Weekly Problems 2015-A

The out-of-play position of the white queen makes the key easier to find: 1.Qc3! (waiting). Now the queen delivers seven mates, all in response to black unguards: 1…b4 2.Qc4, 1…c4 2.Qd4, 1…g3 2.Qf3, 1…h3 2.Qg3, 1…Bd2 2.Qxd2, 1…Be3 2.Qxe3, and 1…Rf6/Re5 2.Qe5. The black rook sets off three other variations: 1…Rg5 2.Bxg5, 1…Rxh5 2.Sxh5, and 1…Rxd5 2.Sxd5. Lastly, 1…Sg8 2.Rxf5 and 1…f6 2.Se6.

Nigel Nettheim: The b6-knight has to take part, which can only be by 1…Rxd5 2.Sxd5, so White must guard e5. Hence 1.Qc3! which also provides for 1…b4/c4/g3/h3, as well as for 1…Re5/Rf6 and 1…Be3. The key is perhaps not the main feature, but the play of the queen against the two pairs of black pawns shows neat symmetry.
Andy Sag: I enjoyed solving Stephen's problem, a waiter with twelve variations including seven queen mates. Why not add a white rook on a1 and make it thirteen, including eight queen mates!? (1…bxa1=Q 2.Qxc1.)

By promoting the three pawns and using the new pieces to self-block, Black is able to arrange various mating configurations, e.g. c8K, b7S, c7R and d7B for Ra8 mate, or h4K, g3S, g4R and g5R for Rh2 mate. These matrices take more than the required 19 moves to set up, however, and the solution sees Black using two promoted pieces to self-block and the third to interpose on the a-file, allowing the king to be mated on a8. 1.h5 2.h4 3.h3 4.h2 5.h1=S 6.Sg3 7.Sxf5 8.Sd6 9.f5 10.f4 11.f3 12.f2 13.f1=B 14.g1=R 15.Rg8 16.Ba6 17.Ka8 18.Sb7 19.Rb8 for Rxa6. If the h6-pawn were to start on h7, there would be a cook: 1-6.Kh2 7.Kg1 8.Kf1 9.g1=R 10.Rg2 11.Kg1 12.Kh2 13-18.Kg8 19.Rg7 for Ra8.

Nigel Nettheim: The attempt with g1=S and h1=R takes one move too long, because the knight’s path is then not minimal. Instead, h1=S gives the knight its most direct path, via the pawn placed on f5 for that purpose. The elegance of this problem impressed me very much when it first appeared, and induced me to try my hand at composing a few similar ones.
Andy Sag: Nice one involving three different sub-promotions and avoidance of checking or moving into check to ensure only one possible solution.

A good withdrawal key, 1.Qa1! (waiting) completes the block. The black queen cannot maintain its focus on two corner squares: 1…Q~rank 2.Qh1 and 1…Q~file 2.Qa8. The g7-knight permits the same queen mates after the self-interferences, 1…Sh5 2.Qh1 and 1…Se8 2.Qa8, while its only other move self-blocks: 1…Sxe6 2.Bc6. Correction play is shown by the black bishop; the random 1…B~ allows 2.Qe5 – a queen mate disabled by three specific bishop moves, which yield 1…Bc7 2.Sxc7, 1…Bxf4 2.Sxf4, and 1…Bb8 2.Qa8. The white queen does more work with 1…Sf~ 2.Qd4 and 1…c3 2.Qa2. Lastly, 1…exd2 2.e4. The e7-pawn has been added to the original position, to prevent the unfortunate duals that follow 1…Be7 and 1…Bf8.

Nigel Nettheim: The key 1.Qa1! is thematic, for the main interest lies in the queens facing off from opposite corners. It suggests the title “She Stoops To Conquer”.

The aggressive key 1.Qc5! is fairly obvious, as White must deal with the two unprovided flights on b2 and c2. However, it’s surprising that no threat is generated by the key. That means Black’s seemingly nondescript pawn moves actually entail specific weaknesses that White will exploit. 1…c6/cxd6 unguards b6: 2.Qxd4+ Kxb4 3.Qb6, or 2…Kc2 3.Qd1. 1…a5 is a distant self-block: 2.Qxd4+ Kxb4 3.c5, or 2…Kc2 3.Qd1. When Black takes the flights, it’s a pity that White answers with the same queen move we have already seen: 1…Kc2 2.Qxd4 (threat: 3.Qd1) Sc3/Sd2 3.Qc3, and 1…Kb2 2.Qxd4+ Sc3 3.Qxc3, or 2…Kc2/Kc1 3.Qd1. A good castling variation completes the play: 1…Sd2 2.0-0-0 (3.Bxd4) Sf3 3.Sxe4, or 2…Sxb3+ 3.Rxb3.

