PageType: HomePage    

KGS 2d, IGS 3k

Perfect vs. Optimal Play

Warning: speculative ideas ahead.

Recently I've had an internal struggle over how I think about the game. There seems to be a school of instruction that focuses on playing moves that increase your chances of winning (Bill Spight's perfect play) vs. optimal play, which maximizes the score (i.e, always seeking the most severe line.)

I had the pleasure of attending a workshop by the strong American amateur Yuan Zhou. I think Yuan might say that this is the difference between trying to win and seeking the truth. In one of his books, he describes the style of Go Seigen as being uncompromising, never giving the opponent more that he deserves. There are games in which he would take what other pros might consider huge risks just to contest a couple of points, when a strategy of perfect play would seek to remove complications in a won game. I'm not sure this was always the case. Earlier games of Go Seigen seem to follow the style of Shusaku, whom most people might consider to be more of a perfect player than on optimal player. As I understand it, the whole idea of the Shusaku fuseki is to preserve black's advantage in a no-komi game, giving white fewer ways to complicate the situation.

Lee Changho, on the other hand, could be described as a player who is comfortable with close games. Because of his endgame skills, pros who seek to win against him often try to start complicated fights in order to make the game less predictable, and reduce his ability to consistently win games if he is ahead by even the slightest of margins. According to at least one pro, there's a whole generation of young Korean pros whose entire fighting style is based on the requirement be able to beat a player like Lee Changho.

Several books (especially from the Japanese tradition) seem to advocate playing sub-optimal moves in handicap games. Michael Redmond's The ABCs of Attack and Defense, Naoki Miyamoto's The Breakthrough to Shodan, Handicap Go, etc. are just some examples. The authors of these books err on the side of advocating moves that might even be considered slack, but which avoid complications. They are certainly not alone and there are many more examples. With respect to even games, Hane Naoki describes his style is a kind of "80%" style---he doesn't try to find the most severe move, just one that is good enough.

On other hand, teachers such as James Kerwin seem to admire pros who are tenacious dogs (like Sakata and Cho Chikun). He has speculated that one of the reasons that they have won so many titles is that they never yield---even when they are ahead, they keep the pressure on.

To date, I've been more of the optimal play type of player (except without the success of the luminaries mentioned above.) I think most players who play on online are like that. It's probably more fun. The great thing about that is that you don't really need to count---you just always play the most severe move. The disadvantage is that it's not a great way to win won games, and I think bringing home a game where you have a clear advantage is an important and interesting skill.

So I might try to change my style, and deliberately choose simpler lines that keep the balance of the game if I get an early advantage. This is not say that I want to play like some of these Monte Carlo engines, soulless robots that choose moves that are almost insultingly slack when ahead. It's more that I would like to learn to win with normal moves.

San-san obsession


Even since I started playing and studying the basic opening moves, I've had an obsession with the idea of playing the 3-3 point in an empty corner, although I haven't actually played it that much myself because my early attempts didn't produce good results. But those early failures don't mean anything to me now, and I'm itching to re-think my game after playing Kobayashi fuseki as black or star point and 3-4 point as white for so long.

Fortunately, there are more sources for study. There's Cho Chikun's book, which deserves another read, and of course pro games. The josekis themselves are simple and the 3-3 in general has a simpifying effect on the game, making one of the corners of the board less interesting for a while to pursue action elsewhere.

San-san practitioners

I did some kombilo searches to determine who the biggest practitioners of the 3-3 point are. Since far fewer players play it as black then as white, it's important to separate them:

Most games in GoGoD as Black (>= 10), with frequency of player's total games where they played 3-3.

      #  Freq    Player
    109 (73.6%)  Fujisawa Hosai
     87 (14.7%)  Sakata Eio
     40 ( 4.8%)  Cho Chikun
     35 ( 9.8%)  Go Seigen
     16 ( 3.7%)  Kitani Minoru
     13 ( 5.6%)  Yamashiro Hiroshi
     13 ( 4.0%)  Ishida Yoshio
     10 ( 1.8%)  Seo Pong-su
     10 ( 6.9%)  Kudo Norio

Most as white (>= 50)

    265 (44.9%)  Sakata Eio
    166 (49.0%)  Ishida Yoshio
    117 (14.0%)  Cho Chikun
    112 (16.1%)  Rin Kaiho
    101 (19.9%)  Otake Hideo
     98 (13.2%)  Hashimoto Utaro
     66 (17.6%)  O Rissei
     66 (11.2%)  Kato Masao
     61 (15.9%)  Fujisawa Hideyuki
     58 (16.4%)  Ma Xiaochun
     50 (9.0%)   Seo Pong-su

So that's plenty of study material. Of course the usual suspects Cho, Sakata, Hashimoto, Go Seigen and Hosai show up, but I didn't know that Kato had dabbled in it so much. I also didn't know there were pros who played san-san either as black or white in more than 50% of their games!

