Four piece handicap (white removes rook, bishop, and both lances) has intrigued me for nearly two decades. It would seem that with all four of his long range pieces missing, white should have no chance against anyone past the novice stage, but this is not the case. Most pros will say that if you can beat them at this handicap, you are a Dan level player. I almost always win giving this handicap to anyone without a Dan rank, and have sometimes won quick games from players as strong as 3 Dan (!). The total score in official U.S. Shogi Federation rated games played at four piece handicap is 36-8 in white's favor. Clearly this handicap is much smaller than six piece or even five, but these results aren't reasonable; they indicate that the players of black just don't know the opening well, or that what they have learned is not correct.

My interest in this handicap is tied to the influence of the late Nada Rensho, 9 Dan pro and once challenger for the Meijin (grand championship). When his results in pro competition went downhill (perhaps for health reasons; he died at only 58), he turned his attention to handicap play, especially four piece. He wrote that the published analysis was both incomplete and inaccurate, and that the usually recommended climbing silver attack was no good. He claimed to be able to defeat any amateur below 4 Dan at this handicap, and proved his point by winning a series of a dozen games against Dan level opponents by an overwhelming 10-2 score, despite the fact that each game was published in a monthly magazine, so each opponent could study the previous games (and the literature) before playing his own game. The key to his success was that he always used his king to defend against the enemy attack, whereas all the literature assumed that white would not place his king in the enemy line of fire.

On my first trip to Japan I earned the rank of 3 Dan, and got the chance to play three games with Nada at four piece, plus two more on a subsequent trip. I won four out of the five, primarily because I already had considerable experience giving the handicap to American players. Nevertheless I was impressed by Nada's ideas, and resolved to work out for myself the best lines of play for black at four piece, since they could not be found in print. Common sense told me that it should not be difficult for any player who is even close to Dan level to defeat anyone with such a huge handicap, if he is only given correct lines of play in the opening. In other words, it should be possible to demonstrate a huge advantage right out of the opening, but proving this to be so was a difficult task!

Most literature on four piece recommends that black play a "climbing silver" attack, showing that he can win easily by dangling a pawn behind enemy lines and using it to deflect a defender. While this works well enough against routine play, it does not work very well when white uses his king as a defender, as Nada always did, or when white gets a pawn in hand early. The other lines recommended in the literature, including the "N1g" attack and the "retreating bishop", are also not very convincing against clever play. These lines will be covered in Part II. Some authors (notably Naito Kunio 9 Dan) have suggested that black just pretend it's a two piece handicap (rook and bishop) game, but only the missing right lance will help him if he plays that way. In fact the absence of his left lance may even help white, as sometimes black plays to win it in a two piece game. Four piece handicap should be much easier than two.

There are probably three reasons for this. One is that professionals like to teach and recommend lines that are likely to be useful in even game play. Another is that the climbing gold does not seem natural, because the gold is the ideal defense piece while the silver is more suited to attack, primarily because if it is exchanged off (the usual goal of the attacker), the opponent gets a less valuable piece. Why weaken your defense and risk the loss of the more valuable gold if the silver will do the job? My answer: the silver is not agile enough for the task, and it turns out that in the classical lines, white needs a silver to defend. Finally, not much attention has been paid to analyzing four piece in the two decades since Nada showed the weakness of the climbing silver.

It is my sincere hope that the climbing gold will come to be recognized as the proper way to play four piece handicap. In my opinion it is much easier to win this way than with the climbing silver, and I consider this discovery to be my greatest contribution to the game of shogi. The following analysis is based on the analysis of the climbing silver by a number of shogi Pros, including the late Kimura Meijin, the late Kato 8 Dan, the late Hanamura Motoji 9 Dan, the late Itaya 8 Dan, former pro Iida 5 Dan, and Shoshi Kazuharu 6 Dan, but modified where necessary by myself due to the use of the climbing gold rather than silver. Wherever I refer to the climbing silver, assume that it is on the same square as the advanced gold, and that the gold is on 5h instead of the silver on 4h. If there are errors, they will be mine.

