Study of the Month - A short history of endgame study castling I

by Siegfried Hornecker
4/30/2022 – Castling is a fascinating move. According to Wikipedia, "it originates from the 'king's leap', a two-square king move added to European chess between the 14th and 15th centuries, which took on its present form in the 17th century". The intricacies of the rules of castling have inspired many studies. Specialized columnist Siegfried Hornecker presents some of them. | Photos: Pixabay

Chess Endgames 1 to 14 Chess Endgames 1 to 14

All endgame DVDs by Karsten Müller in one package! More than 70 hours of instruction! from "Basic knowledge for beginners" (volume 1) to "Practical Rook Endgames" (volume 8) to the ever-popular "Golden Guidelines of Endgame Play" (volume 14).


A short history of endgame study castling I

Due to an error in my last "Study of the Month", a study was omitted that should have been included. It was included as a correction a few days after the article went online, and due to the theme of that study, the retro-legality of castling, I decided to show some more castling ideas this month but without any retro content.

In an endgame study, it is assumed that a player can castle unless proven otherwise, and an en passant key is allowed only if it is proven that the last move was the double-step of the pawn. In this article, I want to provide a short overview of the history of castling in endgame studies - by no means complete. Due to the sheer mass of studies, it will be a multi-parter. (This doesn't mean every month now will continue it, similar to "Check Czech Chess" miniseries.)

Solution to last month's retro study

Due to an error I made, a correction was issued last month with the challenge for interested readers to solve the following study, knowing that there is retroanalysis involved.


Attila Koranyi. 2nd Friendship Tourney 1965. 1st prize

White to move and win

To my knowledge, it is impossible to analyze this correctly with a computer without changing castling rights depending on which line is analyzed, as one castling excludes the other. As such, the database of Harold van der Heijden incorrectly lists the study as incorrect, as there the Black castling rights aren't entered (if they were entered, people would report it also as cooked, as the retroanalytical explanation is not in the database).

White has seven pawns, the f-pawn is missing. The rook on f2 can have been promoted, but in that case, the king on e8 must have moved. If it is not promoted, the king on e1 must have moved. As such, only one of the two castlings is possible. Now per Codex - the governing ruleset for chess composition that is however not complete, and especially regarding fairy chess where rules have strange interactions, so the composers decide usually what happens. It would likely be impossible to create a ruleset that includes everything that might happen on the chessboard, so the Codex covers nearly only the orthodox, normal rules. The retroanalysis for the position is given in detail in the sources, but this is the short explanation: The side to castle first proves by castling that it had the right to castle. As such the other side, as castling rights are mutually exclusive, can't castle anymore.

1.0-0! bxc5 2.c4 g6 3.Kh1 Qa4 4.Nxg6 Qd1 5.Rxd1 Rxg6 6.Rdf1 wins

Apart from 6.Rg1 also winning, the solution is a tactical battle to exploit the checkmate threat on f8. 1.Rhf1 0-0-0 would be Black's proof that he can castle, so Rf2 has come from a1 originally. Importantly, 1.Ng6 bxc5 2.0-0 c4 pins the rook on f2, whereas 1.Ng6 bxc5 2.Rhf1 0-0-0 has Black castle because White didn't castle. This shows the entire absurdity but also interesting logic of such an endgame study showing something that never can occur in a practical game. Although if we look at a game between Heidenfeld and Kerins with three castlings, maybe nothing is impossible...

A viral study and the humble beginnings

However, for our studies this month, we will look at castling without retroanalytical content. Recently a study went a bit viral on Youtube apparently. At least I saw it in two videos, and you can also replay it at the end of this article:


Tigran B. Gorgiev. Shakhmaty v SSSR 1938 (1st semester). Commendation.

White to move and win

After the introduction 1.Bb1 Bxb1 readers should easily be able to solve the study. A chess grandmaster complained in his German language video that for him endgame studies represent chess endgames, so he does usually not consider castling, as for him that is an opening move. The grandmaster generally concentrates on practical play, but admits in his video that he likes to solve endgame studies for training. Another Youtube channel in English had a period recently of showing Gorgiev studies where also this one was included, so (speculation follows) it might be possible the German grandmaster who originated in Greece had seen the study there and tried to solve it.

