Speelman's Agony #52

by ChessBase
5/21/2017 – This week's single Agony game comes from Richard Bailey, an Englishman in his mid sixties. The British Boys Under 16 champion in 1968, he represented England at junior level in 1970 (Glorney Cup and Glasgow Junior International) and 1971 (Danish Junior Open). Richard sent in a game against Vernon Dilworth (1916-2004), the inventor of the famous Dilworth Variation of the Open Ruy Lopez. There is a lot to learn in this week's instructive column.

ChessBase 17 - Mega package - Edition 2024 ChessBase 17 - Mega package - Edition 2024

It is the program of choice for anyone who loves the game and wants to know more about it. Start your personal success story with ChessBase and enjoy the game even more.


Glasgow 1970. The Junior International Invitation Tournament was an innovation, held in conjunction with the Scottish Junior Chess Association. Participants included nine players who had represented their countries in the 1970 Glorney Cup, held just a few months earlier at Bearsden, near Glasgow: Bailey, Bellin and Saverymuttu (England); Findlay, Holt, Rosenberg and Sinclair (Scotland); D. James (Wales) and O'Hare (Ireland). Added to that list were John Nunn (England), and Rajinder (Raj) Bhopal and Allan Radlow (Scotland).

The above photo we see at the very back: Iain Sinclair (l) and Michael Rosenberg (r). In front of Michael are Raj Bhopal and Eric Holt. The next row, from left to right, has Stan Beaton of Cathcart Chess Club, who was assisting with the tournament, R. O'Hare, David Findlay and Allan Radlow (partly hidden). In the front row, from left to right, is David James, John Nunn (15 years old), Richard Bailey, Robert Bellin and Seth[na] Saverymuttu. Here the results:

    S B B R H S B N F J O R Pts
S. Saverymuttu ENG ¦ ½ ½ 1 . . . 1 1 . . 1 5
R. Bailey ENG ½ ¦ 0 . ½ . 1 1 . . . 1 4
R. Bellin ENG ½ 1 ¦ 0 1 1 . . . ½ . . 4
M. Rosenberg SCO 0 . 1 ¦ . . 1 0 1 1 . . 4
E.J. Holt SCO . ½ 0 . ¦ ½ . ½ 1 . 1 .
I.J. Sinclair SCO 0 . . . ½ ¦ 0 . . 1 1 1
R. Bhopal SCO . 0 . 0 . 1 ¦ . 0 . 1 1 3
J.D.M. Nunn ENG 0 0 . 1 ½ . . ¦ ½ 1 . . 3
D.J. Findlay SCO 0 . . 0 0 . 1 ½ ¦ . ½ . 2
D. James WLS . . ½ 0 . 0 . 0 . ¦ ½ ½
R. O'Hare IRE . . 0 . . 0 0 . ½ ½ ¦ ½
A. Radlow SCO 0 0 . . . 0 0 . . ½ ½ ¦ 1

Source: Chess Scotland

Richard played in the British Championship in 1969 and 1970 but says “I never progressed at senior level with a mediocre (dormant) Elo rating of 2340. I stopped playing chess competitively in 1982 and emigrated to the Netherlands in 1983. After a brief flirtation with Volmac Chess Club, I met and married a Dutch lady (prophylaxis worthy of Petrosian – preparing for a Brexit more than 30 years in the future!). I now have two children and four grandchildren. My current hobbies are playing Bridge and Go badly, and singing in a local choir.”

The game Richard sent me is against Vernon Dilworth (1916-2004), the inventor of the famous Dilworth Variation of the Open Ruy Lopez starting 1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.0–0 Nxe4 6.d4 b5 7.Bb3 d5 8.dxe5 Be6 9.c3 Bc5 10.Nbd2 0–0 11.Bc2 Nxf2!?


Richard writes:

“This is a game that has worried me for a number of years since the computer chess programs tell me that Black is winning until very late in the game. I recently loaded Stockfish 8 and given enough time it looks as if it might reach a different conclusion but, at least on my computer, maybe not within the tournament time limit! My opponent is V.J. Dilworth who was a good amateur, but I don't recall him having a rating above 3000. So what exactly is going on here?”

My pat answer after just a short glance was that the engines were hampered by the horizon effect (of not analysing far enough to reach a proper conclusion) and I think that that's certainly correct. But before we look at the game itself, it's worth considering in general just what chess engines are and how we can try to work with them.

Although chess engines are nowadays able to defeat even the world's best players in single combat, they are still really only glorified and extremely sophisticated bean counters. What an engine does is to “maxi-min” the results of a move: that is find the move that maximises the minimum evaluation gained against the opponent's best line of response.

