Friday, 19 May 2017

018. All Rook Endgames are Drawn

Black: tripoduk -, 2017

For five years now I've been playing an ongoing series of friendly games on against someone under the name of “tripoduk”. I generally visit the site first thing in the morning, make a couple of quick moves, and go away again. As my opponent is much weaker than me, and as I don't play against anyone else on, my win-to-loss ratio is very high indeed. In fact there have been no losses at all, and when I'd racked up a score of +89 =2 -0 (98.9%), the site administrators decided I must be cheating and banned me – albeit only temporarily. Once I'd explained the situation to their satisfaction (CC-SIM vs. relative novice), they let me back on under a new name and gave me a Diamond membership too as a titled player. Well, grand.

Since then we've played another 41 games, the score in which is currently +32 =9 -0 (89.0%) in my favour. Still no losses, but you'll notice that my percentage has decreased somewhat. My opponent is getting better. And in the game below, he should have had me.

Most of the games with me as White have been King's Gambits, and the following critical line of Becker's Defence has arisen several times: 1 e4 e5 2 f4 exf4 3 Nf3 h6 4 d4 g5 5 Nc3, followed by 6 g3. John Shaw's King's Gambit book (pp 171-6, 231) has been useful here, and up to 2016 I'd scored 8/8 as White: five wins after 5...d6 6 g3 Bg7 7 gxf4 g4 8 Rg1!; one each against 6...Nc6 and 6...g4; and one more after 5...Bg7 6 g3 fxg3 7 hxg3 d6 8 Be3.

In our most recent game, my opponent repeated the last of those variations – and much more strongly. After his 12th move I was struggling to find any compensation at all for the sacrificed pawn. I continued struggling to find any as the game progressed, and watched rather helplessly as my position declined from worse to bad to lost. In the diagram below, White is completely lost.

The logical plan for Black now is to trade the g-pawn for the white a-pawn, swap the bishops off in the process, and win the rook endgame with two extra pawns.

It's not entirely trivial: the direct 55...Bb2 56 Kxg5 Bxa3? etc leads to a draw (as in the game); but rather than take the a-pawn at once, the switchback 56...Be5! forces the white rook to give way (57 Rd7+ Ke6 58 Rd8? runs into 58...Bf6+), after which Black should win easily enough. Temporizing with 56 Kf5 doesn't help either, since 56...Bxa3 57 Bxa3 Rxa3 58 Kxg5 wins for Black, if only just (58...Ke7! 59 Rd4 Ra1 60 Rxd3 a3 61 Rd2 Rf1! 62 Ra2 Rf3 63 Kg4 Rb3 64 Kf4 Kd6 65 Ke4 Kc5 and so on); while 56 Bd2 Rxa3 57 Kxg5 (or 57 Bxg5 Rc3) 57...Be5! is equally hopeless for White.

The immediate 55...Be5! is good too; e.g. 56 Rd7+ Ke6 57 Rd8 Bb2, when White can't take on g5 and otherwise Black takes on a3 and wins again.

Instead, I was given a reprise: 55...Bc3? 56 Bxc3 Rxc3 57 Kxg5 Rxa3 58 Kf4 and this rook endgame is drawn since Black can make no progress; e.g. 58...Rb3 59 Ke3 a3 60 Kd2 Ke7 61 Ra6 and if the black king advances up the board it gets checked away from behind; or 58...Ra1 59 Rxd3 a3 60 Rd2! a2 61 Rf2 (safeguarding the white king from checks) and if the black king advances up the board to the third rank, it gets checked away from the side. This made me think of Tarrasch's maxim: “All rook endgames are drawn”. By which he meant that even quite favourable-looking rook endgames can be difficult to win, and sometimes they can't be won at all.

As it happens we've just started another game with the same line (up to 8 Be3 so far), so I'll have to find an improvement pretty soon. Hmmm.

Saturday, 6 May 2017

017. Numerical Error

White: G. Benson - Koshnitsky Memorial Tournament, 2002

Most correspondence chess nowadays is conducted on online servers. Rather than replying to your opponent in hard or electronic format, as we used to do, you go and make your move on an actual board. The advantages of this method are significant: there is no chance of moves going astray, the time used for “thinking” is clear, and those errors unique to correspondence chess, such as writing your move down wrong, are ruled out. Obviously, human stupidity can never be prevented entirely; input errors are still possible.

