The Komodo Files – working with a chess engine
3/14/2015 – After an unsuccessful appearance at the London Chess Classic GM Danny Gormally decided that his chess was stuck in a rut. Perhaps it was time to jump on the gravy train and do what all the top players seem to be doing these days: work with an engine. He started his experiment with Komodo, one of the strongest engines around. Part one of his report in CHESS Magazine.
Last year Komodo won the world championship title on two occasions and can call itself "2019 World Computer Chess Champion" and "2019 World Chess Software Champion". And the current Komodo 14 has been clearly improved over its predecessor!
 |
The Komodo Files |
Danny Gormally decided it was finally time to work properly with a chess engine
The London Chess Classic did not go well for me. If you include the Super Rapidplay before the FIDE Open, I drew and lost a number of games I should have been strolling on paper. Clearly my chess was stuck in a rut. Probably had been for years to tell the truth. Perhaps it was time to jump on the gravy train and do what all the top players seem to be doing these days: work with an engine.
I asked Chess & Bridge if they could send me a copy of Komodo, which was rumoured to be one of the strongest engines around, if not the strongest. I remember this Icelandic GM telling me about it a couple of years ago at the Scottish Championships and had vowed to get my hands on it ever since. The DVD duly arrived through the post a day later. In trembling anticipation I loaded it up on my laptop.
It didn’t take long before I got an insight into Komodo’s fearsome strength. In a number of training games I was brutally savaged. Press "Ctrl +" to enlarge the comments in the following JavaScript replay board (and Ctrl-0 to switch back to normal).
[Event "Training Game, Alnwick"] [Site "?"] [Date "2014.??.??"] [Round "?"] [White "Gormally, D."] [Black "COMP Komodo"] [Result "0-1"] [ECO "C88"] [Annotator "Danny Gormally"] [PlyCount "36"] [EventDate "2014.??.??"] [SourceDate "2015.03.01"] 1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Ba4 Nf6 5. O-O Be7 6. Re1 b5 7. Bb3 O-O 8. a4 Bb7 9. d3 d6 10. Nbd2 Nd7 11. c3 Nc5 12. Bc2 Nxa4 13. Bxa4 bxa4 14. Qxa4 { Komodo is happy to play a line where it bags the two bishops, even at the cost of a slight weakening of its pawn structure. This displays its increased positional awareness which is one of the things that the developer of Komodo, Larry Kaufman, wanted to focus on. He said that he was even willing to sacrifice some calculating strength in order to maximise the program's strategic understanding.} Qe8 $1 {A typical computer move, in the sense that it looks ugly but somehow just works. I think it was Peter Leko who said that working with computers changes the way you look at chess, because you have to analyse moves that in the past would have been dismissed as too ugly. Now whether the move is aesthetic or not hardly matters, the point is whether it works or not. Black freezes the development of his rook on f8, but has ideas of perhaps playing ...f5, and bringing the queen out to g6 or h5, starting an attack. And in a concrete way the black queen is shadowing my own queen on a4, which becomes relevant very quickly.} 15. d4 $2 {The mistake comes quickly. Normally when you make an error, the computer flashes up a message along the lines of 'Uncle Komodo doesn't think much of your last move buddy' with an option to take back. Then I'll input an alternative which will be equally bad, meeting with a similar reply of 'Komodo thinks that your previous move wasn't the best in the position, Danny' with the effect of further demoralising the patzer user.} exd4 16. cxd4 d5 $3 {I'm not sure this quite deserves two exclamation marks, but it did make a big impression on me at the time. Many carbon-based life forms might subconsciously overlook this idea altogether as it just seems too dangerous for Black to open up the e-file with the rook on e1 and the queen on e8. However, this just works and as soon as you see this move it makes total sense. Now I can't avoid the opening of the position when the two bishops will start to rule supreme.} 17. e5 ({After} 17. exd5 Nb4 {, Black will quickly recover the pawn with a dream position to boot.}) 17... Nb4 18. Qb3 Qb5 {. Black is starting to take control. I can't recall the rest of the game, but I don't think it took long before my position collapsed altogether. This was a recurring theme. Even with White, it was difficult to get past move twenty without a lost position. You'd play what you think was a reasonable game, believe that the position was fairly unclear, then switch on the evaluation and see that the computer was showing '+1.5' in its favour and with that rising all the time. --- I was thinking surely I'm a grandmaster, I should be able to put up some resistance? But all I could think was that Komodo was extremely good at exposing just how poor my chess had become. Of course, computers are scarily strong now. Even the best players in the world would struggle to play against them, but I was getting outplayed on almost every move. It was a daunting experience and hardly the confidence booster I needed before my next tournament down in sunny Hastings. However, I was determined in any case to take Komodo along and use it for preparation.} 0-1
In my second game in Hastings I was paired against John Anderson. I noticed his theoretical knowledge was fairly good and that he was playing the Slav pretty much every time against 1 d4. To prepare I played a number of training games against Komodo in the Geller variation of the Slav. I think that’s one of the advantages of using an engine. Think how tedious it must have been to prepare for games 40 years ago. You’d have had to lug around these huge great Informators and spend hours flicking through them and other opening manuals.
