(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2] RU[Chinese]SZ[19]KM[7.50] PW[alphago]PB[alphago]C["the timing of switching from aggressive fighting to tidying up" Tan Xiao 8p ->9p Liu Xing 7p Fan Hui 2p (Deepmind) "Tan is number 7 in the world, not yet a world champion, but later this month, we wish him luck in the final of the Chunlan cup" (which he won) ] ;B[pd] ;W[dc] ;B[pp] ;W[dq] ;B[do] ;W[co] ;B[cn] ;W[cp] (;B[dn]C[Liu: in all the alphago games, it very quickly connects here, it doesn't tiger, or tenuki - we think B has a lot of options here, but ? Tan: humans would play C15] (;W[fq] ;B[ce] (;W[fd]C[Tan: it's logic is very interesting] (;B[dh]C[Fan: this is also very interesting Liu: isn't this the same opening as self-play game 6 Fan: my understanding is that alphago wouldn't play on the 3rd line here, and also that wherever B protects, there will always be an invasion point, so just protect one side properly first Tan: so it thinks that after B connects at D6, B is thick now? Because we normally think that D6 is a heavy group, and fear being attacked Fan: not exactly, I think it says that W invading around D10 isn't great direction right now - it isn't the biggest area of space to deal with, it is a strategic attitude - that if W invades and B deals with it in sente, B can get to take the biggest points on the board Liu: and B leaves a good follow up at G5 Fan: B is asking W to invade] ;W[dj] ;B[gn] ;W[fj] (;B[hc]C[Tan: this is very aggressive, and surprising, it would be normal to run out for B - what if W seals B in now?] (;W[ge]C[this game it plays a different move Tan: more solid] (;B[gp] ;W[gq] (;B[fb]C[it just jumps G4 once we would be worried that once W gets H3, the G4 exchange is bad for B Fan: possibly B thinks G4 makes B more solid, and it's helpful in the upcoming fight, and that if B doesn't exchange it now, there might not be an opportunity to later. liu: this knights move I find quite shocking. Though alphago gives us a lot of shocking moves, but this move is ... ... Tan: so B plays a move that just makes the shape uncomfortable, and W doesn't yet need to protect, and if W protects, W helps B make shape. Neither side will follow up it is the sort of move that I would discard after a moments thought Fan: alphago particularly emphasises the balance of territory on the lower left, W has a big corner, so B plays more aggressive for territory here.] ;W[id] ;B[ic] ;W[jd] ;B[kb]C[Fan: so B takes back some territory at least Tan: but it still feels very comfortable for W Liu: everyone should know Tan is a player with very good global judgement, a proud disciple of Nie.] ;W[fh] ;B[ff]C[very interesting right? Fan: still the same point - alphago will not directly defend submissively without aiming at opponent weaknesses] ;W[gf] (;B[qf] ;W[hp]C[Tan: it seems W is playing as we are thinking Liu: very solid, and each move aiming at B's thinness. Fan: C15 and D12 is very thin, and G6 isn't alive Liu: as when B needs to make life with the C15 group, B may make W very solid, such as with C10, but when the G6 group wants to make eyes, it is also in the C8 area, so they seem directly in conflict ] ;B[ec]C[B shows its hand Tan: I did think about this, but it didn't seem to be useful Liu: so did I Tan: because the F18 move must be eyeing this cut, but it didn't look useful] (;W[ed]C[so in the actual game W blocks to fight Tan: we really don't have that shape feeling for the double kosumi to C13 and this block seems like it clearly doesn't work, but alphago must have its deep