24

If you're trying to make the strongest engine possible, absolutely go for NN engines. Traditional engines are great - Stockfish is arguably still the strongest engine on the planet on consensus equal hardware - but they are hard to write. These engines didn't get where they were overnight; they took years and years of work. Stockfish for example has been ...


23

I'm not an expert on quantum computation but my understanding is that quantum computers are not expected to be useful for chess. Quantum algorithms are very good at finding needles in haystacks: the three big quantum algorithms are Shor's factorization algorithm, Grover's database lookup algorithm and the Deutsch–Jozsa algorithm, which essentially ...


23

Computers allowed the creation of endgame tables, which allow the user to know with 100% certainty if a position can be won, and how to do it. Currently all positions containing 7 or fewer pieces are 100% known. I am unaware of any opening line busted by computers. Not that I know of, but computers do find answers to some hard questions. So while computers ...


23

Is the tournament leading to some consensus on how good/bad Carlsen's decision was? Not quite. As others have pointed out, Carlsen's decision was based on factors outside that one game. With a stronger position and a large time advantage, Carlsen most likely could have won game 12, but Caruana had just tied Carlsen in 11 consecutive games, several of which ...


20

That's why I wonder whether there have been any attempts made to provide comparable hardware to both. This is Google you're talking about! So the answer is obviously "No". From the original paper hardware used for initialising and training - Training proceeded for 700,000 steps (mini-batches of size 4,096) starting from randomly initialised ...


20

Is the tournament leading to some consensus on how good/bad Carlsen's decision was? No. Consensus on Carlsen's play and decision has already been reached, I would suggest. Psychologically Carlsen made it clear in the post match interview that his goal before this game was a draw to reach the rapid playoff where he thought (correctly) that he was strong ...


17

This depends on a lot of things, but most importantly: your playing level and the topic. For example, if it is an opening book, it heavily depends on the opening. Some opening variations which are main stream today, didn't even exist in the seventies, or were rarely used (like the Berlin Wall in Ruy Lopez). Others haven't seen that much development (QGD, ...


16

Anand is absolutely right. Computers have revolutionized the way we play chess, think about chess and prepare. Tony Ennis mentions tablebases. That is just the tip of the iceberg. Computers have infiltrated almost every aspect of the game. The only place they're banned is when we sit down to play over the board (and perhaps on some online platforms). At ...


14

IBM claimed the machine could search for 200 million moves per second, while Stockfish in the recent AlphaZero match could "only" search for 80 million per seconds on a modern multi-core machines. But... it was unclear how exactly IBM derived the number. There's no universe definition on how an engine calculates number of moves per second. How it's done is ...


13

In the beginning years of computer chess, people have actually tried to teach computers chess in the same way as they do with humans, explaining strategic concepts like a healthy pawn structure or the initiative. These attempt were soon abandoned because the method you describe was much more successful. Recently, there has been another attempt to let an ...


12

http://tests.stockfishchess.org/tests has information on exactly what you want. They are self-playing tests for improving the Stockfish code base. The draw rate is roughly about 60% - 70%. There're lots of wins and losses. I believe white has more wins but it's not shown on the page. Chess engines aren't perfect, they do make mistakes. If SF makes a bad ...


12

I think it's best if I elaborate on your second point with an example move in the game 1 between AlphaZero and Stockfish which also served to satisfy my curiosity today. the time limit of 1 min/move (How would this disadvantage Stockfish?) Stockfish's performance is dependent upon both the time limit and the hardware configuration, so just think of when ...


12

Otto Blathy is famous for his long mate problems. There are different records depending on whether you allow an illegal starting position, or promoted pieces, or minor duals in the solution, so your answer might be anywhere between 257 and 292 moves. This answer on Puzzling.SE shows you a construction in 270 moves: [Title "Otto Blathy - White to play ...


11

Longest known checkmate is in 549 moves, revealed by Lomonosov Tablebases. You can see the game here. Plus, you don't need to say "endgame" explicitly. Longest checkmates come from endgame tablebases already. And it's not possible to build a middlegame tablebase practically, because of the exponential space complexity. Let's say there is a middlegame ...


10

For what it's worth, here are the rates of decisive games of the TCEC seasons 4 'til 8 superfinals. Season 4 is a bit of an outlier (maybe a rule change?), but after that the percentage of decisive games drops consistently. Season 4: 10/48 = 20% Season 5: 18/48 = 37% Season 6: 19/64 = 30% Season 7: 11/64 = 17% Season 8: 11/100 = 11% Arguably at an ...


10

It will not show how good or bad Carlsen's decision was - humans play and assess differently from computers; there are positions that are extremely easy for computers, but terribly difficult for humans and vice versa. So a computer evaluation of ~+1 might not give any chance for a human win and in many cases a position with equal computer evaluation is an ...