Some concepts I am currently trying to understand better:

Developability, open skirts, and the macroendgame.

In general, this is part of my goal to improve my oyose, the macro-endgame. I have studied professional games since I was 30k, with various levels of understanding as you might expect, and I now feel that one of the biggest parts of the game I have left unstudied is the early endgame. The fuseki and the middle game have general principles that give reasonable results, and I've seen my share of pro lectures and read the approrpiate strategy books, so I'm not feeling I'm missing anything big here. Of course I still make big mistakes in these areas, but mostly I attribute it to poor attention, inexperience, or shallow reading rather than flat out ignorance. I'll readily admit total ignorance of the macroendgame, because the material I've seen on it is rather sketchy. Most endgame books deal with tesuji and late endgame sequence values, which are important, but in many games it seems there are 50-60 moves between the middle game and those more settled postions where I am completely lost and am probably losing lots of points. So I have taken to studying this section of pro games to improve my intuition in this area. There is a feeling of playing away from strength that seems to persist into the endgame further than I would have expected.

The following were added when I was about 12k, but still these are important.

Thickness and compensation

I think that there is a slight problem with the way thickness sometimes described in English translations of joseki books ind in informal discussions. One often sees or hears phrases such as: "white gets thickness in exchange for territory" or "white gives up territory in exchange for thickness." Although technically correct, it leaves out the important point that the thickness isn't immediately worth anything. Although the territory can be counted and will likely remain mostly the same throughout the game-plus or minus some amount depending on sente in the endgame-the value of thickness depends on how well it is utilized. This is common knowledge and is not controversial. The problem with the traditional phrasing is that this it makes it sound like it's an exchange of money for goods. But it's more like an exchange of money for stock options, or maybe like a loan. So a better way of thinking would be:

White's strategy: "I, white, I am going to loan black some territory in exchange for thickness. I hope to use my thickness to collect on this loan with interest. I will therefore keep track of this debt and make sure I get paid in some appropriate form before the end of the game."

Black's strategy: "I, black, I am going to get some territory on loan from white in exchange for giving her thickness. But I am a sneaky cad and I plan to default on this loan before the end of the game. I will keep wihte busy and by the end it will be too late for white to collect."

So if you are white, you need to be a vigilant creditor and remember that black owes you something. At the end of the game you should be able point somehere and say: "this is where I got what was owed to me." If you can't do that, you probably get cheated.

If you are black, you have to make sure you can default, and reduce white's ability to collect.

Ko is your thug to collect to your debt.

It seems logical that thick shapes probably cost some extra or slow moves in order to keep the weakness to a minimum. If a shape has too many weakness, it is not thick, after all. It's a global consideration, but if you are playing thickly there is a good chance that you are owed something in ko threats. The ability to create favorable ko situations or to play in such a way as to profit from the fact that your opponent is afraid to fight a ko would seem to be a requirement for anyone who needs to get maximum benefit from thickness.

Thickness and endgame profit.

Thick play should mean that your opponent has to be more conservative in the endgame. He probably owes you some gote moves. I don't understand this very well, but have heard about it.

Simply connecting is often sente.

Just connect  

Often I have a feeling in tactical situations that connecting is slow, even if it has a sente follow-up. Simply W1 is the proper move here, even though I have the urge to force at a, which is bad, because black needs B2 after W1 anyway. I have to get a better feeling over when solid connections require responses rather than thinking of them as passive moves.

Calvin last edited by Calvin on April 22, 2009 - 05:55
RecentChanges · StartingPoints · About
Edit page ·Search · Related · Page info · Latest diff
[Welcome to Sensei's Library!]
Search position
Page history
Latest page diff
Partner sites:
Go Teaching Ladder
Login / Prefs
Sensei's Library