UNDERPROMOTION IN GAMES

In 1912, in The Theory of Pawn Promotion, Alain White wrote that he had 'never heard of a game where victory was won by a promotion to Rook or Bishop'. In 1936 the great endgame connoisseur Harold Lommer wrote something to that effect too. In my own Chess Curiosities (1985) I mentioned the 'extreme rareness of such promotions', and quoted 12 examples from all of chess history, including trivial ones. Ten years later Harold van der Heijden, in his Pawn Promotion, extended the list to 27. We had entered the database era - he had used a 400,000 game one. Over the following two years, in the magazine EBUR, he showed some 15 new cases. My turn again - in the meantime the databases have grown to almost 2,000,000 games, and if I used the same criteria as in 1985, I could show over 60 examples now.
    It's beginning to spoil the fun. We shouldn't know that much. Sometimes, letting my computer hunt in the databases, I feel like a peeping Tom of chess history - as if I were reading a loved one's secret diary.

But even now, Alain White's question pops up in the newsgroups: has ever a game been won by a promotion to Rook or Bishop? Why, yes - maybe it's time to rephrase his remark: has ever a game been drawn by a promotion to Rook or Bishop? Not that I know of, but it could happen - it doesn't take more than 5 minutes to set up a crude idea that might occur in a game. In the diagram on the left, White must promote. 1.g8N and g8R allow Qe5 mate, and 1.g8Q+ K- 2.Qxb8 Qxb8+ is a simple win for Black. But after 1.g8B+ Black must make a King move and White is stalemate.
    Or take the following two examples from endgame studies. Neither are the initial positions; I just want to show the ideas, which are imaginable in games.

 
White to play 
F. Lazard, 1935 

1.d8R! If 1.d8Q, then Bf4 2.Qd2+ Kf3 3.Qxf4+ Kxf4 and Black wins. 1...Bf4 (Bxh2 2.Rd3+!) 2.Rd2 and Black's King cannot move without causing stalemate, e.g. 2...Bg5 3.Rd5 Kf4 4.Rd2 Bh6 5.Rd6 Kg5 6.Rd2 etc. 
 
 
 

 

 
White to play 
V. Vlasenko, 1967 

If 1.h8Q, then Nf6 or Nc5 followed by Nd7 mate. And if 1.h8N Ng5! 2.Nf7+ Nxf7 and Black wins. But: 1.h8B! A brilliant idea: self-paralysis. If now Nf6 or Nc5, it is stalemate. And after moves that relieve the bind, a drawn endgame of 2 minor pieces vs. 3 remains. 

PS 14 January 2000: As reader Gerhard Josten informs me, the tablebases say Black is winning after 1...Bg4 - and that therefore this study is incorrect. A pity for the beautiful idea.

 
IN THE DATABASES TODAY, a few hundred games with Rook and Bishop promotions can be found, but the majority are silly jokes. Sometimes an underpromotion is used to remind the opponent that it's time to resign. Often he does. Sometimes players, especially young amateurs (but even Topalov has done it in a tournament game against Kramnik), will promote to Rook or Bishop when that new piece has to be captured anyway.
   There are also joke promotions that, without having anything to do with stalemate, still have some practical significance; when the choice between capturing the new piece and not capturing it is not obvious. An example is Salnikov - Bezgodov, Pavlodar 1991 (left), where Black played 61...h1B+ Had he played h1Q+ or h1R, White would have had to capture, but now he had to think. A ploy like that, to confuse the opponent, might be useful in time trouble. White did not take the Bishop (the pawn endgame is dead lost) but after 62.Kf2 Bc6 Black won in a few moves anyway.

Bordering  on silliness, or even rudeness, are those underpromotions where the only point is that an immediate promotion to Queen would have been stalemate. A case in point is Ruben - Sultan Khan, Hamburg 1930, (diagram on the left) where Black resigned after 74.f8R, but 74.Kf6 would have been mate in 3. Varga - Ivanka, Hungary 1991 (diagram on the right) is special in that of White's six legal moves, only the promotions to R, B and N are winning. He chose 63.h8R, and Black resigned.