reading] ;B[dd] ;W[db] ;B[eb] ;W[bc]C[liu: but the W corner is unconditionally dead] ;B[bd] ;W[cg]C[Tan: so W is trying to sacrifice the corner and attack the G6 group Liu: and truly for B to kill the corner, we do need to think for a while how to] ;B[dg]C[for example if B presses here] ;W[ch] (;B[ci]C[oh, so B can actually hane here] (;W[bi] ;B[di] ;W[cj] ;B[bj]C[as before, W can't sacrifice the 2 stones] ;W[bh] ;B[bk] ;W[df] ;B[cf] ;W[de] (;B[ah]C[it plays a weird tesuji looking move, but the result is the same as A11 as normal] ;W[cd] ;B[ai]C[Liu: we don't need to ask Fan how the game is going, the answer is always the same Fan: yes, still completely even Liu: so Tan, what do you think Tan: well, when W played H4, I would definitely have chosen W to play, as B was rather thin. But after this variation, B connected up, and W got points in the upper left corner Liu: my feeling is that from this variation, the R14 for H4 exchange is very good for B, as H4 has very little value in points Tan: W seems to have about 30 points B has 15 on the left, it is hard to count the upper right, but saying 15 points and another corner. Tan: it truly does seem even Fan: it seems the biggest gap we face is how we judge strength and weakness Fan: another thing is like how with Chang Hao (game 1), we said only a group with two eyes is solid thickhess. Before, W was connected on the outside, but there was a problems such as at E11, but now we find W is completely alive] (;W[qq]C[Tan: only move :) for us, it is very hard to choose which direction here because it seems B pretty much has to capture the corner now - if B takes influences, not only does it not seem big, but it may be attacked here] (;B[pq] ;W[qp] ;B[qo] ;W[ro] ;B[rn] ;W[qn] ;B[po] ;W[rm] ;B[rp] ;W[sn] ;B[rq] (;W[qc]C[Fan: but alphago very rarely tries to use centre thickness to forcibly make territory, it normally tries to use it by starting a fight Tan: it seems like W clearly has better ko threats Liu: R2 makes B uncomfortable about this joseki, B doesn't even have one ko threat, not even D4] ;B[pc] ;W[pb] ;B[ob] ;W[rb] (;B[rd]C[in the game B gives W life? Tan: my understanding is that this makes the ko lighter for B Liu: so if B wins the ko, B gains slightly, and if B loses the ko, B gains slightly Tan: but I don't see why B exchanges these?] ;W[qd] ;B[qe] ;W[rc]C[Fan: yes I thought was strange, giving W life so I checked] ;B[ql] (;W[pn] ;B[rl] ;W[oc]C[still here!] ;B[pa]C[accepting this shape!] ;W[qb] ;B[nb]C[we just said living in this shape is unacceptable for W here, but W has a plan] ;W[lc] (;B[md]C[Liu: its special shape instinct] ;W[lb] ;B[kc] ;W[ld] (;B[kd] ;W[le] ;B[ke] (;W[ne]C[Fan: it is probably not kosumi as B is not alive at the top yet, and if B gets thick there, B is thick everywhere, and Q6 isn't settled yet] (;B[pe]C[liu: this looks slow, but actually it seems the strongest move, the only move] ;W[nd] ;B[nc] (;W[re] ;B[rf] ;W[sd] ;B[kf] ;W[lf] (;B[ol]C[B says that as W is alive in the corner and not weak in the centre, P16 is only 15 points or so, whereas capturing Q6 should be more like 20 points] (;W[od] (;B[ka]C[this looks strange] (;W[la]C[if W blocks, W may get 2 points more, but if W is surrounded, may have to spend several more moves inside Tan: if I was playing, I couldn't submit and play N19 In a human game, B's move would make W very uncomfortable, a difficult choice, We would think about the variations, whether W can fight to avoid being sealed in, but we wouldn't be able to work out which was better anyway. At the least