The (Gary) Koshnitsky Memorial was played by email, using international numerical notation. In this system the squares on a chessboard are each given a two-digit code according to its file (1-8; replacing the usual algebraic a-h) and rank (1-8 as well), so that the square f3, for instance, is now 63. Making moves involved sending a four-figure number: the first two digits denoting the starting square of the piece or pawn you want to move, the second two digits indicating its destination. Thus 1 e4 would be sent as 5254.

International numerical was standard for international correspondence games, because different countries have different initials for the pieces arising from their different languages: such that a bishop is B in English, but A (alfil) in Spanish, C (слон) in Russian, F (fou) in French, L (läufer) in German, and so on. Using numbers rather than letters eliminates those discrepancies. The slight drawback in human terms is that moves are harder to visualize. 1 Nf3 may be immediately appreciable, but 7163 you have to think about for a moment. In the game below I failed to visualize a move correctly and duly made a numerical error.

The position in the diagram, arising from a Scotch Game, had been seen before. H.Staudler-V.Piccardo, 19th World CC Championship ¾-final 4 1999, continued 21 Ne3 0-0 22 Nxd5 Rxd5 23 f4 Bg5 24 Rc4 Re8 25 Kf2 c5 and was later drawn. My game went a different way: 21 Be2 c5 (aiming to ease my defence by eradicating the queenside pawns) 22 Ne3!? (an unexpected and dangerous sacrifice) 22...cxb4, and now 23 Bb5+ Kf8 24 Kf2 bxa3 25 Rhd1 gives White a strong initiative for the pawn. I hoped to be able to defend, but it didn't look at all easy. And sure enough my opponent played 23 Bb5+, but not as 23 5225. Instead, 23 6125 came back. Checking his earlier emails I discovered that White's 21st move was not 21 Be2 (6152) at all but 21 Kf2 (5162), which means that he basically has an extra tempo for his attack.

Returning to the diagram again: 21 Kf2 c5 22 Ne3 is now not even a sacrifice, and 22...cxb4? (22...0-0 is necessary here) 23 Bb5+ Kf8 24 Rhd1 gave White a big advantage. Subsequent play led to an endgame with rook vs. bishop and knight which I was unable to hold. Fuckadoodledoo.

Fortunately, that was my only loss in the tournament and I finished in joint second place, with a nice win against OTB GM Colin McNab along the way, and received a cheque for a pleasant amount of Australian dollars. I also, as I discovered years later, achieved a CC GM norm in the process. Damn, if I'd realized that at the time I might have tried to get another one. Nevertheless, I can't quite regard this as a missed opportunity, since in my next major tournament I came last.

Sunday, 26 March 2017

016. The Double Rosentreter

Black: Carpo - thematic tournament,, 2006

In an earlier post, I mentioned in passing the following variation: 1 e4 e5 2 f4 exf4 3 Nf3 g5 4 d4 g4 5 Bxf4 gxf3 6 Qxf3 Nc6 7 Bc4 d5 8 Bxd5 Nxd4 9 Bxf7+ Kxf7 10 Qh5+ Kg7 11 0-0 “with strong play for the two sacrificed pieces in the style of the Double Muzio.” Given that here too White sacrifices the knight on f3 and the bishop on f7, it seems appropriate to dub this the “Double Rosentreter”.

As in the Double Muzio, the practical difficulty for Black is in defending against threats which mount very quickly on the f-file. The problem for White is in obtaining anything more from this than a draw. Indeed, Black could have forced one by playing 10...Ke6, after which White has nothing better than to repeat with 11 Qe5+ Kf7 12 Qh5+ etc. But supposing Black doesn't want that, what then? Then interesting things can happen.

In the following game, 13 Qh6+ was my prepared novelty. I had drawn ten years earlier with 12 Na3 Nf6 13 Bh6+ Kg8 14 Qg5+ and so on. Later, the Russian player (and future CC-SIM) Enver Efendiev did substantially better after 12 Nc3, winning two very short miniatures (including one against my previous opponent); but neither of those featured 12...Nf6.