The splendid key 1.Rf1! yields a flight on g6 and abandons the prominent set variation, 1…Sg5+ 2.Rxg5. The threat is 2.Qb1. Now 1…Kg6+ is answered by the cross-check 2.f5 – also a pin-mate. The set play is changed to 1…Sg5+ 2.fxg5, and 1…Se5 2.fxe5 sees the R + P battery firing again. The c8-knight provides the by-play with 1…Sxd6 2.Sxd6 and 1…Sf6 2.Se7. The white queen is a little underused, though it controls g5 and h6 in the flight-taking variation.

Nigel Nettheim: The fine key allows the impressive cross-check variation; at the same time, it changes the mate after 1...Sg5+. The a5-bishop efficiently counters 1.Bxf7? (2.Rg5/Bxe6) with 1…Bd2! and 1.Rxg8? (2.Se7) with 1…Bd8/Bd2! The low award in the 1913 competition may be explained by the strong international field.

After the key 1.Bg4!, White threatens 2.Qf5. Black commits two self-block errors with 1…Qd5 2.Qxf4 and 1…Sd4 2.Sf2. A couple of self-interferences occurs in 1…Be6 2.Qe5 and 1…Sd6 2.Sc5. Finally, Black unguards twice with 1…Rxg5/Rf6 2.Sf6 and 1…Re5 2.Qxe5, though the latter variation repeats the queen mate seen after 1…Be6. A cleanly constructed position with no white pawns; however, its economy could be improved: shift the g8-knight to h7, and replace the b4-knight and the g6-rook with black pawns on b4 and g7 (1…g6 2.Sf6).

Brian Stephenson: 1...Rxg5, giving a flight, is not set with a mate and it appears clear that 2.Sf6 is the intended mate after that, so the key, guarding f5 to provide just that mate, is soon found. There are two self-blocks (one with white interference), two black interferences and two unguards. Not richly thematic, but very pleasant.

To force Black to mate, a promising plan is to place the white king on a1 for a deflection finish, Bb2+ Bxb2 mate. The scheme requires White to promote to a queen for controlling the black king’s flights on c2, d2, and d1 from d3 (not e2 as the queen would stop the mate), but this takes too long, e.g. 1.e4 2.e5 3.e6 4.Kxb3 5.Ka2 6.Ka1 7.e7 8.e8=Q 9.Qe4 10.b4 11.Bc3 12.d4 13.Qd3 14.Bb2+ Bxb2. In the solution, the white king surprisingly goes to c3 instead, though again the aim is to play Bb2+ (from a1) Bxb2 mate. Now White must guard the king’s flights on b1 and d1, and the quickest way involves a nice underpromotion. 1.e4 2.e5 3.e6 4.Kxb3 5.Kc4 6.b4 7.Ba1 8.Kc3 9.e7 10.e8=B 11.Ba4 12.Bc2 13.Bb2+ Bxb2.

Andy Sag: First tried with king to a1, then c3 and I kept trying to get a promoted queen to d3 to cover b1 and d1 but not b2 or c1, but this could only be done in 14 moves. Then realised that a bishop on c2 would do the trick.

The thematic key 1.Se6! (threat: 2.Qa8) unpins the e5-bishop and grants a flight on d5. The freed bishop defends by vacating e5, and it produces a remarkable number of dual-free variations. In 1…Bd4+/Bc3 2.Sed4, 1…Bf4 2.Sxf4, and 1…Bc7/Bb8 2.Sc7 the direct Q + S battery mates are assisted by the indirect B + S battery which opens to attack d5, while in 1…Bxg3 2.Sxg3 and 1…Bd6 2.Sxd6 it is the R + S indirect battery on the rank that is used to regain control of d5. White employs the B + S battery two more times, once directly if Black takes the flight – 1…Kd5 2.Sg5, and once indirectly with 1…Rxc5 2.Sxc5.