The basic plan is to open the bishop's diagonal, advance and exchange off the rook's pawn, push the adjacent lance pawn twice, and march the gold to 3f, with the silver going to 4h for defense. Then follows P1d Px1d G2e (or G2e first if white cannot sensibly reply ...N3c). Next comes P*1b and a sacrificial breakthrough based on the use of the pawn to deflect the silver from 2b (with a gold on 2b we drop the pawn on 1c). Let's look at the climbing gold, with comparisons to the climbing silver.

Play through the analysis by clicking on "Java viewer" or get diagrams by clicking on links.

Let's start with classical play:

(Java viewer)1...S6b 2 P7f P5d 3 P2f S5c 4 P2e G3b 5 P2d Px2d 6 Rx2d P*2c (For 6...S2b see below.) 7 R2h G7b (If white omits this black can force him to move his gold to a worse square; see below.) 8 P1f (Now 8 P9f P7d 9 P9e P8d is not advisable for black.) P7d 9 P1e S2b 10 G3h (Start of the climbing gold. Standard is S3h.) G7c 11 G2g G6d 12 G3f P5e 13 S4h (A very useful defensive move, guarding three pawns and the silver itself. If white ever plays ...P8e or ...Gx7f black may reply S7h.) G6e 14 P1d Px1d 15 G2e S4d 16 P*1b (Also good here is 16 P*1c Nx1c 17 Gx1d P*1b 18 G1e! with P*1d next, winning knight for two pawns. This isn't possible with a climbing silver.) Gx7f 17 S7h (Best defense against this or ...P8e; the bishop has a retreat if needed, and the king can run either way.) K6b (The king is both safer and more useful off the back rank.) 18 Gx1d P*1c


19 P1a+ (Also good is 19 G1e P8d 20 P*2d Px2d 21 Gx2d P*2c 22 P1a+. Not so good when white has a pawn in hand, as here, is 19 Gx2c (or 19 Sx2c in the climbing silver) Sx2c 20 P1a+ P*2d 21 +Px2a P1d.) Px1d (19...Sx1a 20 Gx2c wins easily.) 20 +Px2a S1c 21 P*1b P8d (This attack is no longer recommended in the climbing silver, due to the defense 21...P2d 22 P1a+ S*2c, with no easy way to continue the attack. But here white has the wrong piece in hand to prevent +P1b, so this attack should be ideal with the climbing gold.) 22 P1a+ P8e 23 N*6h! (White was threatening ...P8f Px8f P*8g with counterplay.) G7e 24 +P1b S2d 25 +P22b! G4b 26 +P1c! K7c (26...Sx1c 27 Rx2c+) 27 +Px1d S23e 28 Rx2c+, intending +P3b and +Px3c. Note how the promoted pawns gain strength as they retreat; generals are almost useless on the ninth rank.

Variation A (K on 4c)

This line is quite similar to the above line, but white emphasizes defense:

(Java viewer) 8...K4b (8...K5b may transpose.) 9 P1e P2b 10 G3h (Again S3h is standard.) P4d 11 G2g K4c 12 G3f P3d 13 S4h P7d (13...N3c 14 L1f intending R1h and P1d.) 14 P1d (If 14 G2e first then 14...N3c is possible, though 15 P1d is still good for black.) Px1d 15 G2e (a diagram) G7c (15...N3c 16 Gx1d intending G1c and P*1d!; 15...S4b? has the point against the climbing silver to play ...S43c to answer Sx1d by ...P2d or ...S2d, but that's nonsense with a gold on 1d.) 16 P*1b (Again 16 P*1c should win knight for two pawns here; white's best reply is probably 16...N3c!.) G6d (Now 16...P3e? and 16...P2d? are reasonable moves against the climbing silver but blunders here; likewise 16...K3c can be met by simply 17 Gx1d P*1c 18 G1e intending N1g-2e, which would be impossible with climbing silver.) 17 Gx1d P*1c 18 P1a+ (Now with a silver on 1d, best is 18 Sx2c+, but here the text is best.) Px1d 19 +Px2a S1c (19...S3c 20 P*1c and 21 P1b+) 20 P*1b P5e (Again, with the climbing silver white would have the tough defense 20...P2d 21 P1a+ S*2c.) 21 P1a+ S2d 22 +P1b G3c 23 +P22b! N7c 24 +P1c! Sx1c 25 +Px2c P*2d 26 +Px1c with an easy win.