After that move, 13 Be5 Qe8 14 Qg5+ Qg6 15 Qxg6+ Kxg6 16 Bxd4 Rd8 saw Black consolidate very easily in Mic.Schulze-J.M.Johansen, IECG 2001. (The two pawns are not worth anything like the extra bishop.) And 13 Bh6+ Kg8 14 Qg5+ Kf7 15 e5!?, from V.Chetvertakoff-F.Arbis, IECC 2001, also seems a bit dubious after 15...Nf5!? 16 Rxf5 Bxf5 17 Qxf5 Qd7 18 Qf2 Qc6; while 15 Qh5+ Kg8 16 Qg5+ Kf7 is another draw (unless Black wants to try 15...Ke6!? perhaps).

Okay, it's probably true that White doesn't have anything after 13 Qh6+ Kf7 14 Be5 either. Houdini resolutely declares it “0.00” and two other games have ended in two more draws after 14...Bg4!. All the same, I think the Rosentreter (with 5 Bxf4), which I first played in 1989, is worth an occasional punt. The lack of any immediate threats allows Black a rather confusing freedom, and in such cases you can easily find that you've put your pieces on the wrong squares. And then it's often too late.

In my game below it is only when we reach move twenty that the engine starts to see some promise in White's position. Even if it's not necessarily all over for Black at this point, it's certainly much easier to play White. Just keep moving forwards. At the end, Black had presumably worked out that he would soon lose a lot of material and so resigned. I'd have played it out myself, but I quite understand the feeling of not wanting to look at a game anymore.

Friday, 10 March 2017

015. Losing From Both Sides

Black: H.J. Hofstetter - 14th CC Olympiad (Final), 2002

In his recent, entertaining book, Your Opponent is Overrated (Everyman 2016), James Schuyler tells of a series of bets he had with Alex Sherzer:

Sherzer set up a simple endgame position with five minutes on the clock and claimed he could win this as White, then turn the board and draw it as Black. Schuyler took the bet. Sherzer won and pocketed the cash. He then said that in another position he could do the same thing twice; i.e. win as White, draw as Black, and then – after having shown how White might win and Black might draw – win with White and draw with Black again. Schuyler took the bet. Sherzer won and pocketed the cash. He then said that in another position he could do the same thing three times: win, draw, win, draw, win, draw. Bet taken. Sherzer won again and pocketed the cash. Having lost enough money and pride by this point, Schuyler gave up.

That story reminds me, in reverse, of Bogoljubow's boast: “When I am White I win because I am White; when I am Black I win because I am Bogoljubow!” Even for me there are certain positions, certain openings, where I expect to win with either colour, but okay, the starting position is not one of them. And there are occasions when I have lost positions from both sides.

In another main line Zaitsev Ruy Lopez (as seen earlier in the Rorschach Knight), the following position (after 28 Qc2) has arisen in two of my games.

In the first (as Black), I played 28...Ra1+ 29 Kh2 Ba6 30 e5 dxe5 (someone had previously drawn with 30...Nxe5 but, rightly or wrongly, I didn't like 31 Re4 here) 31 Nd2 b3 32 Qc3 Nf4 (if 32...Qh4 then 33 Bg5) 33 Ne4 Qf5 34 Bxf4 Qxf4 35 Re1 Rxe1 36 Qxe1 Bb7 37 Nd6 Rb8 38 Qb4 c3 (the only move) 39 Qxb3 c2 40 Qxc2 Bxd5 and now a pawn down for nothing, I eventually lost, U.Strautins-J.Tait, ICCF Champions League 2002.

In the second game (as White), my opponent played 28...Ba6 straight away: 29 e5 dxe5 30 Nd2 b3, the difference being that 31 Qc3 can be met by 31...Qh4 32 Bg5 Qd4, since the e8-rook is still defended by the one on a8. The fact that I had already lost as Black in this variation encouraged me to go for a win as White, so I tried 31 Ne4 Qf4 32 Qd2 Qxd2 33 Rxd2? Reb8 34 Nf6+ Kh8 35 Rg4 Rb6 36 Ne4 f5 37 Rh4 Kg8 38 Be3 g5! and ended up losing again (as you'll see below).

In that game I should have accepted that White had no advantage and opted for 33 Nf6+ Kg8 34 Bxd2. Or else 31 Qc3 anyway, at which point a draw was agreed in R.Bocanegra Moreno-I.Cavajda, correspondence 2010. 29...Nxe5 is also fine for Black, as seen in V.Palciauskas-A.Lanc, 13th CC Olympiad 2004 (which took place later than the 14th Olympiad, since this one was being played by post rather than email).

