(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2]
RU[Chinese]SZ[19]KM[7.50]
PW[AlphaGo]PB[AlphaGo]C[amazing sacrifice by W to wriggle out of a seemingly difficult situation !!!!!
accurate calculation for immense ko !!!!


highlights:

10-12 alphago's (almost) new joseki, for W to efficiently settle and reduce B, while making it hard for B to find a follow up locally
!!

16-23, 26 a new joseki when your approach stone gets kicked. This idea is greatly developed in other self-play games, see game 11
!!!!

49-71
W seems to be in deep trouble when it's very weak group is cut in two in B's huge area. But W finds a brilliant way to sacrifice half. The precision of 66 is particularly notable
!!!!!

208-287
B goes to kill W's corner. It appears to die, but with all the aji in the B area, W finds an enormous ko. It was a very close game, but now, B has to spend a lot of moves inside its own area to kill W, and W's ko threat to kill the top left is enough to win the game. Eventually 1/4 of all stones played this game are captured!

!!!!

297, 298

alphago does strange things when behind/ahead
Is this a weird bug?
!
]RE[W+Resign]
;B[pd]
;W[dd]
(;B[pp]
;W[dp]
;B[pj]C[alphago does sanrensei too, wow I'm surprised

it leaves both 3-3's open!]
(;W[nc]
;B[pf]
(;W[nq]
;B[pn]C[symmetry!]
(;W[oo]C[nice move

B has kept defending solidly on the right, so B is getting increasingly overconcentrated

This move helps greatly in reducing the B potential in the centre. As B is solid, there may be a lot of potential.]
(;B[po]
(;W[pr]C[settling fast, saying it is the biggest area]
(;B[fc]C[standard]
;W[cf]
;B[oc]
(;W[qc]C[new move!

]
(;B[nb]C[making R17 heavier, so W needs to keep going locally]
;W[rf]
(;B[rg]
;W[qg]
;B[re]
;W[rh]
;B[rj]
;W[fq]
;B[fe]
;W[mb]C[wow]
;B[mc]
;W[nd]
;B[lb]
;W[ec]
;B[fb]
;W[og]
(;B[ma]
;W[oi]
;B[oj]
;W[dk]
;B[dg]
;W[cg]
;B[dh]
;W[ch]
;B[di]
;W[bj]
;B[lg]C[looks very powerful]
;W[mi]
;B[ek]
;W[el]
;B[fl]
;W[em]
;B[nh]
;W[ni]
;B[ph]
;W[pg]
;B[oh]
;W[of]
;B[qf]
;W[pi]
;B[qi]
;W[qh]
;B[mh]
;W[sg]
;B[qd]
;W[li]
;B[se]
;W[le]
;B[md]
;W[si]C[subtle and precise