Variation B

(Java viewer) 7...K5b (By omitting ...G7b, white tries to save a tempo and possibly use the gold for defense on his left) 8 P9f (So black punishes him, though he could also just play the above climbing gold strategy. This plan should only be adopted when white has no generals on his three right files and has not advanced his 7 pawn.) G7b 9 P9e P7d (White can defend by 9...G8b, but then black will switch back to the climbing gold attack, with similar play to the above lines.) 10 P9d Px9d 11 P*9b G8b 12 P9a+ N7c 13 +P8a! (My own idea; several books give 13 B6f in similar positions, but 13...P6d or 13...S6d is a bit messy. Sacrificing the promoted pawn gains a vital tempo.) Gx8a 14 B6f G8b 15 Lx9d S6d (else 16 R9h) 16 L9c+ G7b (16...S6e 17 +Lx8b wins two generals for a bishop, with a continuing attack.) 17 +L9b (Planning B9c+.) P8d (17...P7e 18 R9h) 18 R9h S6e 19 B9i and the rook promotes.

Variation C

(Java viewer) 6...S2b 7 R2h G7b (Again, if omitted black will play P9f-9e as above.) 8 P1f P7d 9 P1e S4d 10 G3h P3d 11 G2g N3c (The point; white keeps the enemy gold out of 2e.) 12 G3f (Note that a silver here could be driven back by ...P3e, but a gold can move sideways to 4f if hit.) P*2c (Else 13 P1d Px1d 14 Lx1d P*1c 15 Lx1c+.) 13 S4h K5b (Note the gold on 3f, unlike a silver, prevents P5e, due to G4f.) 14 L1f G7c 15 P1d Px1d 16 R1h and black will break in on the edge.

Nada's Line

(Java viewer) 4...K4b 5 P2d Px2d 6 Rx2d K3b 7 R2h P*2c 8 P1f (Black can also try to punish white for omitting ...G7b by playing P9f-9e as above, but with white's king far away this is not so convincing.) G4b 9 P1e S2b 10 G3h S4d (10...P4d is quite similar but Nada normally played the text move against the climbing silver, as it provides better defense and prepares the annoying ...P5e.) 11 G2g P3d 12 G3f G6b 13 S4h G65c 14 G2e G3c 15 P1d (Another good plan, recommended by M. Sugimoto Pro 5 Dan with the silver on 2e, is 15 P3f, 16 P4f, and 17 N3g, aiming to win silver for knight by N4e. It is probably black's best line with the climbing silver, but as I will show black can do better than this with the climbing gold.) Px1d 16 P*1b (16 P*1c Sx1c 17 Lx1d Sx1d 18 Gx1d is not so clear since the pawn on 2c is adequately protected here.) P5e (Now 16...P1e is effective against the climbing silver, since 17 Lx1e P*1c leads nowhere and 17 S1d can be met by ...G2d, but here 17 G1d followed by Lx1e will eventually transpose to the text line, while 17 Gx1e followed by N1g-2e is also very strong.) 17 Gx1d


P*1c (17...G6d 18 N1g! and the threat of 19 N2e will force 18...P*1c 19 Gx1c, with play like the text but even better for black, since he has an extra knight in the attack. Now with a silver on 1d 17...P2d or 17...G2d are not so easy to refute, but here they are just blunders.) 18 Gx1c Nx1c (18...Sx1c 19 P1a+ is even better for black.) 19 P1a+ Sx1a 20 P*1d! (Here in similar positions from the climbing silver the literature gives 20 Lx1c+ P*1b 21 +L1d which is good but not overwhelming for black; the text seems clearly better.) P*1b 21 Px1c+ Px1c 22 N*2e! S2b 23 Nx1c+! (Rather than 23 Lx1c+ because a lance is more useful to white here than a knight.) Sx1c 24 Lx1c+. At minimal cost in material (gold for silver), black has promoted his lance on a highly effective square near the enemy king, and has many threats. For example: 24...G6d 25 S*1b G*2d (or 25...N*3a 26 P*2b or 25...P2d 26 P*2c) 26 P*2b. There is no defense to 27 P2a+ and 28 +P2b.