And as it happens Black may not be losing in my first game either. Looking at it now, Houdini shows a possible defence: 32...Nxf2! 33 Rf3 Rh1+ 34 Kg3 Qh8 35 Qe3 (my notes indicate that I did consider this, but not...) 35...Ng4! 36 Kxg4 (or 36 hxg4 Qxh6) 36...c3! 37 Qa7 Qh7!, when the computer claims equality in a morass of complications. (I've given a few silicon-generated variations below; the 41...Qh4+ line where Black loses with two queens is worth seeing.)

Sigh. What can I say? When I am White or Black I lose because... it seems I don't really understand chess at all.

Saturday, 4 March 2017

014. Losing With Someone Else's Moves

White: R. Smith - BCCA Centenary Tournament, 2006

The converse to Winning With Someone Else's Moves is losing with them. In correspondence chess this mostly occurs when you haven't prepared properly: when you play an opening variation with which you have very little experience, following someone else's analysis that you haven't checked.

In the BCCA Centenary GM tournament, I decided (for some reason) to abandon my usual repertoire and play "sensible" main lines; e.g. 1 d4, 2 c4 as White, the Semi-Slav and Ruy Lopez as Black. The fact that I was extremely busy around then and unable to concentrate on my games is not an excuse; it just makes my decision more ridiculous. I was consciously playing positions I didn't understand, against strong opponents, without even the time to study them properly. I duly came last on 3½/14 without winning a single game.

The one below is a case in point. I ventured the Marshall Attack against the strongest player in the tournament, CCGM Robin Smith, without having done any specific preparation, and without ever having played it in a serious game before. Reaching the basic tabiya after 13 Re1, I suddenly had to decide what to do next. In a recent-ish issue of Chess Informant (#89), I came across an interesting consultation game (between two teams of OTB masters, played by SMS text messaging), which continued 13...Bf5 14 Qf3 Re8 15 Rxe8+ Qxe8 16 Nd2 Qe1+ 17 Nf1 Bg6 18 g3 b4 (a theoretical novelty)

19 Bxd5 cxd5 20 Qxd5 Rd8 21 Bg5 Qxa1 22 Bxd8 Bf8 23 Ba5 Qb1 24 Kg2 Bxd3 25 Nd2 Qc2 26 Qa8 Qxd2 27 Bxb4 h5 28 Qxf8+ Kh7 29 Qxf7 Qe2 30 Qf4 Qf1+ 31 Kf3 Qe2+ 32 Kg2 Qf1+ 33 Kf3 Qe2+ 34 Kg2 Qf1+ ½-½. This was annotated in detail by one of the team leaders, GM Igor Nataf, and his analysis seemed pretty good.

Sigh. In correspondence chess “seems pretty good” is a terrible mindset. Analysis is either good or it isn't, and either way it needs to be checked beforehand. My opponent deviated immediately from the published game, opting for 19 c4, and only at this point did I begin to examine it all. Nataf's main line ran 19...Nf6 20 Qxc6 (20 d4 is met by 20...Bh5!) 20...Rd8 21 Qb6 Rd7 22 Bc2 Ng4 23 Rb1 Qe2 24 Bd2 Ne5 25 Qe3 Qh5 26 Bd1 Qf5 27 Be2 Bf8, which he assessed as slightly better for White. I didn't much like the look of that. What happens after 28 Re1, for instance? Nataf in fact gave 19...Nf6 a dubious mark.

However, there was also a one-move bracketed note: 19...Nc7, with the accompanying text “!?= Radjabov, Bacrot”. That is to say: interesting, equal, and suggested by world-class grandmasters Teimour Radjabov and Etienne Bacrot. Okay, great, I'll try that. It even sets a little trap: 20 Qxc6?? Bd3 21 Qg2 Re8 and Black wins with ...Bc5 and ...Re2. Hurrah.

A CCGM is not, of course (underlined seventeen times), going to fall for that. Instead, Robin came up with something much stronger: 20 d4! c5 (there's no ...Bh5 here) 21 d5 Re8 (otherwise Ba4 and Qe3 follows) 22 Bd1!, breaking the pin on the c1-bishop and allowing White to drive the black queen away with Bd2 as necessary. Black now has nothing for the pawn, the now connected passed pawn. Several precise, grandmasterly moves later, I resigned.