But even when an underpromotion is the only move, it can still be a silly joke - by the opponent. In Vasiukov - Tukmakov, Erevan 1976 (diagram on the left) White, as a way of admitting he was lost, played 59.Rg1+, and resigned after hxg1R. This sort of thing, where one side invites promotion and the only finesse is that promotion to Queen is stalemate, happens quite often, and my sharpened criterion does not include them anymore.

The distinction between the silly and the serious is not completely clear. What to make of Amelung - Erler, Katharina 1893 (diagram on the right)? That position must have happened more often, but probably no one ever thought of what Amelung, an endgame composer, played here. He announced mate in 3 beginning with 1.a8R. Against 1...Ka2 or a2 a promotion to Queen would have worked as well (2.Kc2 and various Queen moves respectively), but 1...Kb1 2.Qxa3 is stalemate. With a Rook, Rxa3 is then possible, and it is mate next move. And in fact, this position, perfectly mirrored, had already appeared as a 'Mate in 3' by A. Barbe, in 1861.

Coming to serious practical underpromotions - over half of them have the same configuration, the one we already saw in Ruben - Sultan Khan: Black's King is at h7, g6 is inaccessible, and White cannot promote to Queen on f8 because that would be stalemate. (And all mirror situations.) The simplest possible case is also the oldest.
    In Von Guretzky Cornitz - Neumann, Berlin 1863 (diagram on the left), there followed 1...c2+ (a2+ works the same way) 2.Nxc2 a2+ 3.Kb2 bxc2 4.Kxa2 and now Black played the only move c1R, and White resigned.

Equally trivial cases, in all of which there were other moves to win are shown in the 5 diagrams below:
 
Black to play 
Simonson -  Em.Lasker  
New York sim 1892 
 
Black to play 
Alapin - Rubinstein 
Vienna 1908 
 
 
White to play 
Jansen - De Greef 
Dieren 1990 
 
 
White to play 
Makogonov - Dubinin 
USSR 1940 
 
Black to play 
Jordan-West 
Australia 1980 
 
In Simonson - Em.Lasker, Black set a little trap with 66...Kb8 If now 67.c7+?? Kc8 draw. And even after 67.Kd7 there was one last trap: Ka8 68.c7 Ka7 After 69.c8R Black resigned. In Alapin - Rubinstein, Black mated in 4 with 75...Nf2+ 76.Bxf2 gxf2 77.Kxh2 f1R 78.Kh3 Rh1 mate. The same mate occurred in Jansen - De Greef where Black resigned after 64.Nc7+ Bxc7 65.bxc7 Ka7 66.c8R This mate has probably occurred a  good many other times. Makogonov-Dubinin (85.Kh1 g2+ 86.Kh2 gxf1R) and Jordan - West (68...f1R) in both of which White resigned after the Rook promotion, are trivial indeed.
    Just a trifle less trivial were:
 
 
White to play 
Steiner - Morton 
USA 1936 
 
White to play 
Roehrl - Wittmann 
Austria 1981 
 
White to play 
Flores - Letelier 
Santiago 1959 
 
White to play 
Malishauskas-Donchenko 
USSR 1988 
In Steiner - Morton, Black resigned after 112.f8R, but f8Q would have won quicker; Re6+ 113.Kd7 and Black has no further check that maintains the stalemate. In Roehrl - Wittmann, 68.Ra4 was met by f1R and White resigned (but f1Q 69.Rxf4+ Ke2 would have been quicker) In Flores - Letelier some other moves would have won as well, but 66.f8R (and Black resigned) was the easiest (66.f8Q Rf1+ 67.Bf4 Rxf4+ and stalemate.) And in Malishauskas - Donchenko, 59.f8R was easiest too; Black resigned. 59.f8Q Rf6+ would of course be stalemate.


Another theme with this configuration, and which is sometimes used in endgame studies, was first seen in Badestein - Otto, Wernigerode 1952 (diagram on the left.) After 66...f1R White resigned; f1Q? and White has a 'Rambling Rook' which will check forever along the 8th rank.
    In contrast, in all of the following examples promotion to Queen would also have worked. But sometimes it would have been tricky, and in most cases the Rook promotion was the most sensible and practical thing to do: why think if you don't have to.

