Komodo 13 comes out tomorrow

by ChessBase Shop
5/20/2019 – This program thinks like no other chess engine in the world. Inspired by the ideas of AlphaZero and Leela, the authors set out to implement a "Monte Carlo Tree Search". The result: Komodo 13 MCTS – an engine which looks for candidate moves in an incredible new way, and finds moves no other engine can. It enhances the traditional brute force engine that has reached well over 3400 points on the Elo scale.

ChessBase 15 - Mega package ChessBase 15 - Mega package

Find the right combination! ChessBase 15 program + new Mega Database 2019 with 7.6 million games and more than 70,000 master analyses. Plus ChessBase Magazine (DVD + magazine) and CB Premium membership for 1 year!

More...

The latest Komodo in the ChessBase Shop

Komodo 13 cover

Inspired by AlphaZero, Komodo developers GM Larry Kaufman and Mark Lefler have reinvented their engine from scratch over the last two years. The result speaks for itself:

Komodo 13 thinks like no other chess program

The new Komodo 13 MCTS (for "Monte Carlo Tree Search") searches for candidate moves in an incredibly innovative way and finds solutions most engines never see!

Komodo 13 analyzes better than all others

The development of the MCTS technology has paid off especially in combination with "multi-variation mode": In standard computers that have multiple CPUs, Komodo 13 MCTS is the world's best engine for analysing with two or more variations. In addition, Komodo 13 takes on a very human character. In balanced positions where other engines evaluate as "0.00", Komodo looks more closely: the engine gives preference to the side that has the easier game, for instance, when the opponent must find a whole series of difficult moves to remain equal.

Komodo 13 in the ChessBase Fritz UI

What grandmasters say

GM Boris Avrukh: "I am deeply moved by the style of Komodo. In my opinion it‘s the perfect combination between computer accuracy and human positional understanding. I get the feeling it‘s taken positional understanding to the next level. After such an impressive performance I am going to test Komodo in my future work, especially in very positional play, and am really looking forward to working with it."

GM Roman Dzindzichashvili: "I am extremely impressed by Komodo's play. I watched and analyzed every game and it was absolutely flawless positional chess, the likes of which has never been seen before by an engine... or human. Komodo clearly outplayed the other programs in the TCEC tournament."

Komodo mastermind GM Larry Kaufman: "We have just released Komodo 13.01. The normal version is about ten Elo stronger than Komodo 12.3, while the MCTS version is about thirty Elo stronger on one thread, and over forty Elo stronger on four or more threads at blitz time controls. While the MCTS version is not yet as strong as normal Komodo (unless you are using MultiPV), we believe it is among the top five CPU engines, and we claim that it is the number one CPU engine when MultiPV is set to six or more. Komodo 13 MCTS is much better than previous versions in its ability to utilize hardware with many cores. Elo gains in MCTS mode from version 12.1.1 are in the 250 to 350 range."


Box coverKomodo is a three-time computer world chess champion. Whether classical, rapid or blitz chess, in 2018 Komodo 12 won the Computer Chess World Championships in Stockholm, Sweden in all three disciplines. And Komodo 13 MCTS has significantly improved on multi-CPU systems!

Komodo 13: match winner and analysis partner

As with its predecessor, Komodo 13 comes with two versions of the engine. The standard version continues to be the first choice in the match play against other chess programs. The "MCTS" version is recommended for everyone interested in discovering variations and detecting mistakes together with Komodo 13.

Komodo 13: €99.90 (available from May 21, 2019)
Komodo 13 comes with a six-month pass for a Premium ChessBase Account.

System requirements: Minimum: Dual Core, 2 GB RAM, Windows 7 or 8.1, DirectX11, graphics card with 256 MB RAM, DVD-ROM drive, Windows Media Player 9 and Internet access.

Recommended: PC Intel i5 (Quadcore), 8 GB RAM, Windows 10, DirectX11, graphics card with 512 MB RAM or more, 100% DirectX10 compatible sound card, Windows Media Player 11, DVD-ROM drive and Internet access.