As it turns out, there was nothing much wrong with Nataf's analysis. Black has since done perfectly well, answering 19 c4 with 19...Nf6, but playing 27...f6! (rather than 27...Bf8) at the end of the above variation. And Peter Svidler scored an impressive victory as Black against Vassily Ivanchuk after the similar 19 h4 h5 20 c4 Nf6 (where 21 Bd1 did not prove to be anything like as strong).

So, perhaps the title of this post was not the most appropriate. It should perhaps have been ‘Losing After Someone Else's Moves’. Even if I might point to 19...Nc7 as someone else's mistake, actually playing it is obviously my own fault. If you're going to give a casual suggestion (whether from world-class grandmasters or not) a test in a serious game, at least investigate it in advance. Ya drongo.

Robin went on to score an unbeaten 11/14, winning the tournament by 1½ points – an impressive result in what turned out to be his final event. He died in 2009.

Sunday, 19 February 2017

013. Winning With Someone Else's Moves

White: paardesprong - thematic tournament,, 2005

One of the benefits of studying openings is that you can, very occasionally, win games with moves you've worked out entirely at home. Or even with someone else's moves that you've just remembered. My most notable instance was against Colin Crouch at the 1990 Nottingham Congress. In a sharp line of the French, Black answered a classic bishop sacrifice on h7 by giving up his queen for three pieces. This was reckoned at one time to offer Black good play but had since been refuted in the game S.Szilagyi-T.Harding, ICCF World Cup 1986. Colin helpfully played straight down this line, stopped, thought for an hour, and then resigned – and a move sooner than in Szilagyi-Harding, so I didn't have to think of a single move for myself.

This hardly ever happens to me in over the board chess anymore. For one thing, I mostly avoid theoretical main lines nowadays; and, for another, I usually find I've forgotten most of my home analysis, even in my own pet lines. In correspondence chess, it's a different matter. With everything written down (or entered on ChessBase), I do still win games where most of my moves have been worked out (and computer checked) in advance. And it's possible to win games with someone else's moves too.

When investigating the Calabrese Counter-Gambit (1 e4 e5 2 Bc4 f5!?) in the mid 1990s (see Game 009), I got round to considering what would happen if White replied with 3 f4!?. It seemed to me that 3...exf4 was the best response, transposing into a variation of the Bishop's Gambit: 1 e4 e5 2 f4 exf4 3 Bc4 f5!?. Against this, theory recommended that White play 4 Qe2 Qh4+ 5 Kd1 fxe4 6 Qxe4+ Be7 7 Nf3 Qh5 8 Bxg8 Rxg8 9 Nc3 Nc6 10 Re1 d6 11 Nd5 Bf5 12 Qc4 Bxc2+ (intending 13 Kxc2 Qxd5! etc) 13 Ke2.

Here analysis by Igor Glazkov continued 13...Qg6 (L.Hoffer-Grischfeld, London 1882) 14 Kf2! Kd7 15 Rxe7+! and wins, or 13...Ne5 14 Qxc7 Qf7 15 Qxb7 Rd8 16 Kf1! with advantage to White.

But, as it happens, Eric Schiller had refuted all this in his (unfairly disparaged) little book Who's Afraid of the King's Gambit (Chess Enterprises 1989), where he gives 13...Bh4! 14 Nxc7+ (or 14 d4 0-0-0) 14...Kd7 15 Nxa8 Re8+ 16 Kf1 Rxe1+ 17 Nxe1 Qd1 18 g3 fxg3 19 hxg3 (or 19 Qxc2 Qxe1+ 20 Kxe1 gxh2+ 21 Ke2 Nd4+) 19...Bxg3 20 Qe2 Bd3 21 Qxd3 Qxe1+ 22 Kg2 Qf2+ 23 Kh3 Qh2+ 24 Kg4 Ne5+ and Black wins.

This was picked up later (2004) by Thomas Johansson, but did not make it into general circulation. For instance, the one-volume encyclopaedia Nunn's Chess Openings (Everyman, Gambit 1999) stops at 13 Ke2, with an exclamation mark and the symbol denoting that “White is much better”.

