Page 1 of 1
PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 5:42 am
by SuprunP
Hello everyone!
The following is the game between my two AI Go programmes (the most difficult levels). Would you be so kind to comment on it and tell me whether they are any good to study their moves? (I've highlighted some points in the game that are especially confusing to me, could you please elaborate on those?)
Thanks.
(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2]
RU[Japanese]SZ[9]KM[6.50]
PW[White PC]PB[Black PC]DT[2013-02-15]RE[W+5.50]ZT[120]
;B[ee]
;W[cd]
;B[cf]
;W[ec]
;B[fc]
;W[gg]
;B[gf]
;W[ff]
;B[dc]
;W[dg]
;B[ef]
;W[eg]
;B[fe]
;W[cc]
;B[fg]
;W[fh]
;B[cg]
;W[df]
;B[de]
;W[hf]
;B[db]
;W[be]
;B[ce]
;W[cb]
;B[bf]
;W[ae]
;B[he]
;W[hg]
;B[dh]
;W[eh]
;B[ch]
;W[eb]
;B[dd]
;W[da]
;B[ac]
;W[ab]
;B[bb]
;W[ba]
;B[aa]
;W[ca]
;B[bd]
;W[ea]
;B[af]
;W[bc]
;B[ei]
;W[ff]
;B[ge]
;W[fi]
;B[gb]
;W[ga]
;B[ha]
;W[ab]
;B[ad]
;W[di]
;B[ci]
;W[fb]
;B[gc]
;W[ie]
;B[id]
;W[if]
;B[hh]C[Is this attack justified?]
;W[ih]
;B[hi]
;W[hb]
;B[fa]
;W[hd]
;B[hc]
;W[ic]
;B[gd]
;W[id]
(;B[ii]C[Why not J8?]
;W[gi]
;B[ig]
;W[ib]
;B[ga]
;W[ih]C[What is the point of this move, isn't that group already dead?]
;B[ed]
;W[gh]
;B[fg]
(;W[ia]C[What is the point of this move? Why not E1, for example?]
;B[ff]
;W[hi]C[This move also baffles me...]
;B[ei]
;W[be]C[This one defies my logical thinking...]
;B[ae]
;W[di]
;B[ah]
;W[ei]
;B[]
;W[])
(;W[ei]
;B[ff]))
(;B[ib]))
Re: PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 6:26 am
by SoDesuNe
Sorry for derailing but, why do you want to study 9*9 games and if you really want it, then why don't you just take pro vs pro games on 9*9 (
GoGoD has around 100)?
Re: PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 6:59 am
by SuprunP
SoDesuNe wrote:Sorry for derailing but, why do you want to study 9*9 games and if you really want it, then why don't you just take pro vs pro games on 9*9 (
GoGoD has around 100)?
Hello,
SoDesuNe.
9x9 doesn't take up as much as time as 19x19 and it's easier for me to concentrate on 81 points, rather than on 361. On top of that I do think that if I'm for some reason unable to make some substantial progress in 9x9 there's practically no point in learning 19x19 and being constantly frustrated (shouldn't the game bring us joy?).
I do have around 300 9x9 pro games in my library, but being able to analyse and adopt, as it were, some of the moves of reportedly 1-5 dan programmes (and what's more, to adjust them to play slightly stronger than me to use them as a sparring partner wouldn't hurt, would it?)
Thanks.
Re: PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 8:02 am
by Kirby
If you approach the games you watch critically - knowing that any given move could be wrong - I don't see how it could hurt to study them. When you see things that seem wrong, there's a good chance they could be.
That said, the more "good" moves that are made, the more you can be exposed to something meaningful.
If you study only computers, you may limit the diversity of moves you are exposed to.
Still, seeing new ideas from either bots, humans, or borgs can't be that bad.
Re: PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 11:48 am
by palapiku
Bots play strange / unreasonable moves when they are losing. This is not very different from humans, except bots have no shame and sometimes play really unreasonable moves. You have to look out for that.
Similarly, bots will play slack moves when they are winning. This is because the bot just needs to win, it doesn't care about the score.
Re: PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 12:00 pm
by skydyr
palapiku wrote:Bots play strange / unreasonable moves when they are losing. This is not very different from humans, except bots have no shame and sometimes play really unreasonable moves. You have to look out for that.
Similarly, bots will play slack moves when they are winning. This is because the bot just needs to win, it doesn't care about the score.
I don't know about you, but I certainly do this too, to some extent.

Re: PCvsPC 9x9 Are they good enough to study their moves?
Posted: Fri Feb 15, 2013 12:16 pm
by yoyoma
Q: Is Black's attack justified?
I think so. Without this White needs 1 move to defend against a seki when White's liberties are gone. With Black playing a few moves in here first, I think White needs 2 extra moves to defend instead of just 1. (see
http://senseis.xmp.net/?Semedori)
Q: Why not J8?
Black is trying to kill the entire white group. Is that better than J8? I don't know. Maybe they both lose so Black is choosing the more complicated way to lose.
Q: What is the point of this move, isn't that group already dead?
No. If White ignores this ko and allows black to fill, it creates a bulky-5. White would be dead. White's connection in the upper right threatens to start a different ko to connect both White groups.
Q: What is the point of this move? Why not E1, for example?
I assume the game is using area scoring, so this move is worth a point. The two kos can be treated as miai and shared.
Q: This move also baffles me...
"Proper" play would be for white to finish the ko. But white doesn't care because he won already. To humanize the bots thinking, it is thinking something like "I win no matter what I do, but I guess I'll just make 3 eyes instead of just 2 to be extra extra sure".
Q: This one defies my logical thinking...
That's a ko threat. White takes the ko back next.
Posted: Sat Feb 16, 2013 5:32 am
by EdLee
SuprunP wrote:On top of that I do think that if I'm for some reason unable to make some substantial progress in 9x9
there's practically no point in learning 19x19 and being constantly frustrated...
This idea is wrong.

Yes, in the very beginning (much fewer than 100 games), the full 19x19 board can be a bit overwhelming to some beginners.
As a side note, there are pros who hardly ever played 9x9 in their
entire career (
including when they were beginners.

)
Just play (9x9, 13x13, or 19x19) and enjoy.