Life In 19x19

In Chess, they now have computers look at games that were played and analyze their quality. If Go programs were at the same level, they could say "Shusaku was stronger than Shusai, but weaker than Go Seigen" or whatever, and they could also say "this slow game is like ten times awesomer than this quick game" or perhaps "Lee Sedol drunk is better than Cho U sober"...

Alas, it'll be awhile before we'll be made wholly obsolete.

tj86430 wrote:

Vesa wrote:Now, hello, what was the other game Magicwand was so keen on?

I'd guess anything with Korean players.

Mok jinsuk - choi chulhan
i wish i can post the game..but i am not sgf savvy so it will take 30 min for me to create sgf.

i hope someone can post the game.

hyperpape wrote:In Chess, they now have computers look at games that were played and analyze their quality. If Go programs were at the same level, they could say "Shusaku was stronger than Shusai, but weaker than Go Seigen" or whatever.

Alas, it'll be awhile before we'll be made wholly obsolete.

In chess, how do they measure the quality with a computer? Is it based on optimal play, using the computer's evaluation?

Also, it is interesting to note that, for some poker AI programs, it is advantageous for the AI to play in a "suboptimal" fashion, because humans make irrational decisions. So if a computer were to play against humans using a purely rational, game theoretically optimal strategy, it may not gain as much money than a computer AI that plays "suboptimally" to take advantage of human irrationality.

Magicwand wrote:

tj86430 wrote:

Vesa wrote:Now, hello, what was the other game Magicwand was so keen on?

I'd guess anything with Korean players.

Mok jinsuk - choi chulhan
i wish i can post the game..but i am not sgf savvy so it will take 30 min for me to create sgf.

i hope someone can post the game.

[sgf-full](;
EV[2010 olleh kt]
DT[2010-08-10]
KM[6.5]
RE[W+Resign]
PB[Mok Jinsuk]
BR[9d]
PW[Choi Chulhan]
WR[9d]
;B[qd];W[dp];B[pp];W[dc];B[de];W[ce];B[dd];W[cd];B[ec];W[cf];B[db];W[cc];B[kc];W[od];B[fd];W[pf];B[oe];W[pe];B[pd];W[nd];B[of];W[pg];B[og];W[ph];B[ob];W[ld];B[oh];W[oi];B[me];W[mf];B[pi];W[ne];B[mg];W[qi];B[pj];W[qj];B[ni];W[re];B[mc];W[md];B[lc];W[lg];B[oj];W[ql];B[lf];W[nf];B[lh];W[kg];B[kf];W[kd];B[jc];W[rd];B[rc];W[jf];B[jg];W[kh];B[je];W[ng];B[mh];W[nh];B[oi];W[mi];B[li];W[mj];B[lj];W[mk];B[pl];W[pm];B[ol];W[om];B[nl];W[jh];B[lk];W[ml];B[nm];W[nn];B[mn];W[mm];B[on];W[no];B[qm];W[jl];B[jj];W[ij];B[ii];W[ig];B[ke];W[jd];B[ie];W[id];B[hd];W[kj];B[kl];W[km];B[jk];W[hh];B[jm];W[he];B[kn];W[rl];B[pn];W[kp];B[mo];W[mp];B[lm];W[ic];B[iq];W[oq];B[pq];W[jq];B[ip];W[pr];B[qr];W[lo];B[ln];W[ir];B[hr];W[jr];B[nr];W[nq];B[or];W[gq];B[eq];W[dq];B[dr];W[cr];B[er];W[ep];B[hq];W[mr];B[co];W[cp];B[bo];W[fq];B[fr];W[ho];B[jp];W[in];B[jo];W[ns];B[en];W[fn];B[fm];W[fo];B[em];W[bp];B[ch];W[cb];B[hf];W[ge];B[gf];W[if];B[fe];W[le];B[eh];W[os];B[qe];W[qf];B[sd];W[rf];B[bg];W[jb];B[kb];W[ib];B[nb];W[fi];B[fb];W[fh];B[ei];W[eg];B[fg];W[dg];B[gd];W[da];B[gr];W[cj];B[ci];W[ej];B[fj];W[dj];B[bj];W[bk];B[bi];W[fk];B[gj];W[gk];B[gh];W[cm];B[ea];W[eb];B[cn];W[gm];B[db];W[ab];B[cl];W[bm];B[dl];W[bl];B[ck];W[dm];B[ad];W[ae];B[be];W[bf];B[af];W[bd];B[ek];W[dn];B[el];W[hi])[/sgf-full]


Is this the game you meant, Magicwand?

It may be possible that it's the case that a game that's played under faster time constraints has more mistakes (though this has not been proven).

Not proven, but my impression that commentaries on older, slow games seem to point to fewer mistakes than commentaries on modern slap-happy games suggests to me that time does affect quality in that sense. On the other hand, quality as measured by excitement is not necessarily time based, however. Shin Fuseki games were (are?) considered exciting even when they were played with 13 hours each.

I think it is also useful to give some weight to opinion of the now fairly wide spectrum of pros that fast games lead to games they are not very proud to have in their portfolios.

Kirby wrote:In chess, how do they measure the quality with a computer? Is it based on optimal play, using the computer's evaluation?

