All rook endings are drawn...not!
If I had a quarter for every time I'd heard or read the phrase "all rook endings are drawn" (henceforth AREAD), I wouldn't be rich, but I could enjoy a nice vacation from the proceeds. This saying of the great player Siegbert Tarrasch, much maligned for his dogmatic formulations of useful rules of thumb, is obviously false if taken literally. Even if we take it, as suggested in the previous sentence, as a rule of thumb, it still doesn't fare so well. There is a kernel of truth buried under the false and misleading aspects of the cliche, however, and we'll try to separate the wheat from the chaff in this post.
Against AREAD, not only is it literally false (obviously), it's not even true that most rook endings are drawn. Worse still, according to the database-derived statistics given in Andrei Volokitin and Vladimir Grabinsky's Perfect Your Chess (p. 38), bishop endings, queen endings, and knight endings are all more drawish than rook endings (the percentages are 47%, 43%, 40% and 38%, respectively); only pure king and pawn endings are less drawish, with a figure of 27%. Given these numbers, AREAD is not only an exaggeration; it's a risible falsehood. Is there any way to salvage Herr Doktor Tarrasch's reputation, at least with respect to this adage?
I believe there is. What makes rook endings seem more drawish than they really are is the wide range of cases where a one- or even a two-pawn advantage will not suffice for a win. Let's consider some examples.
First, of course, there are the simplest cases of rook and pawn vs. rook. These are almost always drawn if the defending king is in front of the pawn (the exceptions are well-known, so I won't bother detailing them here). Of course this is typical of the other endgames already mentioned, so this isn't too illuminating. There are more interesting draws, though - sometimes even if the king is cut off from the queening square or even the queening file the defender can often draw if his rook can give checks from the long side.
There are plenty of cases with multiple pawns, too. If the pawns are all on the same side, being a pawn down is often tolerable - many such rook endings are not just drawn, they're easily drawn. This is not the case for knight endings, where 4 vs. 3 on the same side is possibly a forced win, even when the weaker side has an ideal pawn formation (e.g. Black with pawns on f7, g6 and h5). It goes almost without saying that a 4 vs. 3 advantage in a pure pawn ending is an incredibly easy win in almost every instance.
Furthermore, in rook endings with pawns on both sides (i.e. the kingside and the queenside, not (just) Black and White), if the defender's rook is very active (e.g. on its seventh rank), a draw can often be achieved, as one of two scenarios will often arise: (1) both sides will go on a pawn-eating spree, typically leading to one of the draws mentioned above, or (2) the side with the extra pawn will be tied down to its protection, and unable to make further progress.
There are cases when two extra pawns are not enough, many of which are quite normal and not something out of an endgame study competition. The best-known case is that of rook plus f- and h-pawn (or a- and c-pawn) against rook. As long as the defender's king is in front of the pawns and not cut off on the back rank, this ending generally should be drawn.
Draws with two extra pawns is a common occurrence when one of the pawns is an a- or h-pawn. Here's a typical example: White: Kb5, Ra8, pa7, pg6; Black: Kg7, Ra1. White can make no progress, because the g-pawn is stuck and the rook can't move without losing his only trump, the a-pawn. White can try 1.Kb6, of course, hoping to move his rook, but the Black rook checks the king away from the pawn and then returns to the a-file (1...Rb1+ 2.Ka6 Ra1+ 3.Kb6 Rb1+ 4.Kc6 Ra1 etc.).
Another, perhaps more surprising way to draw despite a two pawn disadvantage arises when the pawns are connected but blockaded. This pattern is not always a draw, but it is (a draw) often enough for the strong side to be wary of allowing such a blockade. Here's a sample draw: White: Kf4, Rb4, pg5, ph6; Black: Kg6, Ra6 (Kling & Horwitz, 1851). It's a draw no matter whose move it is, e.g. 1.Rd4 Rb6 2.Rd8 Rb4+ 3.Ke5 Rb7 4.Rg8+ Kh7 5.Rd8 Kg6! 6.Kf4 Rb4+ and White isn't making any progress.
In addition to the drawing ideas given above, there are other, more particular cases less susceptible to useful generalization - the materially weaker side has a dangerous passed pawn, or enjoys attacking prospects that might result in perpetual check, or can utilize stalemate tricks, etc. The big point is that there are many typical situations where a material advantage in a rook ending is insufficient to win, and it is in this sense (and this sense alone) that we can extract some genuine wisdom from the poorly-phrased AREAD. (A challenge for the would-be Tarrasches - or better, Tartakowers (who may actually be the culprit behind AREAD) - out there: can you think of a new slogan in place of AREAD that's short, catchy, and gets the main point right?)