it would feel bad to play N19 Fan: and if you win your choice would be the winning move, and if you lose, then the losing move :)] ;B[nf] ;W[me] ;B[nn]C[tan: I think W will invade the B area, as getting sealed in at M13 is too painful, but it seems worthless for W to play in the centre] (;W[pi]C[Liu: perhaps W is thinking of all the aji with the Q6 group, such as with S10 Tan: or P6 O5 O7] ;B[nh] ;W[ni] ;B[oh] ;W[oi] ;B[lg]C[not yet sente but] (;W[of] ;B[oe] ;W[mc]C[still very big as B has no more sentes] ;B[ri] ;W[rj] ;B[rh] ;W[qj] (;B[nk]C[Fan: yes, alphago immediately protects] (;W[mi] ;B[lh] ;W[li] ;B[sl] ;W[kh] ;B[kg] ;W[kp]C[counting ... Tan: the territory is very close on the board, even giving B a little for J16, but W is a little thin, such as at D10, or the M11 group Fan: it thinks it is still a very close game] ;B[lk] ;W[kk] ;B[kl] ;W[ng] ;B[og] (;W[lq]C[still snatching territory] ;B[mr]C[accepting a loss] (;W[jk]C[if B does nothing, K12 is a perfect shape] (;B[kj]C[liu: it felt like there was something wrong with W's shape, but to directly cut here!] ;W[mf] ;B[mh] ;W[ki] ;B[jj]C[liu: so when it is B's turn, B does attack, but when W has already defended, B attacks!] ;W[jl] ;B[jm] ;W[hk] ;B[lj] ;W[ih] (;B[mq] (;W[km] ;B[ll] ;W[lr] ;B[ls] ;W[ks] ;B[ms] ;W[jr] (;B[mp]C[Tan: I counted 68 for B and 61 for W, is it really going to be so close? W is still quite thin] ;W[kn] ;B[ho] ;W[ip] ;B[hj] ;W[im] ;B[bn] ;W[bo] ;B[if]C[sente for J11] ;W[jg]C[here wow] (;B[gk]C[threatening K6] ;W[hl] ;B[gj] (;W[jf] ;B[lo] ;W[ko] ;B[gh] ;W[gg] ;B[ej] ;W[fl] ;B[ik]C[very precise] (;W[jn] ;B[gl] ;W[ei] ;B[hh] ;W[hg] ;B[eg]C[precise] ;W[fg] ;B[je] ;W[ie] ;B[ek] ;W[fk] ;B[ji] ;W[jh] ;B[el] ;W[ef] ;B[bg] ;W[gb] ;B[gc] ;W[fc] ;B[dp] ;W[cq] ;B[ga] ;W[pk] ;B[pl] ;W[fm] ;B[em] ;W[be] ;B[jc] ;W[he] ;B[bf] ;W[hb] ;B[ha] ;W[ib] ;B[ia] ;W[ea] ;B[jb] ;W[ad] ;B[ln] ;W[fp] ;B[go] ;W[ep] ;B[oj] (;W[ph] ;B[sf] ;W[se] ;B[io] ;W[sj] ;B[pg] ;W[lm] ;B[mm] ;W[qr] ;B[rr] ;W[an] ;B[am] ;W[ao] ;B[gm] ;W[qh] ;B[qg] ;W[mj] ;B[mk] ;W[pr] ;B[or] ;W[fn] (;B[hm] ;W[il] ;B[in] ;W[ij] ;B[ii] ;W[en] ;B[bm] ;W[si] ;B[sh] ;W[pj] ;B[af] ;W[ae] ;B[ik] ;W[jm] ;B[jo] ;W[jp] ;B[gi]C[B wins the ko and wins by 0.5]) (;B[en]C[better by 1/3 point, so B wouldn't have to depend on the ko to win as now G11 is sente] ;W[in]C[biggest])) (;W[qi]C[better by 1/3 point, but W still loses])) (;W[il]C[if connect] ;B[fi] ;W[fn] ;B[fo] ;W[hh]C[W wants to try and take sente but] ;B[in] ;W[jn] ;B[gl] ;W[gm] ;B[hm]C[a severe ko])) (;W[eg]C[why not here?] ;B[gh]C[liu: for example here?] (;W[jf] ;B[gg]C[problem]) (;W[hg] ;B[jf] ;W[hh] ;B[gi] ;W[gg]))) (;B[hd]C[B can capture but] ;W[jf] ;B[je] ;W[ie] ;B[he] ;W[hf] ;B[jc] ;W[eg]C[huge and W defends agaist E10])) (;B[lo]C[now this is impossible] ;W[mp] ;B[lp] ;W[mo] ;B[ko] ;W[nq])) (;W[lr] ;B[ls] ;W[ks] ;B[ms] ;W[jr] (;B[mp]) (;B[lo]C[Tan: why not here? Fan: we got the order wrong]))) (;B[] ;W[nq]C[Tan: why didn't W jump here, I can't see where it loses anything] ;B[nr] (;W[or]C[Tan: W still threatens this]) (;W[] ;B[lr]C[Fan: alphago thought this is a bad exchange for W] ;W[kr] ;B[jr])) (;B[lr] ;W[kr] ;B[jr] ;W[nr]C[there is a problem here, so M1 may be sente] ;B[mq] ;W[qr] ;B[rr] ;W[nq] ;B[mp] ;W[pr]C[fan: oh wow, I had still been wondering why as well])) (;B[jh]C[so it is natural to atari but that just helps too as B still has a cutting point at K13 and M8])) (;W[lr] ;B[ho] ;W[ip] ;B[gj]C[alphago still planned this])) (;W[kj]C[if just protect? alphago thought this was also