Nigel Nettheim: It seemed likely that 1…Bd4+ would be allowed, but the threat 2.Qa8 was well-hidden, and the e5-bishop moves are handled with very good variety. The b2-knight is needed to counter 1.Sb7, and the c2-rook could not be replaced by a black pawn because of the cook 1.Qe6.

The waiting key 1.Sg5! gives the black king an extra flight on e5. Now the three available moves of the king form a Y-pattern and they bring about different mates – 1…Ke5 2.Sf3, 1…Kc5 2.Se6, and 1…Kxd3 2.Se3 – so the Y-flights theme is displayed. The three mates are also all models. If the white rook were to start on d7 instead, a valuable try would be added: 1.Sd8? Kxd3! with a changed mate after 1…Ke5 2.Sc6.

Nigel Nettheim: This shows some of the powers of the knights in a nice and simple way.

The thematic try 1.Sd5? (threats: 2.Sxc3/Sc7) sacrifices the knight to three pieces and unpins the black queen: 1…Qxd5/Qb8+ 2.Qb8, 1…Bxd5 2.Rb6, and 1…Sxd5 2.c4. But 1…Sf5! (ensuring that the queen stays unpinned) refutes, since the try-piece on d5 has prevented 2.Bc6. The key 1.Sf5! (2.Sd6) again shows a triple sacrifice of the knight and frees the queen: 1…Qxf5/Qb8+ 2.Qb8, 1…Bxf5/Bd5 2.Rb6, and 1…Sxf5 2.Bc6. The c3-knight causes two self-interferences: 1…Sd5 2.c4 and 1…Se4 2.Sxd4. The latter variation makes clear the advantage of the key over a second try, 1.Sc8? Se4!

Nigel Nettheim: It seemed likely that the queen would be unpinned, after one had noticed 1…Qb8+ 2.Qxb8. The main choice is between 1.Sd5 and 1.Sf5, but 1.Sd5? Sf5! no longer allows 2.Bc6. The g7-pawn avoids letting 1…Qxg5+ make the key obvious. The unpinning key shows just a little similarity to No.222 Hawes & Ravenscroft.
Jacob Hoover: All set mates are preserved after the key, plus: 1…Qb8+ 2.Qxb8 and 1…Se4 2.Sxd4.

The diagram position is solved by 1.Sd8! (waiting), with two variations: 1…Kb5 2.Kc3 Kc5 3.Ra5 (ideal mate) and 1…Kc5 2.Rb3 Kd5 3.Rb5. An equivalent solution is not possible in position (b), but now 1.Rb6! (waiting) works, followed by 1…Kc4 2.Sa7 Kc5 3.Rc6. Simple play but the twinning is clever and only five pieces are needed.

Nigel Nettheim: The twin removes one square (d8) from the reach of the b7-knight, compensating for that with another square (a7). Each part is nice, although no theme is apparent.

The superb key 1.Sc5! (threat: 2.Sxe6) concedes two flights to the black king by cutting off the c7-queen and b5-rook. A battery mate follows one flight-move – 1…Kc3 2.Sxb3, but not the other – 1…Kxe5 2.Qg7. White’s first move is also an active sacrifice, which generates 1…dxc5 2.Qxc5. The black queen has two defences that allow the e3-rook to gain control of the c3-flight: 1…Qc4 2.Sc6 and 1…Qxb5 2.Qxd6.

Nigel Nettheim: The play is excellent, as well as the try 1.Rc5? (threats: 2.Qxd6/Sc6) Qa6! However, the a1-knight and b3-pawn can be removed, adding the variation 1...Qb3 2.Sxb3. Also, the b7-knight might better be placed on d7 white knight [see Diagram 226v below].

In the set play, a random move of the d6-knight unguards b5 and opens the white queen’s line to c5, enabling 2.Sb5. The correction 1…Sxe4 allows 2.Qxe4. Moves of the e8-knight are answered by 2.Qxf6, though there’s a dual with 1…Sg7 2.Qxf6/Qxd6. White cannot preserve all of these variations with a simple waiting move, e.g. 1.Qf8? Sxe4!, and the key is 1.Qd8! (waiting). Now the d7-pawn becomes pinned after any d6-knight move: 1…Sd~ 2.S7e6. Another pin-mate follows the correction, 1…Sxe4 2.S5e6. Lastly, 1…Se~ permits 2.Qxf6 with no duals. The keys in mutates are typically perfunctory, but here the self-confining queen move is a delight.