 
Black to play 
Rajna - Lengyel 
Budapest 1982 
 
Black to play 
Jacobowitz - Grosse 
Dresden 1993 
 
Black to play 
Wolf - Hartmann 
Germany 1991 
 
White to play 
Istratescu - Bancod 
Biel 1993  
In Rajna - Lengyel White resigned after 74...f1R. 74...f1Q was also pretty easy: 75.Re6+ Kd2 76.Rd6+ Kc3 77.Rc6+ Qc4 etc. Also in Jacobowitz - Grosse, promotion to Queen (or e2) was possible: 58...f1Q 59.Rc7+ Kb4 60.Rb7+ Qb5 Rxb5+ Kxb5 62.Kxg1 Kc4 and Black wins. But Black played 58...f1R, and White resigned after a few more moves. In Wolf - Hartmann 76...f1Q would have been just as easy as 76...f1R (and White resigned); 77.Rf3+ Qxf3.
 
Black to play 
Dehmelt - Fedorowicz 
Reykjavik 1986 
And in Istratescu - Bancod too, White could have promoted to Queen; 67.f8Q Rd1+ and White runs to b5 and interposes the Queen. But he played 67.f8R, and Black resigned. In Dehmelt - Fedorowicz finally (diagram on the left), Black already had a Queen on f1, and didn't need another one on e1; 77...e1Q 78.Qxg7+ would be stalemate. 77...Qf2+ followed by e1Q would have been good enough, but after 77...e1R 78.Qh4+ Kg8 White resigned.

THERE IS EVEN an elementary endgame that (although grandmasters normally do not play it to the end) must have happened thousands of times, and in which White can sometimes force the win with a promotion to Bishop, and sometimes to Rook; the endgame of Rook + h- and g- (or a- and b-) pawns vs. Rook.
    If the weaker side keeps his King in the corner, as in Dawidow - Czerwonski, Krynica 1997 (diagram to the left), the procedure given by Euwe in his standard work on the endgame is h2 followed by Kh3 (Kg2 Rf2+). One can also begin with g2+ (see below) but it is more fun to add, apart from the Bishop promotion, a little Rook sacrifice: 75...Rf1+ 76.Rxf1 g2+ 77.Kh2 (77.Kg1 Kg3) gxf1B and White resigned. (Of course a Knight promotion would work here too.)
    It is remarkable that in neither of the following four examples, the opportunity of the Rook sacrifice was used.
 
 
White to play 
Beni - Littlewood 
Lucerne 1963
 
Black to play 
Schichtel - Abend 
Germany 1987 
 
Black to play 
Apicella - Bellia 
Vinkovci 1989 
 
White to play 
Kreiman - Salman 
Philadelphia 1992 
In Beni - Littlewood, White played 67.b7+ Ka7 (Kb8 68.Kb6) 68.Rc8 Rxc8 69.bxc8B and Black resigned. In Schichtel - Abend 68...Rf1 would again have been possible, but Black chose 68...g2+ 69.Kh2 Rf1 70.Rxf1 gxf1B 71.Kg1 Kg3 and White resigned. Same thing in Apicella - Bellia: 88...g2+ 89.Kh2 Rf1 90.Rxf1 gxf1B and White resigned. And in the exact mirror of Dawidow - Czerwonski, Kreiman - Salman, White again let the chance of the Rook sacrifice pass with: 62.g7+ Kh7 63.Rf8 Rxf8 64.gxf8B and wins.
 
 
Black to play 
Lacroix - Bolzoni 
Brussels 1986 
  
Black to play 
Fallone - Vranesic 
Havana ol 1966 
 
White to play 
Wicker - Way 
Copenhagen 1987 
If the weaker side does not keep his King in the corner, a position like that of  Lacroix - Bolzoni can arise - this exact position is again treated by Euwe. There followed 76...Rf8+ 77.Kg5 Ra8 Here Euwe gives 78.h7+ Kg7 79.Re7+ Kh8 80.Kh6 and mate in 2. But again, why miss a Bishop promotion: 78.g7 Kh7 79.Rf6 Ra5+ 80.Rf5 Ra8 81.Rf8 Rxf8 82.gxf8B and wins. In Fallone - Vranesic, White's defense on the second rank led, for a change, to a Rook promotion: 75...h2+ (g2 is even easier) 76.Kg2 Rf2+ 77.Rxf2 gxf2 78.Kxh2 f1R (78...Kf3 is a quicker mate) and White resigned. And in Wicker - Way, yet another underpromotion in this endgame is demonstrated after 51.Kc5 Ka7 52.Rc8 Rxc8 53.bxc8R and Black resigned.
 