Also, it is interesting to note that, for some poker AI programs, it is advantageous for the AI to play in a "suboptimal" fashion, because humans make irrational decisions. So if a computer were to play against humans using a purely rational, game theoretically optimal strategy, it may not gain as much money than a computer AI that plays "suboptimally" to take advantage of human irrationality.

I'm actually not sure how the computer evaluates plays. I suspect it's based heavily on the computer's evaluation of optimal play, yes. There are notable debates here, since a player may deviate slightly from optimal play to create a complicated position that favors their skills. So it's not as if human judges completely defer to computer evaluations.

Maybe I've misunderstood, but the point about poker seems obvious. I feel like it only sounds notable because we tend to have associations with "optimal" play.

John Fairbairn wrote:

It may be possible that it's the case that a game that's played under faster time constraints has more mistakes (though this has not been proven).

Not proven, but my impression that commentaries on older, slow games seem to point to fewer mistakes than commentaries on modern slap-happy games suggests to me that time does affect quality in that sense.
...
I think it is also useful to give some weight to opinion of the now fairly wide spectrum of pros that fast games lead to games they are not very proud to have in their portfolios.

That could be the case. I think that it's a common belief that more time leads to fewer mistakes. I basically want to just note that it might not necessarily be the case - or at least not as much of a case as seems to be brought up. It's not like pro games with the three hour time limits are 10 second blitz, for example.

Part of the reason this came to mind was likely due to a book called, "The Art of Learning" by Josh Waitzin. He was a pretty good chess player when he was younger, apparently, and he claimed that when he looked at his game records, some of his best moves came from instances when he used a "medium" amount of time.

However, when it comes to go, I'm sure that go pros consider the aspects of go better than Josh Waitzin does. But for me, I do not have an exact idea in my head about how time relates to mistakes, so I basically want to say that a difference in time is not an end-all argument to say that there are fewer mistakes - at least it isn't to me.

John Fairbairn wrote:On the other hand, quality as measured by excitement is not necessarily time based, however. Shin Fuseki games were (are?) considered exciting even when they were played with 13 hours each.

True. I also believe that different people might have different ideas about what constitutes the idea of "quality". People have different favorite pros (though a person's favorite pro may not only be because of the "quality" of his/her play), after all.

Kirby wrote:

Magicwand wrote:
Mok jinsuk - choi chulhan
i wish i can post the game..but i am not sgf savvy so it will take 30 min for me to create sgf.

i hope someone can post the game.

Is this the game you meant, Magicwand?

thanks..
i love way choi chulhan plays...
he usually overpowers every opponents he plays.
pure power play and he is winning most games.

if someone want to view really exciting game...i suggest choi chulhan's game.

it is really complicated and exciting.

hyperpape wrote:...

Maybe I've misunderstood, but the point about poker seems obvious. I feel like it only sounds notable because we tend to have associations with "optimal" play.

The reason that I find the part about poker interesting is because it would seem that the best way to play to make money would be to play optimally in a probabilistic sense, taking into account only the cards. But the computer exploits irrationality, as well, which leads to a better result.

Similarly, with the current discussion, while a chess program may be able to evaluate the "quality" of a player's moves taking into account the board position, it's possible that there are human tendencies that are irrational, which the computer program does not take into account.

So maybe a computer program says that Pro X played with better quality than Pro B. But it might be the case that Pro B was, in fact, more adept at taking advantage of human irrationality, and was more successful as a pro. Which of the two pros played with better "quality"? It is hard to say, because Pro B might be able to expect better results.

Kirby: it's an interesting point, but I think it's probably pretty marginal for Go or Chess. Sometimes, it might be good to play marginally to provoke a fight (it seems Lee Sedol does this--I've heard his positional judgment is not good for a pro).

I guess my point with poker is that I never would have thought that best play would be based exclusively on probabilities. Bluffing is just essential.

Huh, I don't really see why black chose to resign there...

daniel_the_smith wrote:Huh, I don't really see why black chose to resign there...

White is up by almost 10 points on the board before komi. On the other hand, if you mean "why this move rather than the previous or the next" I have no idea. It felt like the right time, I guess.

One reason that complicated games are exciting is that there is always the chance that one of the players will misread or otherwise lose control and have his game collapse. Couldn't it be true that in a long (two day) game the players would be able to read out the complications and chaotic fights? I think I read somewhere that in the long Japanese games there are fewer chaotic fights (ran sen) because the players have time to read them out and see that no decisive advantage results. These kinds of fights are, paradoxically, more likely to occur in Japanese speed games, as in the TV tournaments. I suppose then the player starting the "chaotic" fight is gambling that in the speed situation he is less likely to make a fatal mistake than his opponent.

Kirby wrote:

hyperpape wrote:...
Maybe I've misunderstood, but the point about poker seems obvious. I feel like it only sounds notable because we tend to have associations with "optimal" play.

The reason that I find the part about poker interesting is because it would seem that the best way to play to make money would be to play optimally in a probabilistic sense, taking into account only the cards. But the computer exploits irrationality, as well, which leads to a better result.

Isn't what makes poker interesting the people, the mind games, the struggle to read your opponent, and the probability calculations are a nice add-on? Which is all of limited use in our favorite game.

Kirby, I'd take my comments about computers in chess with a grain of salt: http://en.wikipedia.org/wiki/Methods_fo ... er_choices.