Jacob Hoover: A complete block, in which the key changes the mates for the moves of the d6-knight. A very clever mutate that makes use of the half-pin tactic.
Nigel Nettheim: Among the many fine details, I appreciated the use of the top and left board-edges to limit the range of motion of several pieces.

White must provide for 1…Kb8 when the king threatens to escape to the c-file, and the only way is 1.Rf1! (waiting). Now after 1…Kb8, which self-pins the e5-pawn, White has 2.Rc1 forcing 2…Ka8 3.Rc8. If 1…e4+, White avoids the stalemate-inducing 2.Kxe4? or 2.Ke3?, and instead plays 2.Kg2 to prepare for a battery mate, 2…e3 3.Kg1. The composer was an over-the-board IM who won the Australian Chess Championship four times in the 1940s and 1950s.

Jacob Hoover: A great key 1.Rf1! preserves the set continuation after 1…e4+ and meets 1…Kb8 with 2.Rc1, which keeps the king from moving further out.
Nigel Nettheim: It is nice that in the main variation, 1…e4+ 2.Kg2, all white moves are retreating. The h3-pawn forces the final retreat. Could this be called the “platypus” theme, after one of the shyest and most retiring of animals?

The directmate part of the solution is 1.exd6 e.p.! e5 2.Qa8. The en passant key is legal because we can prove by retro-analysis that Black’s last move could only have been …d7-d5. Black couldn’t have just played …Kg8-h8 since on g8 the king would have been in an impossible check from three white pieces. Black also couldn’t have made a pawn capture on d5 or e6 as the last move. This is because White is missing only the two bishops, and the black-squared one couldn’t have reached either of the two white capture squares, while the white-squared one was trapped on its home square f1 by the pawns on e2 and g2 and so couldn’t have been captured elsewhere. One other potential last move, …d6-d5, is ruled out because it would mean White was in check during Black’s turn. That leaves …d7-d5 as Black’s only possible last move.

This popular two-mover with a difference was solved by Nigel Nettheim, Andy Sag, Jacob Hoover, and Dennis Hale.

All possible black moves have been provided with set mates: 1…Re4 2.Sd8, 1…Qxg6 2.Rxg6, 1…f4 2.Rxe5, 1…Bc5 2.Sxc5, 1…Bd4 2.Sxd4, and 1…Ba5 2.Sc5/Sd4. White cannot maintain the block with any waiting move, however, e.g. 1.Rh6? Qxh6!, 1.Rh5? Qxg6!, 1.Rg1? f4!, or 1.Bg3? h2! The flight-giving key 1.Rxf5! entails a threat, 2.Rxe5. Two changed mates occur with 1…Re4 2.Qxd5 and 1…Qxg6 2.Bd7. Taking the flight gives 1…Kxf5 2.Qg4, while 1…Rxf5 is followed by a transferred mate, 2.Sd8. One more variation is the unchanged 1…Bd4 2.Sxd4. A great example of the less common block-threat type.

Nigel Nettheim: Everything seems excellent, including the key and especially the changed mate 1…Qxg6 2.Bd7.
Jacob Hoover: The surprising key grants a capture-flight on f5. An interesting block-threat with a very attractive pin-mate theme.

(a) No. Black is mated by the bishop, but the piece couldn’t have just played to h1. The only way White could have given the check was by playing Pg2xh3; however, with a white pawn on g2, the white bishop could never have reached h1. That means White has no possible last move in the diagram and the position is illegal.

(b) Yes. White’s last move must have been Rb7xh7, capturing a black piece (not a pawn). This black piece had just played to h7, so Black was not at risk of having no possible last move. Therefore the position is legal.

(c) No. White’s only potential last move was Pe4-e5, but Black had no legal move prior to that. Black couldn’t have played Ka7-a8, because on a7 the king would have been in an impossible check from the b6-pawn and h7-queen.

(d) Yes. The last move wasn’t Pd5-d6 (for reasons similar to those in part (c)), but White could have mated with an en passant capture on d6! The following retraction sequence demonstrates how the position could have arisen: -1.Pe5xd6 e.p. Pd7-d5 2.Pe4-e5+ Ka7-a8 3.Pb5-b6+ (or Pxb6+). Note how once the uncaptured black pawn has retracted to d7, it shuts off the white queen, thereby allowing the black king to retract to a7, where the piece is in check from the white pawn only.