The evaluation functions are extremely complex and there are mechanisms (“singular extensions”) to ensure that lines aren't all abandoned at the same search depth but that critical ones are continued until the position supposedly becomes dormant. But if the lines are too long for the engine/hardware combination then a horizon effect can occur where it's fairly obvious to a person that something “ought to be there” but the engine does the arithmetic and decides otherwise.

Working with engines, the most critical thing is to try to use them as a tool, even though they are much better at tactics than us. If, for example, you hate the look of an opening position then there's no point is playing for it, even if your favourite engine is besotted, since you will very likely not be able to play well in a real game. Any chess player above beginner level, where everything is fuzzy, will have a feeling about the positions they reach during a game. When checking with an engine later, watch the computer evaluation and respect it but your gut feeling is more important in practical play.

There are some specific things you can watch out for though. One is a sudden drastic change in the evaluation which can mean only one thing: a more or less hidden tactic. Take this example from a mini-match in Wijk aan Zee where Anatoly Karpov had a disaster in the first game but still fought back to beat Larry Christiansen eventually.

[Event "Hoogovens"] [Site "Wijk aan Zee"] [Date "1993.01.??"] [Round "2"] [White "Christiansen, Larry Mark"] [Black "Karpov, Anatoly"] [Result "1-0"] [ECO "E12"] [WhiteElo "2620"] [BlackElo "2725"] [PlyCount "23"] [EventDate "1993.01.??"] [EventType "k.o."] [EventRounds "5"] [EventCountry "NED"] [SourceTitle "CBM 034"] [Source "ChessBase"] [SourceDate "1993.06.01"] 1. d4 Nf6 2. c4 e6 3. Nf3 b6 4. a3 Ba6 5. Qc2 Bb7 6. Nc3 c5 7. e4 cxd4 8. Nxd4 Nc6 9. Nxc6 Bxc6 10. Bf4 Nh5 11. Be3 {[#]Here Fritz 15 gives White about a quarter of a pawn plus. Karpov played} Bd6 {and suddenly the evaluation jumps to +2.75. Obviously something has happened and indeed after} 12. Qd1 $1 { he resigned immediately.} 1-0

Say you were playing through this game and that Christiansen had also missed Qd1. If you had an engine on and half an eye on it, then you would notice the sudden spike after 11...Bd6, stop and pay more attention. The engine would go “Ping Bd6” and you'd know.

Another important phenomenon is flat lining. This occurs in the endgame where every line ends in exactly the same evaluation and doesn't shift. And what it means is that whatever the value given, the endgame is really drawn.


For the sake of argument, we'll say it's White to play. With the pawn on the seventh, Black can always check the white king away from b6 and with care – for example after g4-5 he mustn't allow the king to settle on e6 when 1.Rd8 Ra6+ 2.Rd6 Rxa7 3. Rd7+ wins – he can draw.

However, computers “think” otherwise. When I chose to have three lines Fritz15 was all 4.14, Komodo 10 all 2.99 and Houdini 5 all 2.46. In all cases, once the evaluations are established they don't shift even if you leave the engines running for much longer. If the position were winning then these would go up to stratospheric levels given time but they flat line. because there is no way for the “advantage” to grow. White has significantly more beans than Black but can't make coffee with them.

On now to Dilworth v Bailey. Richard provided notes which I've used and I've added my own thoughts as JS, Apart from the odd thought, these start at move 20.