it seems like W may be reading far into the future when playing these sorts of moves, as it doesn't appear to influence the game in the short term.]
(;B[ri]C[why here?]
(;W[jh]
;B[kf]
;W[if]
;B[ke]
;W[gg]
;B[fk]
;W[dj]
;B[il]
;W[ef]
;B[fm]
;W[id]
;B[hi]
;W[kd]
;B[me]
;W[ll]
;B[gp]
;W[en]
;B[fn]
;W[hc]
;B[ig]
;W[ih]
;B[hf]
;W[hg]
;B[ie]
;W[he]
;B[jf]
;W[gf]
;B[ol]
;W[qq]
;B[cc]
;W[eb]
;B[cd]
;W[de]
;B[db]
;W[ea]
;B[be]
;W[iq]
;B[km]
;W[lm]
;B[jo]
;W[gq]
;B[cq]
;W[rk]C[if this is a good move, it requires deep reading, as it may just become suicide and lose points]
;B[ro]
;W[hp]
;B[ho]
;W[go]
;B[oq]
;W[or]
;B[np]
;W[hn]
;B[hm]
;W[dq]
;B[eo]
;W[do]
;B[mq]
;W[nr]
;B[io]
;W[fi]
;B[dr]
;W[er]
;B[ej]
;W[gn]
;B[in]
;W[fo]
;B[mo]
;W[lp]
;B[mp]
;W[gi]
;B[hj]
;W[lo]
;B[hh]
;W[if]
;B[gh]
;W[fh]
;B[hf]
;W[nl]
;B[nk]
;W[if]
;B[fg]
;W[om]
;B[pl]
;W[eg]
;B[hf]
;W[pe]
;B[qe]
;W[if]
;B[gm]
;W[fp]
;B[hf]
;W[pm]
;B[qm]
;W[nm]
;B[ql]
;W[kn]
;B[kl]
;W[kk]
;B[jk]
;W[jm]
;B[jl]
;W[jn]
;B[ff]
;W[ge]
;B[mr]
;W[im]
;B[gd]
;W[hd]
;B[eh]
;W[fd]
;B[ed]
;W[ee]
;B[fd]
;W[fa]
;B[kq]
;W[kp]
;B[dl]
;W[cl]
;B[rr]
;W[gb]
;B[kj]
;W[lj]
;B[ki]
;W[lh]
;B[mf]
;W[kh]
;B[hk]
;W[rq]
;B[sp]
;W[sq]
;B[ns]
;W[rs]
;B[qs]
;W[sr]
;B[ss]
;W[bb]
;B[cb]
;W[rs]
;B[bp]
;W[br]
;B[ss]
;W[qo]
;B[rs]
;W[op]
;B[on]
;W[sj]
;B[qj]
;W[rn]
;B[qn]
;W[rp]
;B[qp]
;W[so]
;B[jq]
;W[jp]
;B[qo]
;W[sl]
;B[rm]
;W[sm]
;B[sn]
;W[pq]
;B[no]
;W[rn]
;B[sf]
;W[nj]
;B[sn]
;W[da]
;B[ca]
;W[rn]
;B[mk]
;W[lk]
;B[sn]
;W[bc]
;B[ab]
;W[rn]
;B[lq]
;W[sn]
;B[ir]
;W[hr]
;B[is]
;W[ms]
;B[ls]
;W[kr]
;B[jr]
;W[ks]
;B[os]
;W[ps]
;B[ms]
;W[nn]
;B[oq]
;W[op]
;B[nf]
;W[sp]
;B[mn]
;W[ln]
;B[rl]
;W[sk]
;B[oe]
;W[sh]
;B[oo]
;W[ml]
;B[ng]
;W[mj]
;B[qk]
;W[ok]
;B[nk]
;W[mk]
;B[oq]
;W[mm]
;B[rg]
;W[op]
;B[rf]
;W[nk]
;B[oq]
;W[ba]
;B[bd]
;W[op]
;B[ip]
;W[hq]
;B[oq]
;W[bf]
(;B[qr]
(;W[ae]
;B[cr]
;W[bo]
;B[ei]
;W[ii]
;B[ij]
;W[ci]
;B[jd]
;W[jc]
;B[aa]
;W[ac]
;B[jb]
;W[je]
;B[co]
;W[bn]
;B[jd]
;W[if]
;B[fj]
;W[gc]
;B[cm]
;W[cn]
;B[hf]
;W[kc]
;B[ib]
;W[ic])
(;W[if]
;B[je]
;W[jc]
;B[ae]
;W[ei]
;B[gj]
;W[gc]
;B[ia]
;W[jj]
;B[ij]
;W[ii]
;B[hf]
;W[ib]
;B[ja]
;W[ha]
;B[kb]
;W[jb]
;B[ka]
;W[ld]
;B[lf]
;W[if]
;B[jg]
;W[af]
;B[ac]
;W[hs]
;B[fj]
;W[fg]
;B[hf]
;W[ce]
;B[if]
;W[dc]
;B[aa]TW[ga][fb][hb][fc][ic][ed][fd][gd][fe][df][ff][ag][bg][dg][ah][bh][dh][eh][ai][bi][ci][di][ji][ki][aj][cj][kj][ak][bk][ck][al][bl][dl][am][bm][cm][dm][an][bn][cn][dn][in][ao][bo][co][eo][ho][io][jo][ko][ap][bp][cp][ep][gp][ip][aq][bq][cq][eq][ar][cr][dr][fr][gr][as][bs][cs][ds][es][fs][gs]TB[ba][la][na][oa][pa][qa][ra][sa][bb][mb][ob][pb][qb][rb][sb][bc][nc][pc][qc][rc][sc][ad][nd][od][rd][sd][ne][pe][of][mg][og][pg][qg][sg][qh][rh][sh][si][sj][gk][ik][rk][sk][gl][hl][sl][sm][rn][sn][so][op][rp][sp][nq][pq][qq][rq][sq][kr][lr][nr][or][pr][sr][js][ks][ps]C[B+4.5]))
(;B[ae]C[W just wins the ko locally if B responds]
;W[op]
;B[cr]
;W[hs]
;B[oq]
;W[qr]
;B[rs]
;W[pk]
;B[rr]
;W[lr]
;B[qs]
;W[op]
;B[js]
;W[kr]
;B[oq]
;W[ss]
;B[rs]
;W[ks]
;B[qs]
;W[op]
;B[lr]
;W[kr]
;B[oq]
;W[rr]
;B[qs]
;W[op]C[W wins]))
(;W[rk]C[on other hand, while the aji looks bad, perhaps it isn't so terrible as B doesn't need to play S10]
;B[ol]))
(;B[qj]C[this is B's other choice to kill, and it seems cleaner, so I'm not clear why B doesn't play here