Composer: Rather off-beat – a black Rex Solus retro with multiple parts, in miniature.
George Meldrum: This lightweight setting gives an instant appeal and the first three tasks pass the eye with an enjoyable smile only to conceal the complexity that is yet to yield in the last task that is really a blast.
Nigel Nettheim: Clever twins.
Dennis Hale: Bravo to Ian! This is a fine retro twin incorporating a range of clever ideas.

The highly thematic key 1.Sc3! not only unpins the black rook but also grants a flight on f6. The freed rook has four defences against the threat of 2.Qxg7, provoking two pairs of battery mates: 1…Rxd6 2.Sd5, 1…Rf4 2.Se4, 1…Ra4+ 2.Sxa4, and 1…Rd1+ 2.Sxd1. The flight-taking 1…Kf6 is answered by 2.Qe7. Most moves by the g5-knight defeat the threat but allow the same queen mate: 1…Se4/Sf7/Sh7 2.Qe7, while the correction 1…Se6 permits 2.Sd7.

George Meldrum: Whenever I see the name Eaton I expect something enjoyable. Nice key and the variations 1…Rxd6 2.Sd5 and 1…Rf4 2.Se4 are superb.
Nigel Nettheim: The key is good, though the immediate appeal of countering the rook checks made it easy to find. The g4-pawn is not needed; the h3-knight, which prevents some cooks, could be replaced by a black pawn on h6, producing a more natural and more economical position (1.Sxg5? hxg5! and 1.Sf2? h5!).

White mates are prepared against most of Black’s moves, and the key 1.Sf1! (waiting) completes the block. The star variation, made possible by the sacrificial key, is 1…exf1=Q+ 2.Rc2 – here the white queen exploits the opening of the e-file to control e5. A random move by the black rook produces 1…R~ 2.Qf2; the correction 1…Rg2 cuts off the black bishop and allows 2.Rf4. Since 1…Bg2 reciprocates the interference, enabling 2.Qf2, the Grimshaw theme is shown. The other bishop move is a square clearance: 1…Bxf3 2.Sxf3. The well-utilised queen gives four further mates: 1…h3 2.Qh4, 1…d2 2.Qxd2, 1…S~ 2.Qc3, and 1…a4 2.Qb4.

Jacob Hoover: A pair of Grimshaw interferences in 1…Rg2 2.Rf4 (this was the point of guarding e3 with another piece) and 1…Bg2 2.Qf2.
Nigel Nettheim: Wonderful. The provision for the h4-pawn with 1…h3 2.Qh4 is very pleasing; and 1…exf1=Q+ 2.Rc2 is brilliant, e5 being newly protected. Evidently the composer died a few years after this was published, aged not quite 35.

The white queen takes a step back with 1.Qg8! to threaten 2.Qb8. Black’s d4-knight, the thematic piece, defends by opening a line for the rook on d3. In completing its move, however, the knight also closes various black lines of defence, and four such self-interferences are exploited by White: 1…Sb5 2.Sxc6, 1…Sc2 2.Sxd3, 1…Se2 2.Re4, and 1…Sf3 2.Qg3. A fifth move by the knight commits the error of self-block: 1…Sf5 2.Sf7. There’s one variation of by-play, 1…f5 2.Qxh8.

Jacob Hoover: Square-vacating key. A very classy (near) knight-wheel theme.
Nigel Nettheim: Five eighths of a knight-wheel. The h8-bishop prevents a dual after 1…f5; however, it could better be replaced by the white king.

The key 1.c5! guards d6 and traps the black king in the f4-e5 cage. There is a full-length threat: 2.Sb4 Ke5 3.Sxc6+ Kf4 4.Se7 Ke5 5.Sg6. But this is stopped by Black’s only legal move 1…Ke5, since now 2.Sb4? Kf4 3.Sxc6 is stalemate. White plays 2.0-0 instead, subtly motivated to allow the knight to attack from the king-side and to support the f2-pawn with the rook. 2…Kf4 3.Se1 Ke5 4.Sg2 fxg2 5.f4. Note that the try 1.0-0? could transpose to the solution after 1…Ke5 2.c5, but it’s refuted by 1…c5! Bob Meadley calls this problem “a classic” in his recent paper on the composer.