good] (;B[lq] ;W[lp] ;B[mq] ;W[kq] ;B[mp]) (;B[mq] ;W[lq])) (;W[mq] ;B[io] (;W[kj] ;B[jp] ;W[jq] ;B[iq] ;W[jr] ;B[hq] ;W[hr] ;B[jh]) (;W[jq]C[alphago thought this game was also too close to call] (;B[gj]C[liu: woah tan: it seems so strange, as G11 is not sente either] ;W[jh] ;B[kj] ;W[ki] ;B[fk] ;W[ej] ;B[gi] ;W[gh]) (;B[jh]C[for us, we would expect this] ;W[ki] ;B[ij])))) (;W[li]C[Tan: I had thought the shape was jump] ;B[sl]C[then alphago would play here])) (;B[] ;W[on]C[Tan: isn't there a problem here] ;B[no] ;W[pl])) (;W[] ;B[mc] ;W[na] ;B[oa] ;W[qa] ;B[ma] ;W[mb])) (;W[lg]C[too slow] ;B[oi]C[B is almost completely alive] ;W[ea] ;B[gc] ;W[fc] ;B[ga] ;W[jc] ;B[jb] ;W[ia]C[but the liberties of the W stones are very short])) (;W[ma]C[B is unconditionally better if here as B saves a ko threat and a liberty])) (;B[la]C[natural] ;W[ma] ;B[ka] ;W[mb])) (;W[nn]C[if W responds] ;B[od] ;W[kg] ;B[jg] ;W[lh] ;B[jh] ;W[pi] ;B[oh])) (;B[od]C[if B captures, ?] ;W[kg] ;B[jg] ;W[lh]C[liu: Q11 and K12 seem miai])) (;W[od] ;B[me] ;W[mf] ;B[mc]C[this would be terrible for W])) (;B[nd]C[Tan: there is also this move] ;W[od] ;B[pe] ;W[oe] ;B[mb] ;W[me] ;B[la] ;W[re] ;B[rf] ;W[sd] ;B[kf]C[we had valued the B top and corner as 20 points and B is still not alive, so the W thickness is still showing its value])) (;W[nd]C[Tan: what is B afraid of here?] ;B[nc]C[Tan: probably this] ;W[od] ;B[me] ;W[pe] ;B[lf]C[Tan: though it seems this is also possible both sides gain a lot in their own way, this requires some judgement])) (;B[od]C[Liu: what if B just defends here, because isn't B alive already?] ;W[me] ;B[jc] ;W[ne] ;B[nd] ;W[re] ;B[rf] ;W[sd]C[this does look very good for W Fan: like we said before, alphago can't accept being forced to spend a move on the inside like K17])) (;B[lb]C[if B answers] ;W[nc] ;B[mc] ;W[md] ;B[mb]C[Tan: B could submissively play this if it wasn't for the fact that the Q15 cut seems to make B fall apart] ;W[pe] ;B[od] ;W[nd] ;B[oe] ;W[pf]C[Tan: even if W dies in the corner, W can sacrifice very happily, this will be a disaster for B])) (;W[qk]C[weren't we always taught to play this?] ;B[rl] ;W[rk] ;B[pl] ;W[sl]C[liu: but have you thought about this problem] ;B[pa]C[isn't Q9 a ko threat?] ;W[re] ;B[rf] ;W[sd]C[W lives painfully and T14 is sente] ;B[pk]C[tan: oh, now I understand after hearing Liu's explanation, I didn't even need Fan to teach me here! Fan: well, I wouldn't have been able to explain it! Liu: you didn't investigate this Fan: I was getting panicky just now!])) (;B[pa] ;W[qa] ;B[qb] ;W[qr]C[comfortable] ;B[rr] ;W[pb] ;B[cc] ;W[cb] ;B[qb] ;W[be] ;B[bf] ;W[pb] ;B[dp] ;W[oa]C[once W captures, the whole B stones look weak after O18 and R12, and on the lower left, it is hard to choose from C3 or E3]) (;B[ql] ;W[pn] ;B[rl] ;W[oc]C[W was planning this Liu: somehow B doesn't have the ladder] ;B[pa] ;W[qb] ;B[nb] (;W[re]C[Tan: now I start to understand] ;B[rf] ;W[qe] ;B[pe] ;W[pf]C[B's shape is not good, and W has M17 too]) (;W[qd]C[if W still did this in the game it would be as if W lived like this] ;B[qe] ;W[re] ;B[rd]C[even this throw in is a good exchange] ;W[rc] ;B[rf] ;W[sd]C[as T14 is sente, and B has sente to fix the outside]))) (;W[qh]C[Tan: I had felt though that W might make a bit of a moyo like this though])) (;B[qp] ;W[pq] ;B[oq] ;W[or] (;B[nq] ;W[nr] ;B[mq] ;W[pk]C[it doesn't seem like there is much point for B to take this as B is not very solid, and W has the background of the centre thickness]) (;B[nr] ;W[nq] ;B[op] ;W[mr] ;B[pr] ;W[ns] ;B[qr]C[Tan: perhaps B can say that B still has the potential of the