[Event "Hoylake Chess Congress Rd. 2"] [Site "?"] [Date "1970.04.11"] [Round "2"] [White "Dilworth, V. J."] [Black "Bailey, Richard"] [Result "1-0"] [ECO "E67"] [Annotator "Bailey,Richard"] [PlyCount "56"] [EventDate "1970.??.??"] [SourceDate "2015.07.13"] 1. d4 Nf6 2. c4 g6 3. g3 Bg7 4. Bg2 O-O 5. Nf3 d6 6. O-O Nbd7 7. Nc3 e5 8. d5 a5 9. e4 Nc5 10. Qc2 Ne8 11. b3 Bd7 ({Immediately} 11... f5 {is better. It is not yet clear where the bishop should be}) 12. a3 f5 (12... a4 $1 13. b4 Nb3 14. Ra2 f5 $11 {JS: One of Black's big problems in the Classical King's Indian is getting the minor pieces to good squares. White's space advantage can be a huge boon, but it leaves weak squares in its wake, and when enemy minor pieces get to them they can be very disruptive. The knight on b3 can go to d4 at some annoying moment, and I'd be distinctly uncomfortable as White.}) 13. Bg5 Qc8 ({ Simply} 13... Nf6 {is better}) 14. b4 Na6 (14... Nxe4 15. Nxe4 fxe4 16. Nd2 Bh3 17. Bxh3 Qxh3 18. Nxe4 h6 $11) 15. Qb3 f4 16. gxf4 {At this point Stockfish (and other programs) give Black the advantage since the bishop on g5 is in trouble} exf4 17. e5 dxe5 18. Rae1 ({Stockfish 8 prefers} 18. c5) 18... h6 $17 ({Stockfish 8 suggests as strongest} 18... Bh3 19. Bxh3 Qxh3 20. Ne4 h6 21. Be7 Rf7 22. b5 Rxe7 23. bxa6 Rxa6 $17) 19. Nxe5 (19. d6 Nxd6 (19... cxd6 20. Nd5 $14) 20. Nxe5 f3 21. Nxf3 a4 22. Qc2 Bf5 23. Qd2 hxg5 24. Nd5 Re8 25. Rxe8+ ( 25. Ne7+ Rxe7 26. Rxe7 Bf6 27. Re3 Nf7 $17) 25... Qxe8) 19... Bxe5 (19... Bh3 20. d6 f3 21. Nd5 Kh7 22. d7 Bxd7 23. Bxf3 Ba4 24. Qxa4 Rxf3 25. Bf4 $16) 20. Rxe5 hxg5 {[#] At this point the initial view of most chess programs is that Black is winning. Only after lengthy analysis does Stockfish decide that the position is equal. JS: My gut feeling here is that White has very serious play but it will take a few moves to develop. There may or may not be good defences and against a decent engine, and against an engine a human as White would probably lose. But against another person, even a very strong player, you'd have excellent chances. Since the lines are long before White gets proper threats, engines will certainly start off by stating that "Black is winning". They will also throw in moves like axb4 or a4 to lengthen the lines, when White's compensation may well go beyond their horizons. What you have to do, when analysing with them is to ignore the assessments until the positions become clear to you and follow your gut, feeding the engines the moves you want to play and waiting for the assessments to change. Perhaps they won't and you're simply wrong, but often your pattern recognition will prevail against their brute force.} 21. c5 ({Stockfish prefers} 21. Rfe1 f3 22. d6) 21... f3 22. d6+ Kh8 (22... Kh7 23. Re7+ Kh6 24. h4 $13 gxh4 25. Ne4 g5 26. Qe3 Rg8 27. Bxf3 {JS: Once you get here, engines realise that Black is in trouble.}) 23. Nd5 fxg2 24. Rfe1 {JS: It's about here that Fritz 15 on my machine starts to realise that things aren't so simple. It had been misled by analysing lines with Rxg5 rather than Re7 from afar (a pawn is a pawn to a bean counter). But now we're close enough for Re7 to register.} Ng7 (24... cxd6 25. Re7 Nxc5 26. bxc5 Qxc5 27. Qb2+ Kg8 28. Qa2 (28. Nb6 Bb5 29. Nxa8 Qc4 30. R1e2 $11 {JS Not sure about the =. This is a complete mess to the human eye but the line} Nf6 31. Nc7 Bd7 32. Qd2 Rf7 (32... Qxc7 $2 33. Qxg5) 33. Qxd6 Rxe7 34. Qxe7 Qxc7 35. Qxf6 Qc1+ 36. Kxg2 Bh3+ 37. Kxh3 Qf1+ 38. Kg3 Qxe2 39. Qxg6+ {is certainly very good for White and, I imagine, fairly accurate if far from the human compass.}) 28... Qxf2+ 29. Qxf2 Rxf2 30. Kxf2 Bc6 31. Rxe8+ Rxe8 32. Nf6+ Kf7 33. Nxe8 Bxe8 34. Kxg2 Bc6+ {JS: this looks as though Black should be okay with so many pawns.