With the engine it’s just much quicker; you learn how to play an opening exponentially faster. I can charge up Komodo and just play some games with it. So I was playing the Anderson side, the black side of the Slav, and Komodo would take White, just so I could get a feel for how the computer handles the position. Once I felt I’d mastered one sub-variation, I’d move on to the next. Then the plan was to try and replicate Komodo’s play over the board. However, the reader may have already spotted a flaw in this plan...
[Event "Hastings"] [Site "?"] [Date "2014.??.??"] [Round "?"] [White "Gormally, D."] [Black "Anderson, John"] [Result "0-1"] [ECO "D15"] [Annotator "Danny Gormally"] [PlyCount "80"] [EventDate "2014.??.??"] [SourceDate "2015.03.01"] 1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 dxc4 5. e4 {I chose this admittedly risky gambit because I felt I had more chance of blowing him away in the opening with something like this. If you use engines a lot in preparation, then it makes sense to play sharply when you can use the computer's great power in calculation. That's why Carlsen is famous for avoiding opening variations that are extremely sharp and where there's a danger he can lose in preparation. He doesn't want to lose to an inferior player who uses computer analysis against him, but rather wants to outplay them with his special understanding.} b5 6. e5 Nd5 7. a4 e6 8. Ng5 $5 {An interesting idea that has been championed by the Ukrainian player Eljanov, amongst others.} (8. axb5 Nxc3 9. bxc3 cxb5 10. Ng5 Bb7 11. Qh5 g6 12. Qg4 Bd5 {is the main line. However, White is not better here and many games have shown that Black has quite serious chances himself by playing ...a5 and ...b4 later on.}) 8... h6 $1 ({In my opinion,} 8... Be7 9. h4 $1 {is quite dangerous for Black and} h6 10. Nge4 b4 11. Nb1 Ba6 12. Nbd2 c3 13. Nc4 Bxc4 14. Bxc4 {would have been similar to the line I was hoping to get in the game.}) 9. Nge4 b4 10. Nb1 Ba6 11. Nbd2 Nf4 $1 {This is the problem and makes the whole line somewhat dubious for White. In fact this move has been known about for years. The question was could Komodo find any improvements for White?} (11... c3 12. Nc4 Bxc4 13. Bxc4 cxb2 14. Bxb2 {was the main focus of my preparation. White plays Qg4, castles kingside and eventually puts a rook on d1 with obvious chances of an eventual attack against the black king. By contrast it's surprisingly difficult for Black to free himself without falling apart completely:} Be7 15. Qg4 O-O 16. O-O Nd7 17. a5 {with excellent compensation for the pawn.}) 12. Qg4 Nd3+ 13. Bxd3 cxd3 14. O-O Qd5 15. Re1 Nd7 16. Nf3 {Up to this point this was all preparation.} c5 $1 {Anderson continues to find the best moves. What irritated me at this point was that 16...c5 was clearly the most natural move to play over the board, but in my preparation I had failed to really pay much attention to it.} ({I think the computer was preferring to play some weird stuff like} 16... Bc4 {, but for a human it makes much more sense to open up the position for the two bishops.}) 17. dxc5 $6 {This was where I finally deviated from the engine. I couldn't recall at the board what it had been suggesting.} (17. Nd6+ {was what it wanted to play when I checked the game with it later.} Bxd6 18. exd6 c4 {was what I didn't like at the board as I thought these pawns were starting to look very scary. Of course, though, the computer doesn't fear ghosts and thinks White is doing very well after} 19. Qxg7 O-O-O 20. Qxf7 c3 21. bxc3 bxc3 22. Qxe6 Qxe6 23. Rxe6 d2 24. Nxd2 cxd2 25. Bxd2 {, as he already has a number of pawns for the piece. Over the board, this all seemed too scary, but the problem was the alternative was even worse.}) 17... Nxc5 {Now the knight is targeting b3 and I'm probably just in trouble.} 18. Nd6+ Bxd6 19. exd6 Nb3 20. Bf4 Nxa1 21. Qxg7 O-O-O 22. Rxa1 d2 $2 {Around about here Black starts to throw away his advantage.