This should also set the tone, after the serious topic last month a rather lighthearted approach for this month should balance things out, as this month's endgame studies are for entertainment and education purposes only and not relevant to practical play.

While early composers already used castling positions (i.e. wKe1 and wRa1 or wRh1, or mirrored for Black), they did not use actual castling in the solution. The earliest example I found where castling really happens is a study by Julius Mendheim:


Julius Mendheim, Handbuch des Schachspiels 1843 (by von der Lasa)

White to move and win

1.Qxf4+ Kxf4 2.Rf1+ Kg4 3.Rf4+ Kxf4 clears the way for 4.0-0+ Kg4 5.Ne3+ Kh4 6.Nf3 mate

Here the castling is used to unpin two knights. As such it has a logical component: After 2.-Kg4 the main plan 3.Ne3+ Kh4 4.Nf3 mate fails because both knights are pinned. The foreplan 3.Rf4+ Kxf4 4.0-0+ Kg4 unpins the knights. Unfortunately the study does not strictly exhibit the logical school's purity, as also 3.Ne3+ Kh4 4.Rxh3 mate would be possible if the knight was not pinned. With some modifications the study could be made to be "logical", moving Bb3 to c8 would be one idea, but the exact execution of such an improvement would be beyond the scope and theme of this article and as such is left to other composers.

Early composers in many cases used gamelike positions where castling would have been allowed but had no role in the solution. The artistic usage of castling in fact had no strict conventions and so even great composers might struggle to include it in their endgame studies.


Leonid Kubbel, Rigaer Tagblatt 1909, commendation

White to move and win (incorrect)

1.Bf5 is a strong move that abuses the situation of Black. However, as Black can have moved with his Ph6 in the move before this, and White could have captured a piece in his previous move, Black can draw with 1.-0-0 as Thomas Rayner Dawson pointed out in "The Chess Amateur", December 1915. The study is easy to correct by mirroring the position on the d-file/e-file axis: bKd8 and bRa8 can't castle.

In fact, the same error was repeated by other composers later.

If I talk about the castling study by Selezniev, Tidskrift för Schack 1921, you certainly will know that this study is meant:


Alexey Selezniev, Tidskrift för Schack 1921.

White to move and win.

However, there is another study with the same source, just five problem numbers later. Something does not add up, however, as the previous problem was number 133 from December 1921, and the next one supposedly is number 138 from January to February 1921. I assume an error in the database, with it being from 1922.


Alexey Selezniev, Tidskrift för Schack 1922 (version by Pal Benko, Chess Life, December 1999).

White to move and win.

The retroanalytical content here is easy to solve: Black must have moved king or rook last, so he can't castle. Thus 1.b6 a:b6 2.Ra1 Kf8 3.Kf6 etc. wins

The endgame studies so far had castling more as a gimmick that made up the entire solution. However, the Soviet School emerged, themes begun to become aesthetically pleasing, complex. Multiple styles with geometrical motifs or just endgame-like studies coexisted, and do to this day. While the themes have become more oriented towards logic nowadays, back then tactical and geometrical play dominated - including dominations. Not only the Platov brothers were able to compose in multiple styles, and Vasily Platov on his own created the following study where castling is used only in the introduction.

New ideas


Vasily Platov, Shakhmaty 1924.

Black to move, White wins.

It is notable that not every endgame study requires White to begin. While White fulfills the stipulation, it could be that the composer finds that the best value is with Black to play, usually starting with a move that puts pressure on White.

1.-Ra2 is exactly such a move. It is impossible to take the rook unless White desires a quick loss. The pawn on the a-file hangs. 2.Rd1? b2 is devastating. So what can White do? Well, the other main variation (why is it one?) is given as 1.-Rf2 2.0-0-0 Rc2+ 3.Kb1 and White wins. So there castling is an option. But here it isn't, as after 2.0-0-0 Ra1+ 3.Kd2 b2 nothing is gained. A conundrum!