For further research, I've included some of the material above - both directly and as examples of the principles discussed here - at this link. Readers might also want to check out this typically fine article by Karsten Müller on the Chess Cafe website (this duplicates and extends several of the examples from my game list, as I noticed a little after the fact), this Wikipedia article on rook endings, and on a very practical level, this old post from my previous blog.
(show)
Against AREAD, not only is it literally false (obviously), it's not even true that most rook endings are drawn. Worse still, according to the database-derived statistics given in Andrei Volokitin and Vladimir Grabinsky's Perfect Your Chess (p. 38), bishop endings, queen endings, and knight endings are all more drawish than rook endings (the percentages are 47%, 43%, 40% and 38%, respectively); only pure king and pawn endings are less drawish, with a figure of 27%. Given these numbers, AREAD is not only an exaggeration; it's a risible falsehood. Is there any way to salvage Herr Doktor Tarrasch's reputation, at least with respect to this adage?
I believe there is. What makes rook endings seem more drawish than they really are is the wide range of cases where a one- or even a two-pawn advantage will not suffice for a win. Let's consider some examples.
First, of course, there are the simplest cases of rook and pawn vs. rook. These are almost always drawn if the defending king is in front of the pawn (the exceptions are well-known, so I won't bother detailing them here). Of course this is typical of the other endgames already mentioned, so this isn't too illuminating. There are more interesting draws, though - sometimes even if the king is cut off from the queening square or even the queening file the defender can often draw if his rook can give checks from the long side.
There are plenty of cases with multiple pawns, too. If the pawns are all on the same side, being a pawn down is often tolerable - many such rook endings are not just drawn, they're easily drawn. This is not the case for knight endings, where 4 vs. 3 on the same side is possibly a forced win, even when the weaker side has an ideal pawn formation (e.g. Black with pawns on f7, g6 and h5). It goes almost without saying that a 4 vs. 3 advantage in a pure pawn ending is an incredibly easy win in almost every instance.
Furthermore, in rook endings with pawns on both sides (i.e. the kingside and the queenside, not (just) Black and White), if the defender's rook is very active (e.g. on its seventh rank), a draw can often be achieved, as one of two scenarios will often arise: (1) both sides will go on a pawn-eating spree, typically leading to one of the draws mentioned above, or (2) the side with the extra pawn will be tied down to its protection, and unable to make further progress.
There are cases when two extra pawns are not enough, many of which are quite normal and not something out of an endgame study competition. The best-known case is that of rook plus f- and h-pawn (or a- and c-pawn) against rook. As long as the defender's king is in front of the pawns and not cut off on the back rank, this ending generally should be drawn.
Draws with two extra pawns is a common occurrence when one of the pawns is an a- or h-pawn. Here's a typical example: White: Kb5, Ra8, pa7, pg6; Black: Kg7, Ra1. White can make no progress, because the g-pawn is stuck and the rook can't move without losing his only trump, the a-pawn. White can try 1.Kb6, of course, hoping to move his rook, but the Black rook checks the king away from the pawn and then returns to the a-file (1...Rb1+ 2.Ka6 Ra1+ 3.Kb6 Rb1+ 4.Kc6 Ra1 etc.).
Another, perhaps more surprising way to draw despite a two pawn disadvantage arises when the pawns are connected but blockaded. This pattern is not always a draw, but it is (a draw) often enough for the strong side to be wary of allowing such a blockade. Here's a sample draw: White: Kf4, Rb4, pg5, ph6; Black: Kg6, Ra6 (Kling & Horwitz, 1851). It's a draw no matter whose move it is, e.g. 1.Rd4 Rb6 2.Rd8 Rb4+ 3.Ke5 Rb7 4.Rg8+ Kh7 5.Rd8 Kg6! 6.Kf4 Rb4+ and White isn't making any progress.
In addition to the drawing ideas given above, there are other, more particular cases less susceptible to useful generalization - the materially weaker side has a dangerous passed pawn, or enjoys attacking prospects that might result in perpetual check, or can utilize stalemate tricks, etc. The big point is that there are many typical situations where a material advantage in a rook ending is insufficient to win, and it is in this sense (and this sense alone) that we can extract some genuine wisdom from the poorly-phrased AREAD. (A challenge for the would-be Tarrasches - or better, Tartakowers (who may actually be the culprit behind AREAD) - out there: can you think of a new slogan in place of AREAD that's short, catchy, and gets the main point right?)
For further research, I've included some of the material above - both directly and as examples of the principles discussed here - at this link. Readers might also want to check out this typically fine article by Karsten Müller on the Chess Cafe website (this duplicates and extends several of the examples from my game list, as I noticed a little after the fact), this Wikipedia article on rook endings, and on a very practical level, this old post from my previous blog.
(hide)