}) (24... Nxc5 25. bxc5 cxd6 26. Rxe8 Bxe8 27. Qc3+ Kh7 28. Nf6+ Kh6 29. Re7 Bf7 30. h3 Qxc5 31. Ng4+ Kh7 32. Nf6+ Kh6 (32... Kh8 33. Qb2 Rac8 34. Ne4+ Qe5 35. Rxe5 dxe5 36. Qxe5+ Kh7)) ({JS: Engines are fantastic at finding defences when matters get tight. Here they suggest} 24... Nf6 {[#]The first line given is then The first line given is} 25. Re7 ({While I was inputting 25.Re7 Ba4!, there was an engine spike as Fritz 15 realised that} 25. Qd3 $3 {is winning. Of course, if I were analysing on my own this is a move I might well consider eventually but sorting it out without an engine would be quite difficult. With its help once it opines over +3, it's pretty trivial.} Bf5 (25... Nxd5 26. Qxg6 $1 Rf6 27. Qh5+ Kg7 28. Rxg5+ {and mates.}) 26. Qd4 Nxc5 27. bxc5 cxd6 28. Re7 Qxc5 29. Qxf6+ ({or simply} 29. Nxf6) 29... Rxf6 30. Nxf6) (25. Qc3 $2 Nxd5 26. Rxd5+ Kg8 {defends}) 25... Ba4 $1 ({I stared at this for a little while wondering why not Bg4 or h3 so I asked my friendly engine and it bashed out} 25... Bh3 26. d7 $1 {winning}) 26. Qc3 Qf5 27. R1e6 Qxd5 28. Rxf6 Qd1+ 29. Kxg2 Qg4+ 30. Kf1 Qd1+ (30... Bb5+ $4 31. Ke1) 31. Kg2 { with a draw.}) 25. Re7 Be6 $2 {JS: This loses, as a spike in engine assessments confirms. But it was already very difficult for Black.} (25... Be8 26. Qc3 Rf7 27. R1e6 (27. Rxe8+ Qxe8 28. Rxe8+ Rxe8 29. b5) 27... axb4 28. axb4 Nxc5 29. bxc5 cxd6 30. Rxg6 Qf5 31. Rh6+ Kg8 32. Rxf7 Qxf7 33. Nf6+ Kf8 { [#]Fritz now gives a long line starting 34. Rh8+, but it seems much more human to me to try to close the net with cxd6, though this turns out not to be enough if Black finds the best defence:} 34. Rh8+ (34. cxd6 Qa2 35. Qf3 Qb1+ 36. Kxg2 Ne6 37. Nd5+ Nf4+ 38. Nxf4 Kg7 39. d7 Kxh6 40. dxe8=Q Rxe8 41. Qh5+ Kg7 42. Qxe8 gxf4 43. Qe5+ Kg6 44. Qxf4 {With the b pawn a runner Black should certainly be okay, and indeed when I asked the Lomonosov seven piece tablebase at Moscow State University, any sesnsible move draws.}) 34... Ke7 35. cxd6+ Kxd6 36. Ne4+ Kd5 37. Rxe8 Rxe8 38. Qb3+ Kxe4 39. Qxf7) (25... Bg4) (25... axb4 26. axb4 Bg4 ({JS: Lengthens the analysis path for the engines but only makes a real difference if Black can throw in Nxb4 or Nxc5 at some point.} 26... Nxb4 27. Qc3 Rg8 28. Nf6) (26... Be8 27. Qc3 Rf7) 27. Qb2 Rg8 28. Rf7 Be6 29. d7 Bxf7 30. dxc8=Q Rgxc8 31. Re7 Rf8 32. Nf6 $14 ({JS:} 32. Qf6 {is actually stronger.})) 26. Qc3 Bf7 (26... Rf7 27. R1xe6) 27. d7 Qd8 28. R1e6 Nb8 (28... Nb8 {and Richard resigned, not waiting for} 29. Rxg6 Bxg6 30. Qxg7#) 1-0

Submit your games

Did you enjoy the column and instructive analysis by GM Jonathan Speelman? Do you wish you could have a world-renowned grandmaster analyzing your play? You can! Just send in two of your games: one success story (Ecstasy) and one loss (Agony). Tell why you chose them, where or when they were played, and if they are selected, not only will you get free detailed commentary of your games by one of chess’s great authors and instructors, and former world no. 4 player, but you also win a free one-month Premium subscription to ChessBase Account.

A one-month Premium subscription to ChessBase Account, means that for one month you get:

  • Premium access to the Playchess server with ratings, simuls, lectures, and live commentary of top games.
  • Access to all Web apps with no restrictions, such as the Cloud database (MyGames.chessbase.com), and more!
  • Full access to the Video archive, which not only includes all the past lectures by Daniel King, Simon Williams and others, but also a large number of full ChessBase products you would normally need to buy in the ChessBase Shop, but that you can view for free as a Premium subscriber.

Reports about chess: tournaments, championships, portraits, interviews, World Championships, product launches and more.


Rules for reader comments


Not registered yet? Register