In another recurring configuration the Black King is on h6, h5 is guarded, and the promotion is on g8 (and mirrors.)
 
White to play 
Sajtar - Benko 
Budapest 1954
 
Black to play 
Garcia - Bellon 
Torre Grande 1990 
 
White to play 
Torre - Gufeld 
Kuala Lumpur 1994 
The position of Sajtar - Benkö can be won in many ways, including by promotion to Knight, but the quickest win, mate in 3, is 86.g8R (Black resigned.) Garcia - Bellon had a finish that has been used in endgame studies: 61...g1R and White resigned. Promotion to bishop would have won too. And in Torre - Gufeld we see two grandmasters playing out the same little joke: 66.g8R and Black resigned.
 
 
White to play 
Stahlberg - Czerniak 
Buenos Aires 1941
 
Black to play 
Chan - Depasquale 
Laoag 1985 
 
Black to play 
Serper - Navrotescu 
Oakham 1988
A few dressed-up examples. In Stahlberg - Czerniak, Black resigned after 51.g8R. Promotion to Queen would work too (there will be an end to the checks of the Rook), but again: why think. In Chan - Depasquale, 1...g1B was the only winning move: 2.Rd8+ Kf7 3.Rh8 Nd5 4.Kg3 Be3 5.Rxh5 Bxf4+ 6.Kf3 Kg7 7.Kf2 Bh6 and White resigned. And in Serper - Navrotescu Black had various ways to win (not 46...g1Q 47.Rh7+ Kxh7 48.Re7+ with a draw, but 46...Ra1 is quicker) but it was of course irresistible to play 46...g1R and create a unique position with five Rooks on the board simultaneously.

We come to more diverse cases. In ascending degree of interest:
 
 
Black to play 
Pinzon - Del Pozo 
Lima 1959 
 
Black to play 
Grishchuk - Hua 
Szeged 1994 
In Pinzon - Del Pozo, after 60...Re8 61.Rb8! Rxb8 the Bishop promotion 62.axb8B (after which Black resigned) was a nice, but in fact a weak move: the Zwischenschach 62.g6+ followed by 63.axb8Q would have won much quicker. But in the largely identical position Grishchuk - Hua (from the Boys under 12 World championship), the underpromotion was necessary: 66...Rf8 67.Rb8 Rxb8 68.axb8B and Black resigned. (Of course in both cases, promotion to Knight would have worked too.)
 
 
White to play
Gluzman - Sabic 
Sydney 1992 

Black had already played on for too long. Not even 91.g8R made this clear; after Rxc6 he played on seven more moves. 91.g8Q was possible; 91...Rd1+ 92.Kc5 and already the good checks are over. 
 
 
 

 
Black to play
 
Ashley - Bezold 
Bermuda 1997  

Just about every legal move would win here, and I would have expressed my dismay over White's for playing on with simply 74...f1Q 75.Re3+ Kc2 76.Rc3+ bxc3 and mate next move. Black chose 74...f1R and White resigned. 
 
 
 

 
Black to play
 

Ahlander - Krasenkov 
Malmö 1994  

White resigned after 50...g1R; 50...g1Q 51.Qxd7+ would have been stalemate. 

 
 
 
 
 

 
White to play
 

Tomic - Winzbeck 
Dortmund 1993  

White decided the game with the nice little trick 43.Rd8 Rxd8 44.cxd8B and Black resigned. 

 
 
 
 
 

 
Black to play
Nedela - Schmaltz 
Stockerau 1992  

After 54...b1R (54...b1Q? 55.Rb6+ is a draw) White resigned, but 54...Rc4+ and 55...b1Q would have won more easily; Rh7+ is met by Rc7. 
 
 
 
 

 
Black to play
Weitthoff - Maki 
USA 1988  

58...b1R and White resigned two moves later; 58...b1Q 59.Rxf5+ would have been a draw.
(PS May 2016): The rook promotion is the simplest win, but as Ricky Demer shows, 58...b1Q is also winning: 59.Rxf5+ Kh6 and now 60.Rh5+ (60.Rf6+ Qg6) Kg7 61.Rh7+ Kf6 and regardless of White's checks, the Black King goes to a1 and the Queen can interpose on a2.
 
