The study in question is the following:
White: Kd6, Rh3, Pb4 Black: Ka8, Na6, Nf8, Pb6
White to Play and Win
I first noticed this study in the book “Modern Endgame Studies for the Chess Player” by Hans Bouwmeester (Study #67). It also appears in “1234 Modern End-Game Studies” by Sutherland & Lommer (Study #582), and in a modified form in Kasparyan’s “Domination in 2545 Endgame Studies” (Study #57).
While playing out the solution in Bouwmeester’s book I found myself asking how well his analysis would stand up to a modern analytic interrogation. To assist me, I collaborated with the Stockfish engine (10 x64) working under the ChessX GUI. The results were very interesting. It is helpful to the cause of truth that the initial position has only seven pieces. I think the Nalimov Tablebases have extended to seven, so one could just look it up. I can’t, not having access, but you know what I mean… But what we can say here is that any capture will drop us into the six player Tablebase which Stockfish seems able to reference as required.
Let’s first consider the solution Mark Liburkin wanted to achieve: 1.b5 Nb8 2.Rh8 Ngd7 3.Kc7 Ka7 4.Re8 Nf6 5.Rxb8 Ne8+ 6.Kd7 Nc7 7.Ra8+ Nxa8 8.Kc8 Nc7 9.Kxc7 and White wins the King & Pawn endgame. The lovely finish was supposedly played by Capablanca against Emmanuel Lasker in a casual game in 1914 (see Study #56 in the Kasparyan book).
Okay, so I’m feeding this line to Stockfish and it says initially +0.48, but as we move along that doesn’t change. There is no win after 2.Rh8. The defensive mistake is playing for the trick with 4... Nf6. Much better is 4... Nc5! 5.Rxb8 Ne6+ 6.Kc8 Nc5 repeats, or 6.Kd6 Nd4! So we can’t take the Knight on b8. The big squeeze with 5.Re7 looks good, but just 5… Ka8 6.Kxb6 Na4+ and Stockfish declares this to be a Tablebase draw. Also, if 4.Rh3 Nc5 5.Ra3+ Nca6+ 6.bxa6 b5 and again a Tablebase draw.
The winning line is 1.b5 Nb8 2.Rf3!! Stockfish starts at +56 but is soon announcing mate. So I’m pretty sure this is winning, but it’s not what Liburkin intended! There is one other defence that should be mentioned: 1.b5 Kb7!? But Stockfish dismisses this with 2.bxa6+ Kxa6 and declares wins for both 3.Rh6 and 3.Rf3 per Tablebase.
A curious side note relates to the Kasparyan “amendment” to the original study. Kasparyan’s version starts with the White Pawn already on b5 and the Black Knight on b8. His “solution” begins with 1.Rh8 which doesn’t actually win.
Kasparyan writes (from the English edition of 1980):
So, was Bron right? Stockfish thinks not. After 1.b5 Nc5 2.Re3!! and White is winning! I won’t give any variations. You can test it yourself. It is hard to argue with these engines!
Anyhow the combination of ChessX as the GUI and Stockfish as the analytic engine provides us all with an incredible tool to dig at the truth. And it can be surprising what you find!
White: Kd6, Rh3, Pb4 Black: Ka8, Na6, Nf8, Pb6
White to Play and Win
I first noticed this study in the book “Modern Endgame Studies for the Chess Player” by Hans Bouwmeester (Study #67). It also appears in “1234 Modern End-Game Studies” by Sutherland & Lommer (Study #582), and in a modified form in Kasparyan’s “Domination in 2545 Endgame Studies” (Study #57).
While playing out the solution in Bouwmeester’s book I found myself asking how well his analysis would stand up to a modern analytic interrogation. To assist me, I collaborated with the Stockfish engine (10 x64) working under the ChessX GUI. The results were very interesting. It is helpful to the cause of truth that the initial position has only seven pieces. I think the Nalimov Tablebases have extended to seven, so one could just look it up. I can’t, not having access, but you know what I mean… But what we can say here is that any capture will drop us into the six player Tablebase which Stockfish seems able to reference as required.
Let’s first consider the solution Mark Liburkin wanted to achieve: 1.b5 Nb8 2.Rh8 Ngd7 3.Kc7 Ka7 4.Re8 Nf6 5.Rxb8 Ne8+ 6.Kd7 Nc7 7.Ra8+ Nxa8 8.Kc8 Nc7 9.Kxc7 and White wins the King & Pawn endgame. The lovely finish was supposedly played by Capablanca against Emmanuel Lasker in a casual game in 1914 (see Study #56 in the Kasparyan book).
Okay, so I’m feeding this line to Stockfish and it says initially +0.48, but as we move along that doesn’t change. There is no win after 2.Rh8. The defensive mistake is playing for the trick with 4... Nf6. Much better is 4... Nc5! 5.Rxb8 Ne6+ 6.Kc8 Nc5 repeats, or 6.Kd6 Nd4! So we can’t take the Knight on b8. The big squeeze with 5.Re7 looks good, but just 5… Ka8 6.Kxb6 Na4+ and Stockfish declares this to be a Tablebase draw. Also, if 4.Rh3 Nc5 5.Ra3+ Nca6+ 6.bxa6 b5 and again a Tablebase draw.
The winning line is 1.b5 Nb8 2.Rf3!! Stockfish starts at +56 but is soon announcing mate. So I’m pretty sure this is winning, but it’s not what Liburkin intended! There is one other defence that should be mentioned: 1.b5 Kb7!? But Stockfish dismisses this with 2.bxa6+ Kxa6 and declares wins for both 3.Rh6 and 3.Rf3 per Tablebase.
A curious side note relates to the Kasparyan “amendment” to the original study. Kasparyan’s version starts with the White Pawn already on b5 and the Black Knight on b8. His “solution” begins with 1.Rh8 which doesn’t actually win.
Kasparyan writes (from the English edition of 1980):
“In the original version of the endgame White’s Pawn stood on b4 and Black’s Knight on a6 instead of b8. The solution started with the moves 1.b5 Nb8. In 1957 V. Bron proved that Black can save himself by 1…Nc5! 2.Rh8 Ne6 3.Ke7 Nd4 4.Rxf8+ Kb7=. That is why the endgame is presented in a revised version.”
So, was Bron right? Stockfish thinks not. After 1.b5 Nc5 2.Re3!! and White is winning! I won’t give any variations. You can test it yourself. It is hard to argue with these engines!
Anyhow the combination of ChessX as the GUI and Stockfish as the analytic engine provides us all with an incredible tool to dig at the truth. And it can be surprising what you find!
Comment