right side and the lower side is not so big? Fan, Liu: but perhaps due to W's great centre power, if B tries to build the right side, it just helps W, whereas if B blocked the other side, W's ponnuki might still be attackable Fan: the difference between the two blocks won't be much, but it is just that alphago slightly leans towards the other one]))) (;AE[pd][qf]AB[pc][qe]C[Tan: B would definitely prefer to shift the stones for territory in this position])) (;B[cd]C[B can't connect anymore] ;W[eg]C[B can't even connect up underneath, so B falls apart] ;B[ai] ;W[ag])) (;W[di] ;B[bi]C[Tan: counter-sacrifice strategy] ;W[df] ;B[cf] ;W[eg] ;B[bg] ;W[eh] ;B[bh]C[Tan: now that B doesn't have to short the liberties and B leaves a back door for D6, this is far too much better for B Liu: it is like W sacrificed the corner for no reason Fan: the territory loss is not just a little now.])) (;B[cf] ;W[di] ;B[cc] ;W[cb] ;B[cd] ;W[bb]C[it seems like B can kill W, but gets totally sealed in and so W gets solid, D6 totally loses the option of connecting up to the other group and the territory from killing the corner isn't actually that big it seems like both sides are striving for particularly high efficiency of every stone Such as how B wanted to connect up to make H4 look inefficient Liu: or like someone said, against alphago, if it plays to sacrifice its stones, you've probably already lost the game. Fan: you don't fear it fighting, but you fear that it sacrifices Liu: its sacrifice ability is too strong Fan: really, its judgement is too strong])) (;W[dd]C[it seems W can just defend] (;B[ed] ;W[de] ;B[ee] ;W[df]C[this is very comfortable for W Fan: oh no, this ... Tan: B is all on dame Fan: and F14 looks silly Liu: it doesn't make sense]) (;B[cg]C[alphago was preparing this weird move ] (;W[ci]C[W can't play here] ;B[ed] ;W[de] ;B[df] ;W[ee] ;B[fg]C[W in trouble as G13 and F15 are miai]) (;W[df] ;B[bj]C[Fan: once B connects up, H4 looks less efficient Tan: but for me, as W to see B directly connect up on the second line, I would feel quite happy, as B still have many points, and there is still aji with the attachment Fan: but for alphago, if B connects up, it thinks W spent too many moves locally without much gain])))) (;B[ip]C[Liu: I wanted to play here Tan: me too Fan: even me too Tan: because Q16 now has the support of L18, so it seems smaller] ;W[qq]C[alphago would play here, it thinks this is good timing, and it seems the R14 move has something to do with the variations on the upper left it isn't that it will affect the upper left fighting, but more that after that fighting, it will be in a good position Tan: because it feels like if W is to apporach R14 or invade R17, it seems smaller than to play in the lower right, normally we say to play in the wider area])) (;B[ip]C[still very big]) (;B[cc]C[normally we would think to settle this group] ;W[cb] (;B[bc]C[alphago descends directly] ;W[dd]C[alphago says this is not even, actually a bad move] ;B[bh]C[woah Tan: but D16 makes B so thin]) (;B[dd]C[generally it doesn't play here because W can B17, though even if W goes E17, W is fine] ;W[bc]C[Liu: I am slightly shocked W will atari here] ;B[ec] ;W[cd] ;B[db] ;W[cc] ;B[ed] ;W[be] ;B[cf] ;W[gh]C[though W is split up, W's shape is still good])) (;B[fh]C[Tan: or jump out as H17 for G15 seems to be a good exchange]) (;B[dd]C[tan: or here])) (;B[fb] ;W[fo]C[perhaps W is scared of this?] ;B[fn] ;W[em] ;B[ep]C[but B has this, and W probably has to defend in gote] ;W[fp] ;B[en] ;W[ip])) (;W[fh]) (;W[fg]C[Liu: in