 
 
 
 

 
White to play
Boniface - Pugh 
Bristol 1995  

A study-like intermezzo: after 63.Nd1, Black had to play 63...b1B, because 63...b1Q or b1R 64.Nc3+ Bxc3 would be stalemate. Although the position was then technically won for Black, he did not manage to win, and a draw was agreed at move 92. 
 
 
 

 
White to play
Kholmov - Ehlvest 
Wolgodonsk 1983  

After 72.Ra1, there followed  h1B! (72...h1Q  73.Ra8+ Qxa8 would have been stalemate). Now it was easy; after 73.Rf1 Rh8 74.Rf7 Re8 75.Kc5 e5 76.Kd6 Bb7 White resigned. 
 
 
 
 

 
White to play
Bremel - Kertis 
Budapest 1948  

Again a finish that is doubtless also the finish of some endgame studies. 1.a8R+! and Black (probably; I do not have the score of the whole game) resigned. 1.a8Q+ would have been stalemate after Kb4 2.Qb7+ Kc3 3.Qxb1 
 
 
 

 
White to play
Reshko - Kaminsky 
Leningrad 1972  

61.a8Q (or a8R) Qf7+ 62.Qxf7 would be stalemate. After 61.a8N Qa7 it is very hard to see how White can make progress. What else? 61.Qb8 or 61.g5 allow mate in 1. But 61.a8B, which White played, is a fairly easy win. There followed: 61...Qb3 62.Qd7 Qg8 63.Bd5 Qf8 64.Bf7 Kh8 65.Qe8 Qxe8 66.Bxe8 Kh7 67.Bf7 Kh8 68.Kg6 h5 69.Kxh5 Kh7 70.Be8 Kg8 71.Kg6 and Black resigned.

Sokolsky - Ravinsky, USSR 1938 (or 1933) (see diagram below) is still by far the most interesting, complicated and mysterious of practical underpromotions. For one thing, I have never been able to find the entire game, nor the exact year, nor even its conclusion. The move number is the closest I have been able to come. The position is always dispatched with as '66.a8B and White won.' If he couldn't have won in another way, how he won, or why he should win at all, is never mentioned.
    Let's first look at some other tries.
 
  
White to play 
  To start with, l.a8=Q(R) Rc2+ 2.Ka1 Rc1+ etc., and the checks can never be stopped.
    Or 1.Rb8 Rc2+ (Rc8? 2.Ka3! and Black is in zugzwang: 2...Rd8 3.a8Q Rxb8 4.Qe4 winning or 2...Rxb8 3.axb8N! and Black gets mated) 2.Ka3 Ra2+ 3.Kb4 Ra4+ 4.Kc5 Ra5+ 5.Kd6 Ra6+ 6.Ke7 Rxa7+ 7.Kd8 Rd7+ 8.Kc8 Rc7+ 9.Kd8 Rd7+ 10.Kxe8 Rd8+ and draw.
    A good try seems 1.a8N because the Knight could be negotiated to e7 to give a deadly check, but again Black uses the immured stalemate position of his King: 1...Rc8! 2.Ra6 (2.Nc7 Rxc7 3.Rb8 Rc2+) Rxa8 3.Rxa8 stalemate.
    Vacating b6 for a future Knight, for example with 1.Rb7, is refuted by Rc8 and there will be no Knight.
    All Black has to watch out for in these variations is the zugzwang mentioned above, e.g. 1.Rb4 Rc8 2.Rb7 Ra8?? 3.Rb8 Rxb8 4.axb8=N and mate. But this is easily avoided.
    Well then, 1.a8B. For a long time, the only analysis I saw on this position was in my own book in Dutch, Schaakkuriosa (1974), where it says Black now draws by chasing the Bishop: 1...Rc8 2.Be4 Bc6 3.Bb1 (the winning chances after 3.Bxh7+ are illusory) Be4 4.Ba2 Bb1! and the Rook will have its perpetual.

But the Dutch master and endgame composer, Cortlever, refuted this strange Bishop chase.
 