} (22... Kb7 $1 {was a very strong plan, trying to relocate the king to a8 where it's completely safe:} 23. Qxf7+ Ka8 {and next move Black will play ...Bb7 and start a counterattack of his own. I think White is in big trouble here.}) 23. Bxd2 Qxd6 24. Qxf7 Rd7 25. Rc1+ Kb7 26. Bxb4 Rxf7 ({Of course, the engine spots} 26... Qxh2+ $1 {instantly. We both missed this during the game, although White would still have some chances to hold the resulting ending.}) 27. Bxd6 Rd8 $2 (27... Rd7 $1 28. Be5 Rc8 {was much better. }) 28. Ne5 $1 {I think Anderson had missed this. Now he can't easily exchange rooks anymore.} Rg7 29. Bb4 (29. Bc5 $5 Rd5 30. b4 $1 {is another clever tactical idea that the computer points out. That said, I doubt it is really a winning attempt: for example,} Rxe5 31. Bd4 Rgg5 32. f4 Rd5 33. fxg5 Rxd4 34. b5 Bxb5 35. axb5 hxg5 {with a drawn rook and pawn ending.}) 29... Rd5 $6 { Failing to call my bluff.} ({After} 29... Rd4 $1 30. Bc3 Rxa4 {, White has no good discovery with the knight on e5.}) 30. Bc3 Rc7 31. Re1 Bc4 32. f4 ({After the game I was kicking myself for a number of stupid things I did in this ending, not least of which was 'why didn't I just play} 32. a5 {- ?'. I was concerned that Black might move the bishop away from c4, take on c3 and then take on a5, when I might have problems dealing with his passed a-pawn. However, for one thing White would have dangerous threats of his own on the kingside and, secondly, it would be psychologically very difficult for Black to make this decision to sacrifice given he's had this extra exchange for almost the entire game. This was the first sign I was starting to get nervous and lose my objectivity.}) 32... a5 33. Kf2 Bb3 34. g4 $2 ({Around about here I began to get a very obvious rush of blood to the head. My first instinct was to play the obvious} 34. Ra1 $1 {, protecting the pawn. Then I think I'd actually have very good chances to win the game as Black's rooks are not very effective and he lacks an obvious point of attack. White, by contrast, has a simple plan of expanding slowly on the kingside. All this I sensed during the game, but in mounting time trouble I suddenly lost my head. --- The fact that I sensed Anderson was starting to lose the thread of the game and that White was now better led me to completely lose my objectivity. I thought why don't I just throw all these pawns at him and see what happens? Who cares about a pointless a-pawn which is doing nothing on the queenside? Those were my rushed thoughts at the time, but I still had enough time to think things through.}) 34... Bxa4 35. g5 hxg5 36. fxg5 Be8 37. g6 a4 $1 {This was the point I had missed. By simply pushing this pawn Black does indeed create nasty counterplay on the queenside, which I have great problems dealing with.} 38. h4 a3 39. bxa3 $4 { With 40 seconds left on my clock, I panic completely.} ({The move my hand wanted to play was, of course,} 39. h5 $1 a2 $1 40. h6 Rxc3 41. bxc3 Rxe5 {. My calculation stopped here as I assumed White must be lost, but the engine points out a beautiful study-like draw with} 42. Rxe5 a1=Q 43. g7 $1 {(the only move, but easily sufficient)} Qb2+ 44. Re2 Qb6+ 45. Re3 {. With more time on the clock I might have found this idea, but with so little time left it's almost impossible.}) 39... Rxc3 40. g7 Rd2+ {. So my first real experience of preparing with Komodo ended in near total disaster, although that wasn't the machine's fault. Only when I was forced to think for myself did the old human frailties come to the surface. For my next game with White I tried to make this whole Geller Gambit work, but it just seemed too risky. Besides Keith Arkell seems to be playing at 2600 strength these days just by outplaying people positionally, so why couldn't I do the same? Indeed, I my next game I chose rather than 5 e4, 5 a4 Bf5 6 Nh4 and won a long game.} 0-1
So my first real experience of preparing with Komodo ended in near total disaster, although that wasn’t the machine’s fault. Only when I was forced to think for myself did the old human frailties come to the surface.