self-play game 6, the same position, W turned to seal in, very normal, but the next move was so shocking] ;B[hg]C[liu: it's hard to understand this move just from looking at it tan: it is very hard for us to think of this move with our logic liu: even if we think of it, we still wouldn't have dared to play it normally, we would think that we don't want to be sealed in by F13 as we have to leave and make W more solid, but the meaning of this move is to eye W's weaknesses first, and ask W to defend, and get W to help B defend B's weaknesses like this. Fan: yes alphago never directly defends, but tries to defend by attacking first. We can notice this concept in almost every game it plays])) (;B[fh]C[fan: given that W turns to seal in, it isn't as if this is bad direction, some games B may play this, some B may play H17])) (;B[ci]C[Liu: this would be more normal]) (;B[di]C[or here Liu: fan's meaning seems to be that this is also ok, but it seems to leave problems still, and doesn't impact much on W so] ;W[qc]C[even if W tenukis, when it is B's turn to play on the left, it is harder to find a good move])) (;W[di]C[Tan: what if pincer Fan: alphago would pincer C11, or D12 or D11 normally] ;B[ed]C[and press as usual] ;W[ec] ;B[fd] ;W[gc] ;B[gd]C[alphago would probably play here])) (;W[cf] ;B[eq] ;W[er] (;B[dp] ;W[cr] ;B[fq] ;W[fr] ;B[hq]C[Liu: isn't this very similar?]) (;B[fr]C[Fan: I've wondered the same, but alphago will play here] ;W[ep] ;B[fq] ;W[dp] ;B[go] (;W[en] ;B[ck]) (;W[dr]C[Tan: what if W just connects? Fan: I'm not actually sure] ;B[cj]C[perhaps we can compare this to the earlier variation B is faster here, W has more liberties] ;AE[cj][go][fq]AW[eo]AB[dm][ep][gq]C[if this is good for B, perhaps the above variation is fine for B too])))) (;B[ce]C[Tan: high approach is also possible, but this is more common] ;W[dn]C[alphago will cut] (;B[ed] ;W[ec] ;B[fd] ;W[gc] ;B[cm] ;W[en] ;B[dk] ;W[qq]) (;B[fd]C[humans have played this Fan: I haven't checked this] (;W[ed]C[Tan: we play here]) (;W[de]C[Fan: alphago would probably play here it doesn't like to be sealed in] ;B[df] ;W[ee] ;B[cd]))) (;B[dm]C[Fan: I am also enthusiastic about this question. When we were young we were taught this joseki, and for a while, humans almost stopped playing D6 connect. But slowly now, we connect again. And against Lee Sedol, alphago played tiger. But gradually alphago stopped playing tiger. Normally in response to tiger, alphago will tenuki or play F4] (;W[cf]C[in this case, alphago would play here] ;B[eq] (;W[er]C[alphago would play very simply] ;B[dp] (;W[cr]C[Tan: oh this?] ;B[fq] ;W[fr] ;B[hq]C[Tan: oh? probably if W played C3, B would extend Fan: alphago thinks this position is already better for W Tan: we normally think that getting sealed in like this so early isn't good for W Fan: but in the context of the 3-3 invasion, W is thinking W got 2 moves on the outside, and lives in this corner in sente. Fan & Liu: and this corner life seems to be a lot better than the 3-3 invasion Fan: and W's C14 reduces the value of B's influence perfectly Liu: Ok, I gradually understand Fan's meaning, but I have a new question - if B connects at D6, why can't W tenuki? There doesn't seem to be much difference? ]) (;W[cq] ;B[fq] ;W[fr] ;B[hq]C[now, G3 is more comfortable])) (;W[dp]C[Liu: we thought W should generally fight back] ;B[ep] ;W[eo] ;B[dn] ;W[er] ;B[fr] ;W[dr] ;B[gq]C[perhaps B's shape outside is alright, and B6 is sente due to B3 placement. W is sealed in])) (;W[fq]C[fan: alphago will basically not play here])))