On the other hand, W may have more points if W lives with T11, and B may have less liberties in some variations if there is a capture race, and B needs to play S11 to capture W.]))
(;B[pg]
;W[ph]
;B[oh]
;W[oi]
;B[nh]
;W[pi]C[B's cut means W is mostly alive]
;B[mi]
(;W[pk])
(;W[ob]C[W seems to be threatening to living in the corner in sente]
;B[ma]
;W[pb]
;B[pe]
;W[rb]
;B[nf])))
(;B[rd]
(;W[hc])
(;W[ob])))
(;B[ob]C[it played here game 11]
;W[ec]C[this sort of thing?])
(;B[nd]C[human normal, but less pressure on the corner as W has P18

I have no clue what alphago would do here though]))
(;W[ec]C[the standard human thought process

I can't seen anything too wrong with this]
(;B[fd]
;W[ic]
;B[nd]
;W[ie])
(;B[nd]
;W[fd]))
(;W[nd]C[no good as Q14 is high to attack, and Q10 helps make B too solid, so there is no point playing here, but where else can W play?]
;B[jd]))
(;B[oc]C[if]
(;W[nd]C[B doesn't have a great follow up]
(;B[jd]
;W[hc])
(;B[fc]
;W[hc]))
(;W[qc]C[though even now, alphago would probably 3-3])))
(;W[fq]
;B[oq]))
(;B[qq]C[I'm not actually clear about how to punish this]
(;W[gq]C[it seems W has to defend first, and then this is bad aji so it isn't so easy for B to attack O3, but also when W invades at R8, it is annoying])
(;W[po]C[this may also be a problem as W has the ladder]
;B[qo]
;W[qn]
(;B[qm]C[B seems fine for now]
;W[rn]
;B[pm])
(;B[pm]C[probably not so good]
;W[ro]
;B[qp]
;W[qm]
;B[pl]
;W[rk]
(;B[op]
;W[qi])
(;B[qk]
;W[rl]
;B[rj]
;W[rp]
;B[rq]
;W[sn]
;B[op]
;W[jp])))))
(;W[qq])
(;W[pr]
;B[]
;W[oo]C[later it seems entirely possible B will respond differently]
(;B[on]C[Q4 is lighter now])
(;B[qq]C[dual purpose of defending against P5 and Q2])))
(;W[qc]C[alphago liked to approach and then invade 3-3 in last year's self-play but my feeling that it thinks this makes B too solid, in fact it still wants to aim at the aji at R12, and maybe even threaten to attack B. So it likes to invade 3-3 directly, or not invade at all (at least not yet)]
;B[pc]
;W[pb]C[whichever way B blocks, B gets quite solid. So though W gets quick cash for territory, there is the feeling that it helps B fix the weaknesses. I don't know if this analysis is correct though.

furthermore, we know from master that alphago tenukis here, which makes W's choiceto invade 3-3 look even worse.]
;B[qd]
;W[rb]
;B[rc]
;W[qb]C[comparing to the direct invasion of 3-3, it looks like B Q14 is overconcentrated but perhaps it is important for B to fix that weakness, but W's shape isn't great to go so much on the 2nd line, so it hard to tell.

At least B gets sente.

Hmm, actually I'm not so sure. W does feel content with this.

]
;AE[pb][qb][rb][nc][pc][qc][rc][qd][pf]
;B[]
;W[qc]C[tewari]
;B[qd]
;W[pb]C[a pretty bad move but not too bad]
(;B[pc]
;W[rb]
;B[rc]
;W[qb]
(;B[md]C[this is still good])
(;B[]
;W[nc]C[W getting this makes W happier]
(;B[pf]C[but B plays here???

so submissive])
(;B[pg]C[this seems slightly better]
;W[oe]C[this is annoying though])
(;B[nd]C[but why not more efficient?])))
(;B[nc]C[B should probably follow up])
(;B[md]C[or here])))
(;W[qc]C[a bit less appropriate]
;B[qd]
;W[pc]
;B[oc]C[B hanes to take gote makes more sense with Q10 support]
;W[ob]
(;B[nc]
;W[nb]
;B[mc]C[normal joseki, it is difficult for W not to play S16])
(;B[nb]C[maybe])))
(;B[qp]C[expected for master]))