In my opinion the Nada line is very effective against the climbing silver, but the climbing gold handles it nicely.

Gold defense

This is like Nada's line, but with the gold on 3c and silver on 2b reversed. It is rarely seen, because a gold on 2b is bad style, but it does have the merit of preventing P*1b. Here too the climbing gold works better than the climbing silver, because black must aim for P*1c.

(Java viewer) 1...K4b 2 P7f S6b 3 P2f K3b 4 P2e S4b (Instead of 4...G4b which would transpose to Nada line.) 5 P2d Px2d 6 Rx2d P*2c 7 R2h G3a 8 P1f P5d 9 P1e G2b 10 G3h S65c 11 G2g P4d 12 G3f P3d 13 S4h S3c 14 G2e S6d 15 P1d Px1d 16 P*1c


P5e (Both captures are bad, since 16...Gx1c 16 Lx1d is very strong, and 16...Nx1c 17 Gx1d P*1b 18 G1e followed by P*1d wins material; again this would be impossible with a silver on 1d) 17 L1f!! (Better than 17 Gx1d P*1a or 17 Lx1d Nx1c.) G5b 18 R1h P2d 19 Gx1d P*1a 20 R2h and white has no decent defense to 21 Gx2d. This is a remarkable variation, because both sides could capture an edge pawn two different ways, yet the best move for both sides was to decline! This line is quite tough to defeat with a climbing silver, but the climbing gold can handle it quite well.

Iida's defense

This line is similar to the classical but white attacks with silver instead of gold, and exchanges pawns on 7e. Again the climbing gold works well, though in this line I cannot say that it offers any advantage over the climbing silver; the analysis would be the same.

(Java viewer) 1...G65b 2 P7f P6d 3 P2f G3b 4 P2e G6c 5 P2d Px2d 6 Rx2d P*2c 7 R2h P7d 8 P1f S6b 9 P1e S2b 10 G3h S7c 11 G2g S8d 12 G3f K6b 13 S4h P7e 14 Px7e Sx7e 15 P1d Px1d 16 G2e P6e 17 P*1b (Again 17 P*1c is a decent alternative with the climbing gold.) P6f (To block the bishop.) 18 Px6f N7c 19 P1a+! (The normal continuation 19 Gx1d P*1c 20 P1a+ is a bit slow here, as white is poised to counterattack, but because white has exchanged pawns on 7e giving black a pawn in hand this quicker attack is possible) Sx1a 20 P*2d! (Not 20 Gx1d? S2b.) S2b (20...Px2d 21 Gx2d P*2b 22 P*1c with Lx1d next or 21...S2b 22 P*2c with P*1c next ) 21 Px2c+ Sx2c 22 P*2d S1b 23 Gx1d P*2b 24 P*1c winning a knight with a continuing attack.

Early ...P3d

White can prevent the exchange of the rook's pawn altogether by an early ...P3d, but the opening of the bishop's diagonal should prove fatal.

(Java viewer) 1...G3b 2 P7f S2b 3 P2f P3d?! 4 P1f (now there's no point to 4 P2e as 4...S3c or 4...G3c prevents 5 P2d, but a quick knight attack works) G3c 5 P1e P4d 6 N1g G4c 7 N2e K4b 8 L1g P2d (Else 9 R1h.) 9 Nx1c+ Sx1c 10 P1d S2b 11 R1h (11 P2e is also strong.), with P1c+ next.

To summarize, the climbing gold works better than the climbing silver in nearly every variation because it guards 2d while on 1d, it guards 3e while on 2e, and it can retreat from 1d to 1e or from 2d to 2e. Also, white needs a silver, not a gold, in the main line. The only downside is a slightly weaker defense. The evidence is overwhelmingly in favor of the climbing gold.

1. The main goal is to promote pawns, then exchange them for generals and clear the way for the rook to promote.
2. Two attackers (rook and gold) cannot overcome two defenders (gold and silver), but the use of a pawn to deflect a defender makes the attack work.
3. Concentrate on your right side as black, unless white totally abandons his right side.
4. If white guards 2c three times, focus your attack on 1c.
5. One or two defense moves are all that are needed; attack before white can counter.
6. As white, try to block out the enemy bishop, and/or get a pawn in hand early.