Low approach to 3-5 point
I would like to examine the practical experience with the 3-5 point as a background for looking at the different joseki. As a first step I have summarized the historical results as found by searching the Go Games on Disk (May 2002 version) using kombilo. DaveSigaty
I ran the search on the otherwise empty quadrant plus one extra line above and to the left. This minimizes but does not completely eliminate continuations that are directly based on the positions to the left or above.
Numerically these are the top 11 continuations where Black plays next. Note that the search position occurs very frequently but that the players very often tenuki from here. Thus there is a big question of whether Black plays next (= a 5-3 point situation) or White plays next (= a 3-4 point situation). We will only look at the 11 plays for Black shown.
Note: read the following information as (in the case of the first line): Black plays at a in 86 cases (9 times immediately on the next move, otherwise with one or more plays in between), Black won 40.7% of the games in which this move was played while White won 54.7%. The remainder of 4.6% were cases with no winner = jigo, unfinished, result unknown, etc.
Ba: 86 (9), B40.7% - W54.7% three-space low pincer Bb: 103 (16), B44.7% - W54.4% two-space high pincer Bc: 202 (14), B39.1% - W56.4% two-space low pincer Bd: 468 (85), B38.5% - W56.6% taisha Be: 251 (33), B49.4% - W45.8% one-space low pincer Bf: 301 (38), B53.5% - W44.9% Bg: 43 (5), B41.9% - W55.8% Bh: 355 (95), B45.1% - W49.0% Bi: 381 (88), B42.0% - W52.0% Bj: 316 (62), B41.8% - W50.0% Bk: 56 (2), B42.9% - W53.6%
For me it is fascinating that only two of the continuations for Black have winning records (Black's winning percentage is higher than White's): the one-space low pincer at e and the knight's move at f. The actual experience for all other moves is that White wins more often than Black. In particular the taisha joseki (Black at d) has a very clear advantage for White with the best winning percentage in actual practice. Despite this, the taisha has been the most popular continuation by a fairly wide margin.
Speculating on why the results might be so poor for Black, we can see that playing 5-3 and letting White enter at the 3-4 point cedes the corner territory to White. Black does so in expectation of taking compensation on the outside. However, it seems that Black needs to pressure White if he is to gain sufficient compensation for the loss of the corner. Black does not seem to be able to do this consistently with simple extensions up the side or loose pincers. For her part White has the corner. If Black pincers, White chooses whether to sit tight in the corner or fight it out. Black can normally dictate matters on one side but can not easily do so on both sides simultaneously.
Alex Weldon: These statistics also do not tell the whole story. 4-3 is not the only approach to 3-5. 5-4 is also a common approach. Also, it's common that the 5-3 player will get another move in the corner and make a shimari. To judge whether 5-3 is a good move or not, you'd have to include the statistics from those and other (e.g. 3-3) possibilities. If the statistics for the opponent making the high (5-4) approach were favourable for the 5-3 player, for example, that would tell us that to make 5-3 effective, you have to arrange the situation so that the low approach is unfavourable for the opponent. Looking at these statistics alone, all we can say is that you don't want to play 5-3 if the opponent is going to be able to invade 3-4, unless you're planning on pressing or making a one-space low pincer.
The 5-3 results also tell the player who opens with 3-4 something. When the opponent approaches at the 5-3 point, think seriously about tenuki. The corner then converts into a 5-3 opening which is difficult for the opponent to exploit. Even better in a komoku - tenuki scenario, the opponent did not plan to play 5-3 to begin with and is suddenly confronted with the need to choose a continuation :-)
- Bill That really does not follow. You need to compare the results of tenuki vs. local reply.
- The fact that it does not follow does not mean that it isn't true. Tenuki is always an option.
Andre Engels: I find a different result with MasterGo with my own additions. It gets to exactly 50%. The big difference seems to be that MasterGo has less Taisha games. The total number is also lower, but still I think the results are interesting. Between brackets are the results as black and as white. In the database as a whole black scores 55%, white 45%, for a player playing in this position the score is 56% with black, but only 41% with white.
a: 102 times, 47% (40% - 52%)
b: 151 times, 48% (57% - 39%)
c: 174 times, 47% (54% - 40%)
d: 365 times, 46% (53% - 36%)
e: 242 times, 54% (57% - 42%)
f: 353 times, 55% (64% - 44%)
g: 13 times, 38% (57% - 17%)
h: 78 times, 40% (36% - 44%)
i: 75 times, 38% (56% - 32%)
j,k: not found
Other moves: 3 space high pincer: 12 times, 64% (70% - 0%)
The reason the last two moves are not found is probably that MasterGo tends to not include moves that are significantly closer to a stone in another corner.
Looking at these data, it seems that black has little to fear of playing in this position - most answers score around his expectation, and one (f) scores considerably higher. White might have more to worry about, but would especially have to shirk away from the taisha (36% out of 145 games is a low percentage).
Another interesting look is the variation of the result over time (x is the three space high pincer). *-marked moves scored 50% or more in the specified period. 'perc' shows the full percentage in the give period.
total perc a b c d e f g h i x other before 1950 144 28 17 37 24 19 1 18 27 1 1951-1970 162 45 3 27 23 24* 19* 29 2* 14 18* 1 2* 1971-1979 156 47 3 26 14 38* 31* 27* 1 8* 4* 2* 2 1980-1984 164 55 5* 12* 18* 49* 26* 40* 1 6 5 2* 1985-1987 192 59 9* 16 16* 54* 30* 51* 1* 5 7 3* 1988-1990 173 51 7 22* 14* 49 23* 39* 3 8* 4* 1 3 1991-1993 153 49 11 11 23* 42* 17* 31 1 7* 2 3* 5 1994-1996 152 52 18 14* 11* 21 31* 44* 2* 6 3 2* 1997-1999 149 49 25* 14* 19 23 20 37 1* 5 1* 4* 2000-2003 123 48 21 6* 19* 22 14 35 1* 5*
Looking at these data, I see even less reason to consider this position as bad. What I think is going on is that in the old days, a likely way for this position to come up would be in the Shusaku fuseki, with White having the 5-3 stone. The Shusaku fuseki gives a good percentage for Black, especially in the days before komi, and the resulting bad percentage for White results in a bad position for this position as well.