9

So, in terms of creating the strongest chess engine possible, should I go neural network or hard-coded? Don't choose a NN unless you have access to ridiculous(A few hundred Nvidia V100s). Training a NN to play chess takes so much hardware. See the people contributing to Lc0 to train over 200 million games. Since you will probably have trouble accessing the ...


8

Maybe you can take a look at TalkChess, a forum dedicated to computer chess. I found a recent thread that might be interesting for you: Progress in 30 years by four intervals of 7-8 years A couple of matches between (former) top engines are played on the same hardware. The test suggests that in the recent years (2002-2017), the gain is mainly made by ...


8

First step: Define your goals/reasons I think this is the predominant factor. Which of these best fits you? (Choose only one) You want to enjoy a fun, challenging coding task You want to create an extremely good chess engine You want to learn about how chess engines work You want to learn/practice coding skills You want to learn/implement computer science ...


7

Well, Aaron Nimzowitsch wrote a book called "My System" in the late 1920's, and it's still worth reading today, so just because a book is old doesn't mean it's bad. It's hard to say whether a book has stood the test of time just knowing what year it was published, but one thing is for sure: many lines can be faulty, as computers were not used to verify ...


7

I don't think it'll ever happen because: Who's going to pay? Deep Blue was covered by IBM, Deep Fritz was covered by Chessbase, but Stockfish is a free open-source engine. It's not fun for Carlsen to play an odd match with Stockfish, but he will likely be beaten convincingly if the match is fair. Human vs computer in chess is not interesting anymore. The ...


7

Let's take a look what they are Polyglot is an open source format. It is free to use, free to distribute, the file format is in public domain. You won't land yourself into any potential legal disputes with Polyglot, as long as you follow the GPL license for the source code. CTG file format is owned by Chessbase. It's a closed format, and is not open for you ...


7

What you have to understand is, first, that any rating is just a comparison of players within that pool. So a 2200 on chess.com is not the same as 2200 on lichess.com, or on ICC. They are probably close, but it is not the same. Second, you need to realize that your three game sample is just not enough to draw any valid conclusions. You are human, so even if ...


6

You are somewhat correct. In theory, each position can be correctly categorized in one of three ways -- win for the side to move, loss for the side to move, draw, all presuming "best play" by both sides from that position. That last caveat is required to make the ideas of "will win", "will lose", and "will draw" well-defined. But that caveat also means that ...


6

This article from Chessbase answers most of your questions. This was the position in which Kasparov had set a "computer chess" trap. [fen "r1r1q1k1/6p1/p2b1p1p/1p1PpP2/PPp5/2P4P/R1B2QP1/R5K1 w - - 0 35"] [White "Deep Blue"] [Black "Gary Kasparov"] He was expecting 36. Qb6 Rd8 37. ab Rab8 38. Qxa6 e4 [fen "1r1rq1k1/6p1/Q2b1p1p/1P1P1P2/1Pp1p3/2P4P/...


6

In contrast to Rauan Sagit, I'm going to offer some reasons I think any human should (assuming they don't get lucky and play some computer-level games themselves) lose such a match. Elo differential: As of Nov.15, 2014, the highest-ranked engine on CCRL 40/40 is Komodo with an Elo of 3303 on relatively modest hardware of an Athlon 64 X2 4600+ 2.4 GHz (...


6

This question is based on a number of misconceptions. One, engines are not self-learning software, they are not "trained" on a set of games. They just go through a huge number of possible variations and evaluate the end positions based on fixed factors set by the human programmers. Second, engines can be configured to use opening databases, but they hardly ...


6

Stockfish, Houdini and Komodo (many other engines too) can all be configured like what you said. This is known as Contempt. In Komodo, the name is Drawscore (same thing). If you give enough negative value to the option, the engine will not push for a win even it sees it. http://support.stockfishchess.org/kb/advanced-topics/engine-parameters


6

I've noticed AlphaZero played 1.d4 and 1.Nf3 in http://www.chessgames.com/perl/chess.pl?tid=91944&crosstable=1. Possibilities: Mutlithreading (any software engineer can tell you it's not deterministic) Monte Carlo Tree Search (the algorithm draws random paths, so it can't be deterministic) Please note while the model's parameters are "fixed" (your own ...


6

You're likely looking at premoves; the players know (or at least have a good guess at) what their opponent's move is going to be, so they can already 'ready' their own move by taking the piece and dropping or just moving it to the destination square. Of course, if the opponent makes another move, they have to cancel their move (if it's still possible; that ...


Only top voted, non community-wiki answers of a minimum length are eligible