even have a sense of scale for this order of magnitude.

Think about it this way: the average branching factor of a legal chess move is 35: there are about 35 legal moves that could follow. For Go, the average branching factor is 250. In StarCraft, how many "legal moves" do you have available at a given game state, on average? Tens of thousands? Millions?

By the way, this is made significantly more difficult by SC and LoL not being turn-based games the way Chess is.

Maybe. There are indeed many possible moves, but how many represent important strategic choices that take learning to understand, and how many are just the obvious thing that even beginners will see how to do after a little study? (there are plenty of "broken" RTS games where there are millions of choices but there's a clear winning strategy in building lots of one particular unit as fast as possible - StarCraft is the exception and not the rule in sustaining interesting gameplay years after its release).

And if we're talking about computer playability, the real-time nature of these games cuts both ways; in StarCraft analysis we literally talk about how many "actions per second" a player was able to execute, a field where a computer will naturally have an enormous advantage (part of why I prefer to play Supreme Commander).

For a second, I was disappointed that chessgames.com didn't have any of the TCEC games to step through. But the article links to the TCEC site, and it has its own interface with much more information about what the engines are "thinking": http://tcec.chessdom.com/archive.php

So the article is by Ken Regan? I'm pretty sure he's the best player I've ever faced, in a casual game when he was checking out grad schools (I think) and dropped by the Harvard math department.

We castled on opposite sides. Not the best kind of game in which to have a chance against a vastly superior player. His attack, quite predictably, punched through faster than mine did.

The WC match was held in November between Anand and Carlsen. Fake title!

This is a totally click-baity title. Obviously it should read "The New Computer Chess World Champion", as it has nothing to do with a New Chess World Champion, though the title is obviously meant to get your attention by implying that it does.

(edit: accidentally deleted this comment)

If it's click-bait, it's poorly done click-bait, because the actual content is more interesting than either of those titles.

Well since Komodo will kick the butt of any human it played is it not the real world champion?

Only in the sense that a team of 11 Ford pickups would be the NFL champion.

Ford pickups can't play football.

A better example would be a motorcycle vs. Lance Armstrong.

Not really. A motorcycle can't ride a bicycle any more than a Ford pickup can play football.

Computer programs can play chess better than humans, following the same rules that humans do.

Humans usually cannot take books or a printed compilation of all the chess games in history, while computers have a database with a lot of chess games. If humans could take a whole library with them to the match, things would be different, but then, the reading/writing speed of a computer is much higher then the human reading/writing speed, so I think the comparison between human muscles and a gasoline engine is fair. Kasparov playing with no time limit and his whole library at hand would still be a match for any computer.

There's no chess rule stating that a human cannot memorize large opening tables.

I'm sorry you've been voted down, because yours is a common misconception. No human player can beat the modern chess engines even running on a smartphone without an opening book.

It's incredible how quickly game engines have advanced. Way back in 2007 Rybka destroyed all human challengers without an opening book. The current chess engines are much stronger and running on much faster hardware. There's simply no comparison any more.

Or Lance Armstrong vs an unassisted human.

We added "computer" to the title.

Reply on news.ycombinator.com