So, so far I've won three times as Black after 13 Ke2. The game below was the shortest – and in that one, too, I didn't have to think of a single move for myself.

Saturday, 11 February 2017

012. A Bust to the Bishop's Gambit

Black: tarby - thematic tournament,, 2015

I once wrote an article (again for the BCCA magazine) entitled “A Bust to the 7...Qc7 Winawer - ?”. It was based on my games with a home-made plan involving pushing the kingside pawns. My record over the board against the main line is indeed quite favourable: P41, W31, D9, L1 (grrrr). All the same, the title was a little joke, a nod to Fischer's famous “A Bust to the King's Gambit”, in which he proposed that 2...exf4 3 Nf3 d6! wins by force.

Actual busts of opening variations are pretty rare. Sooner or later someone usually finds a way to refute the refutation. In both the supposed “busts” above, theory has indeed moved on substantially. All the same, I think Black's position is objectively better against my line than White's position is against Fischer's – which is as you'd expect. Robert James Fischer, the 11th World Champion, was one of the strongest players in the history of chess. Naturally, therefore, his opinions (at least on chess) are going to worth more than mine, a mere correspondence IM (SIM) with a current, rather specious OTB rating of 219 ECF (roughly 2350 Elo).

GM John Shaw is also much stronger than me, with a current ECF rating of 237 (roughly 2500 Elo), but he is not Fischer either. In his recent megatome The King's Gambit (Quality Chess 2013), Shaw includes a chapter “The Refutation of 3 Bc4?!”, in which he recommends 3...Nc6 and claims a definite advantage for Black. Subsequent investigators have disputed this.

The critical line in question runs 1 e4 e5 2 f4 exf4 3 Bc4 Nc6 4 d4 Nf6 5 Nc3 Bb4 6 Nge2 f3! 7 gxf3 d5 8 exd5 Nxd5 9 0-0 Nxc3 10 bxc3 Bd6.

Here Shaw writes: “What is good about White's position? Nothing, unless you think four pawn islands versus two is a plus. The following game confirms Black's superiority.” That was Belanoff-Simmelink, correspondence 2007, which continued 11 Ng3 0-0 12 Ne4 Be6 13 Bxe6 fxe6 14 Rb1 b6 15 d5 Ne7 16 dxe6 Qe8 17 Nxd6 Qg6+ 18 Kh1 Rad8 19 Ba3 cxd6 20 Qe2 Rf6 21 Rbe1 Nf5 and “Black eventually converted his advantage”.

Shaw also comments that “The simple 12...Be7 maintains an edge”. As it happens, I had previously lost a game following that very move: 12...Be7 13 Kh1 Na5 14 Bd3 f5 15 Ng3 c5 16 Rg1 cxd4 17 Rb1 dxc3 18 Nh5 Rf7 19 Bf4 g6 20 Be5 Bc5 21 Rg2 Qh4 22 Bxc3 Qxh5 23 Bxa5 Be6 24 Qf1 b6 25 Bd2 Bd5 26 Rg3 f4 27 Rh3 Qg5 28 Rg3 fxg3 29 Bxg5 Bxf3+ 0-1 tsmenace-Carpo, 2005.

There is nothing new in chess. In his earlier book The Fascinating King's Gambit (self-published 2004), Thomas Johansson, too, had flagged 6...f3! as “Black's best option” and continued the line to 12 Ne4, which he assessed as unclear, writing that “White's pawn structure may not be the healthiest, but on the other hand White still has more influence in the centre and a half-open g-file.”

So who is right? In my game, White might improve with 19 Qe1 or simply 17 cxd4, intending 17...Qxd4 18 Bh6 Rf7 19 Nxf5, but I'm not keen on the position after 12...Be6. Instead, on the ChessPublishing forum, Stefan Bücker proposed 11 Qd2 and 12 Qg5, writing that “When the queens are exchanged, there is not the slightest reason why Black's fewer pawn islands should be a factor. White's center may well be more important.”

Which brings us finally to the game below, where I had a chance to test Stefan's idea out. As you'll see, Black kept the centre pawns under control, but didn't manage to do much else. And another 2016 correspondence game quickly ended in a draw too. So it seems that, while Shaw's recommendation leads to a position that you might not want to play over and over again, White is probably not worse here.

Unless, of course, you know different...