2.0-0-0 (still!) 2.-Ra1+ 3.Kb2! This seems not to work, as Black has checks on the d-file and White must eventually go to the c- or a-file, allowing Black to reposition the rook. Yet, let us see...

3.-R:d1 4.a7 Rd2+ 5.K:b3 Rd3+ 6.Kb4 Rb4+ 7.Kc5! Ra4 8.Kb6 Rb4+!

Exactly! Black has an important check, forcing White to move his king either to the a-file or to c5 which repeats moves. So the tension is high: 9.Ka5 Rb1 10.d6!

Now Black can give checks and reposition his rook, but it will turn out that his king is too far away. The rest is straightforward: 10.-Ra1+ 11.Kb6 Rb1+ 12.Kc5 Rc1+ 13.Kd4(Kb4) Rd1+ 14.Kc3(Ke3) Rc1+ 15.Kd3!(Kb3!) Rc8 16.d7 Ra8 17.Kd4(Ke4) wins. Different routes but all with the same idea were available to the king.

Understanding the versatility of castling, it can be used to ensure a study is correct when the solution does not include castling, but the try would be refuted by it. An example with rather simple play:


Sergey Kryuchkov. Zadachy i etyudi 1927.

White to move and win.

Information about the composer is welcome. The solution is not 1.Kd6? 0-0!, or 1.-f:g6 2.Ke6 0-0!, but rather 1.Be5! B:e5 2.R:e5+ Kf8 3.Re8+! K:e8 4.g7 Rg8 5.f6 wins by zugzwang. Note that the basic idea of the trick 3.-Kg7 4.f6+ might come up in practical play also.

The "anti-castling theme", as I call it, was invented by Josef Moravec in 1931. The study is world-famous as much as the Selezniev study from 1921.


Josef Moravec. Duvtip, 29 April 1931.

White to move and win

Black threatens to castle queenside, so the physical obstruction 1.Bb8! temporarily prevents this. After 1.-d2 2.Bd6 the restriction is removed but the third rank is open: 2.-0-0-0 3.Rc3 mate

A revolutionary idea with difficult sidelines.

Simple castling ideas were shown again by composers, which we will skip, as they are similar to studies already shown here. Instead we pick up again at Herbstman combinding castling with stalemate.


Alexander Herbstman, Shakhmaty July 1937 (correction by Mario Garcia 2012). White to move and draw

1.0-0-0+ K:e5 2.R:g1 B:f3 3.Rg3 Ke4 4.Rg4+ B:g4 stalemate is a forced main variation, but shows for the first time the combination of castling and stalemate. The rook on b2 does not prevent castling, but it does prevent the king from later moving to b1. Similar ideas on the kingside would require attacking h1, which due to attacking the rook would be even more forced.

The original setting of Herbstman's study had bKd6 and no knight on e5. The refutation there is interesting, starting with 1.0-0-0+ Bd4!, so it is also included in the replayable studies below.

With heavy material usage, J. Barendregt in De Schaakwereld, 18 June 1937, forced a stalemate on the kingside after castling. As this is also the first study combining the idea - I don't know which one was published first, it is reproduced here.


J. Barendregt, De Schaakwereld, 18 June 1937.

White to move and draw

1.0-0+ Rf3 2.R:f3+ Q:f3 3.Qg7+ K:g7 4.h6+ forces stalemate but lacks anything else that is interesting.

Multiple castlings were incorporated into endgame studies in 1938 and 1939 (apparently for the 1938-1939 tourney), both by giving White a choice and by having both sides castle. Unfortunately, the study where White castles in both directions contains a destructive dual - it can be replayed below -, so only the other one is reproduced here.


Henri Rinck, Magyar Sakkvilag 1938 (published November 1939), 2nd prize.

White to move and win.

As Black wants to check on the h-file, the study starts with the spectacular pair of moves 1.0-0-0 0-0 and then White's attack will force the triumph of the soon-to-be passed pawn. Just be careful with the move order, as 2.Rh1? Rf6 defends. 2.Bc3 Bb7 3.Rh1 Rf6 4.B:f6 e:f6 5.Rh8+! K:h8 6.e7 wins

It has a tragicomical component how helpless Black is after the second move, so only despair remains.