– Part two will follow soon. In it Danny tells us how he used Komodo
to work out a line virutally to checkmate after White's seventh move –
|
Chess Magazine was established in 1935 by B.H. Wood who ran it for over fifty years. It is published each month by the London Chess Centre and is edited by Richard Palliser, Matt Read and John Saunders. The Executive Editor is Malcolm Pein, who organises the London Chess Classic. CHESS is one of most popular English language chess publications and one of the very few in A4 colour format. Chess Magazine (ISSN 0964-6221) is published by Chess & Bridge Ltd, 44 Baker St, London, W1U 7RT. Tel: 020 7288 1305 Fax: 020 7486 7015. Email: info@chess.co.uk, Website: www.chess.co.uk Twitter: @CHESS_Magazine |
|
Subscription Rates:United Kingdom Europe USA & Canada Rest of World (Airmail) Subscription GB / USA. Subscribers can claim a special discount when ordering products online. For instance: Komodo Chess 8 (PC-DVD) RRP £64.95 Subscribers £58.46. UK based readers receive free postage on their order. |
Komodo Chess 8 – Bringing creativity back to chess
Komodo is a chess program that is different from the rest. Its search makes greater use of extensions than any other top engine, which results in a slightly lower average depth, but helps the actual playing strength significantly. It also allows Komodo to see deeper than any other engine, even if it is displaying a lower search depth.
The evaluation of Komodo differs from its main rivals because, unlike the automated tuning generally employed, Komodo's evaluation represents a blend of both automated tuning and the judgment of a grandmaster and computer expert (Larry Kaufman). The programming team have avoided terms and weights that don't make sense to him, even if they happen to test a bit positively. Automated evaluations are subject to rather large sample error, and applying some chess judgment appears to be beneficial, both in results and in the reasonableness of reported evaluations.
Komodo is primarily known for superb positional play. Of course it also one of the top few engines in tactical strength as well, but the programmers have not been willing to sacrifice positional play just to score better on tactical problem sets. It is generally recognized that all good engines are far stronger tactically than any human player, but that when positional judgment is involved, top grandmasters are still superior in many positions to any engine. Therefore it makes sense to emphasize positional play rather than tactical skill: it is better to improve one's weakest point rather than just to further improve the strongest feature.
Larry Kaufman of the USA is a man of many talents. Not only is he a grandmaster and the
2008 World Senior Champion, but he’s also a shogi expert and the brains behind Komodo 8!
Komodo is especially useful for opening analysis, because as an opening specialist Larry Kaufman has always paid close attention to checking whether the program's evaluations in the opening agree in general with accepted theory. Another point in which Komodo excels is the play and evaluation of positions with material imbalance, which it handles more correctly than other top engines. The endgame of Komodo has been improved by the use of Syzygy Tablebases, which are considered the best for actual play and game analysis as they provide only the most essential information to save time and memory.
Another unique feature in Komodo is its superior handling of multiple processors, using a method that is quite different than the usual one. This is most noticeable when using eight or more cores. There is little doubt that Komodo 8 is and will remain the top rated commercial chess engine on most rating lists.
Komodo Chess 8 includes:
- The Komodo 8 engine, which can support up to 64 processor cores and 16 GB of hash memory
- The new Deep Fritz 64-bit program interface (+ 32 bit program interface)
- Online access to the world’s largest analysis database “Let’s Check”, with over 200 million extensively analyzed positions (free access to “Let’s Check” until December 31, 2016)
- Access to ChessBase engine cloud
- Premium membership to Playchess.com for six months
- Database with over 1.5 million games
System requirements
Minimum: Pentium III 1 GHz, 2 GB RAM, Windows Vista, XP (Service Pack 3), 7/8, DirectX9, 256 MB graphics card, DVD-ROM drive, Windows Media Player 9 and Internet access for program activation, access to Playchess.com, Let’s Check and program updates.
Recommended: PC Intel i7 (Quadcore), 4 GB RAM, Windows 8.1, DirectX10, 512 MB graphics card, 100% DirectX10-compatible sound card, Windows Media Player 11, DVD-ROM drive and Internet access for program activation, access to Playchess.com, Let's Check and program updates.
Price: €79.90 (€67.14 without VAT for customers outside the EU; $86.62 without VAT). Languages: English, German. ISBN: 978-3-86681-442-4; EAN: 9783866814424. Delivery: Download, Post