Nigel Nettheim: Excellent! An alternative setting is possible by reflecting the position along the e-file and then placing the rook on a1: 1.g5! Ke5 2.0-0-0. The a1-rook has more freedom than the h1-rook and 0-0-0 is a rarer move in chess games.
George Meldrum: Set up the chess set to solve this one… no need… hmmm… maybe I will need to after all. Nice!
Jacob Hoover: I usually don't like stalemate-avoidance problems, but this one had a delightful anticipation theme that I very much enjoyed. I also enjoyed the fact that one of the moves was a castling move and that the black king was mated by a bunch of pawns. Great job, Mr. Watson.

After 1.Rb5! White threatens 2.Bd5, which fires the Q + B battery while covering the e4-flight. Black’s queen and b8-bishop defend by pinning themselves on e5, anticipating that the threat move will unpin them and allow them to interpose on f6. However, White exploits these self-pins – called Schiffmann defences – with 1…Qxe5 2.Seg3 and 1…Bxe5 2.Sd6 (also 1…Bd6 2.Sxd6). If Black takes the flight, White utilises the Q + B pair again, but now as an indirect battery: 1…Kxe4 2.Bg6. Lastly, the self-block 1…Sxe4 enables 2.Qc8.

Jacob Hoover: The thematic defences are self-pins that allow pin-mates from the e4-knight.
Nigel Nettheim: Finding the key gave a real “light-bulb moment”. Not 1.Ra5/Rc5? Be8!

The diagram shows a complete block position, with every black move given a set mate: 1…Se~ 2.Sf3, 1…Sg~ 2.Se6, and 1…B~ 2.Bxc3. The set play cannot be retained with any simple waiting move, e.g. 1.Kg1? Be3+! The flight-giving key 1.Se3! (waiting) brings about an almost complete change of play: 1…Se~ 2.Sc2, 1…Sg~ 2.Sf5, 1…B~ 2.Qxc3. Now 1…Bxe3 is a correction move that disables the queen mate, but due to the self-block it allows 2.Bxc3 once more. An added mate results when the black king captures the offered knight: 1…Kxe3 2.Qe4. A well-constructed mutate that increases the number of variations from three in the set to five in the actual play.

Nigel Nettheim: A good key with a double sacrifice. The description “all change here” does not quite apply, because of 1…Bxe3 2.Bxc3.

The black king has two diagonal flights, one of which is unprovided or without a set mate. The waiting key 1.Sxg4! completes the block, whereupon both flight-moves provoke pin-mates by the b5-knight: 1…Kf5 2.Sd4 and 1…Kd5 2.Sc7. The other black moves all permit queen mates, either by opening white lines – 1…R~ 2.Qf6, 1…f6/f5 2.Qd7, or by unguarding mating squares – 1…Sf~ 2.Qe5, 1…Sb7 2.Qxf7.

Nigel Nettheim: After 1.Se~ Kd5, 2.Sc7, but only 1.Sxg4! works. The black king’s other flight is also self-pinning, doubling the theme.
Jacob Hoover: Any move by a black unit other than the king allows the queen to mate at point-blank range.

This problem was originally cooked by 1.Sxf3; it's fixed by adding the g4-pawn but the latter results in a capture-key. Nigel Nettheim has found a better correction that avoids such an aggressive key – see the setting below (1.Sh3!) which uses a black bishop instead to stop the cook. Another advantage of this version is that 1.Se6? becomes a good try that’s defeated only by 1…g5!

In each solution, both white bishops are captured to facilitate Black’s promotion to a second white-squared bishop. The first game goes 1.a4 g5 2.Ra3 Bg7 3.Rb3 Bxb2 4.Rxb7 f6 5.Rb4 Ba6 6.Bxb2 Bxe2 7.Be5 Bd3 8.Bf4 gxf4 9.h4 f3 10.Rh3 fxg2 11.Rf3 gxf1B 12.Rff4 Bh3. The second part involves another precise sequence of play, one that brings about the thematic exchange of places for two pairs of pieces: 1.h4 g5 2.Rh3 Bg7 3.Rb3 Bxb2 4.Rxb7 f6 5.Rb4 Bb7 6.Bxb2 Bxg2 7.Be5 Bh3 8.Bf4 gxf4 9.a4 f3 10.Ra3 fxe2 11.Rf3 exf1B 12.Rff4 Bd3. So the identities of the two white rooks and the two black bishops in the diagram are swapped across the two phases.

Andy Sag: Had a bit of fun with this. I finally cracked it after realising that the black g-pawn can reach f1 via e2 instead of g2.