White to play 
After 1.a8B Rc8 2.Be4 Bc6 he continues with 3.Bxc6. Now Black must threaten his stalemate-perpetual. 3...Ra8 (3...Rd8 4.Bd5 and 5.Bxf7+ or 3... Re8 4.Be4 and 5.Bxh7+) 4.Ba4 Re8 (4...Rc8 5. Re6! followed by 7.Be8) 5.Ka3! and Black must give up the e-line, enabling Re6 after all: 5... Ra8 (see small diagram) 6.Re6! fxe6 (or 6... R8- 7.Be8) 7.Kb4 Kf7 (the Rook had to guard the eighth rank against 8.Be8) 8.Bc6 Rb8+ 9.Kc5 and  Black is helpless against the threat of 10.Be4 folowed by Bxh7 (9...Rxb3 10.Be8+)

The question remains as to how White wins if Black does not chase the Bishop. We return to the position after l.a8=B (see diagram).
 
 
Black to play 
    There are now four important possibilities:
    (a) 1...Rh5 2.Be4 Rxh6 3.Ka3 (zugzwang; but 3.Rb8 Rxf6 4.Rxe8+ etc. may also win) Rh5 4.Rb8 Re5 5.Bc6 etc. (Cortlever)
    (b) 1...Rc7 Black tries to maintain the seventh rank. 2.Bd5 Bd7 (2 ...Rd7 3.Rb8 Rd8 4.Bxf7+) 3.Rd6 and Black is paralysed; if 3...Rc- 4.Bxf7+, and if 3...Be8 4.Bxf7+ Kxf7 5.Rd8 (this would also have followed after other captures), and Black has no defense against 6.Rxe8 and/or g8Q+.
    (c) 1... Rc8 2.Bd5 Knowing, by Cortlever's proof, that it doesn't have to let itself be chased, the Bishop chooses the best square. 2...Rd8 (2... Rc- 3.Bxf7+ etc.; 2...Bb5 or Ba4 3.Bxf7+; 2...Bd7 3.Rb7 threatening 4.Bxf7+; now 3...Rd8 4.Bc4 Be8 5.Re7 as well as 3...Be8 4.Re7 lead into the main variation) 3.Bc4 Rc8 (Black must keep to the eighth rank, e. g. 3...Rd2 + 4.Kc3 Rf2 5.Bxf7+ Kxf7 6.Re6 Rg2 7.Re7+ etc.; or 3...Bd7 4.Rd6 Be8 or R- 5.Bxf7+) 4.Rb7 Rd8 (4...Rc- 5.Bxf7+) 5.Re7 and now Black can only play his Rook on the eighth rank where it will be collected by the white King, e.g. 5...Rc8 6.Ka3 Rb8 7.Bd5 Rd8 8.Kb4 Rb8+ 9.Ka5 Rd8 10.Kb6 Rc8 11.Kb7 Rd8 12.Kc7 Rd7+ 13.Rxd7 Bxd7 14.Bxf7+ etc.
    (d) 1...Re5 Preventing the strong move Bd5. Some very queer possibilities now arise.
        (d1) 2.Rd6(?) Bc6! 3.Rd8+ (3.Bb7 Re8! 4.Ba6 Bb5 5.Bc8 Bd7 draw) 3...Be8! (3...Re8 4.Rxe8+ Bxe8 5.Be4 Bc6 6.Bxh7+ and White wins) 4.Rd2 Bc6 5.Bb7 Re8 6.Ba6 Bb5 7.Bxb5 Re2 draw.
        (d2) 2.Bf3! and now the Bishop will reach the vital c4-f7 diagonal without Black obtaining his perpetual; e.g. 2...Bc6 3.Bd5! or 2...Bb5 3.Bd5! or 2...Rf5 3.Be4 (3.Rb8 Re5) 3... Re5 4.Bd3 Bb5 5.Bc4 and White wins by bringing his Rook to e7 (5... Re8 or Bc6 6.Bxf7+).

In summary: 66.a8=B is winning, and is White's only winning move in Sokolsky - Ravinsky.
 
IN CHESS CURIOSITIES, in 1985, I gave 12 examples of underpromotion in games, of which 5 were 'above triviality as well as of uncontested authenticity.' In this survey, I gave 47 examples, of which perhaps half could be called interesting.
    We're waiting for that drawing underpromotion now.
 
(c) Tim Krabbé 1999
With many thanks to Harold van der Heijden.


Comments and suggestions
Top of the page  |  Main chess page  |  Main page