John Selman showed the first underpromotion combined with castling. This nearly completed a task which we will look at in the future.


John Selman, Tidskrift v. d. KNSB 1938, 2nd commendation.

White to move and win

The play is forced: 1.b7 0-0! 2.c7 Kh7! prepares a stalemate. White can avoid it easily, however the correct pawn must be chosen: 3.b8Q? g1Q+ 4.K:g1 h2+ 5.Kh1 Rf1+ with perpetual check on the f-file, or stalemate. So the winning moves are 3.c8Q R:c8 4.b:c8B! winning as opposed to 4.b:c8N? Kg8 5.Nd6 Kf8 6.Nb5 Ke7 7.Nd4 Kd6 8.Nf3 Ke5 9.K:h3 Kf5 and 10.-Kf4 with positional draw.

It is 1938. The time circle has closed to the study that went viral recently. This shall mark the end of this first part of an article that will in its entirety be very extensive. We already again have incorporated more studies than usual and have only looked at a small bracket of interesting studies with castling. As such, this will have to become an entire (mini-?)series of articles.

Many years ago, when I still was a child, the trainer in my chess club - also named Siegfried - said that undeveloped rooks stand "in the garage". The comparison to cars might not be farfetched, as the rooks cruise over the board quickly. So is castling a kind of pit stop for them? Getting fuel, overhauling the vehicle, tuning the engine? We will continue our driving lesson soon...


Discussion and Feedback

Join the discussion via the comment section or submit your feedback to the editor.


Siegfried (*1986) is a German chess composer and member of the World Federation for Chess Composition, subcommitee for endgame studies. His autobiographical book "Weltenfern" (in English only) can be found on the ARVES website. He presents an interesting endgame study with detailed explanation each month.
Discussion and Feedback Submit your feedback to the editors


Rules for reader comments


Not registered yet? Register

Lovuschka Lovuschka 5/3/2022 06:28
Yes, isn't it annoying how those typos happen?
Here it is EG 26, page 291; not EG 6, page 291.

The comment about Kryuchkov is correct, another typo.
Frits Fritschy Frits Fritschy 5/2/2022 12:38
Sotadas, the retroanalysis can be found in the game window.
Sotadas Sotadas 5/2/2022 05:04
”The retroanalysis for the position is given in detail in the sources”, but I couldn't find the sources; where are they? Thanks.
Frits Fritschy Frits Fritschy 5/1/2022 11:35
According to, the Russian Sergei Petrovic Kryutchkov (mind the spelling) lived from 1907-1942 and composed 19 studies. They can be found at Van der Heijden, but there the spelling is Kryuchkov. 'Kryuckhov' as under the diagram seems to be misspelt.,_%D0%A1%D0%B5%D1%80%D0%B3%D0%B5%D0%B9_%D0%9F%D0%B5%D1%82%D1%80%D0%BE%D0%B2%D0%B8%D1%87 may give some more information (born in Moscow?), but not in a language I have any knowledge of (Czechian?).
Karsten Müller Karsten Müller 5/1/2022 09:20
nieder t-bird: The diagram is right, but "The solution is not 1.g6? 0-0!, or 1.-f:g6 2.Ke6 0-0!, but rather 1.Be5! B:e5 2.R:e5+ Kf8 3.Re8+! K:e8 4.g7 Rg8 5.f6 wins by zugzwang." is wrong and should be " not 1.Kd6?..."
nieder t-bird nieder t-bird 5/1/2022 07:37
The diagramm of the Kryuchkov study is wrong as it already shows the g-pawn on g6.
Lovuschka Lovuschka 5/1/2022 01:12
It seems this has already been corrected? If not, please point out where exactly it is given wrong, so it can be corrected. Thanks for the good perception!
Mamack1 Mamack1 4/30/2022 10:45
The Gorgiev study is White to play and *win*, not draw.