Note that you can only order Komodo Chess 13 tomorrow, May 21, 2019, when it will be available in the ChessBase Shop. The good news is that it will be at a special discount price.


About the Komodo authors

Don Dailey

The original author of Komodo was Don Dailey, who along with GM Larry Kaufman developed the program Komodo. Dailey started chess programming in the 1980s, and was the author and co-author of multiple commercial as well as academic chess programs. He died of an acute form of leukaemia in November 2013, but before his death handed over the code and all assets to Mark Lefler.

Mark Lefler

Mark Lefler is an American computer games and chess programmer. He is author of the computer chess program Now. Mark is a graduate in Engineering Science and Mechanics, and has been employed as a Security Engineering Officer by the US State Department. His non-computer interests include performing contact magic, music, and gaming.

Larry Kaufman

Lawrence C. ("Larry") Kaufman is a chess grandmaster, a title which he earned after winning the 2008 World Senior Championship. A longtime researcher in computer chess, Kaufman helped write the opening book for the pioneering chess program Mac Hack, co-developed Socrates II, worked on Rybka 3 and on many other research and commercial chess engines. He has written several books and articles, including The Evaluation of Material Imbalances. Larry has reached a considerable level of competence in several other board games. He is one of the strongest Shogi players in the West. He learned the game by the traditional method of studying handicap theory and is one of the leading experts in this field. Go and Xiangqi are among his other past interests.




The ChessBase Shop contains hundreds of products covering all aspects of chess for enthusiasts from beginner to grandmaster. Visit shop.chessbase.com
Discussion and Feedback Join the public discussion or submit your feedback to the editors


Discuss

Rules for reader comments

 
 

Not registered yet? Register

XSammaelx XSammaelx 5/30/2019 10:19
p.s.

I tested Stockfish versus Lc0 in a 'win material versus winning pawn ending' scenario, and Stockfish takes hours (on my CPU) to get that the latter might actually be better. Leela got it within a minute on weak (GPU) hardware. That's (a good example) diff where MCTS AI versus a traditional program might differ.

Can Komodo 13 do that? I'd be curious.
XSammaelx XSammaelx 5/30/2019 10:01
@AhmetBurak23

Just to be clear, MCTS gives an added dimension to analysis that even the best alpha-beta engines, like Stockfish, lack.

But can a person do better than forking over money for Komodo 13? Yeah, probably. Even with less than a stellar graphics card Lc0 can probably give better analysis given extended run times.

I was just trying to bust the ad hype.
Gulfstream321 Gulfstream321 5/29/2019 03:01
Fake news/ Stockfish is better than komodo
AhmetBurak23 AhmetBurak23 5/29/2019 08:03
"Analyzes better than all others." This means you don't know anything about Stockfish Development Versions.
SFDevs are CLEARLY BETTER than all of the engines except LC0 because LC0 learns. They are nearly equal in A Powerful Computer but in normal computers, SF is clearly better than LC0 too.
I mean Stockfish is the BEST!!!!!!
XSammaelx XSammaelx 5/24/2019 09:05
My first impression of this article is "ad hype".

"Analyzes better than all others", uh, yeah. I guess could be somewhat justified as saying "Don't have a good graphics card? We'll get you MCTS functionality in a more generic way." But does it measure up to Lc0 playing strength? I guess we'll see when CCRL 40/40 ratings come out down the road.

"Engine gives preference to the side that has the easier game, for instance, when the opponent must find a whole series of difficult moves to remain equal." sounds intriguing, but the devil must certainly be in the details. This sounds derivative of the whole Monte Carlo process, and would apply to engines like Lc0 also.

Color me skeptical about this article's claims, but I'll certainly eat crow if that's in order.

It might be kinda sad but the distributed testing model of Stockfish outdid proprietary chess engines in tuning the evaluation, and now self-play neural networks seem to be making the chess programmers obsolete altogether. Oh well, I did fork over a lot of $$$ to you proprietary guys over the decades.
1