(;FF[4] CA[UTF-8] GM[1] GN[KataGo 7 komi self-play - game 51] PC[https://online-go.com/review/664786] PB[KataGo] PW[KataGo] BR[9p] WR[9p] OT[Error: time control missing] RE[?] SZ[19] KM[7] RU[NZ] ;B[qd] ;W[dp] ;B[cd] ;W[qp] ;B[fd] ;W[oc] ;B[pe] ;W[nq] ;B[mc] ;W[pk] (;B[pi]LB[cq:A]C[Both players start with a peaceful opening. In particular q11 is a bit surprising, as modern AI likes to enter the 3-3 point (A) early on (all comments on this game by Antti Törmänen, JPN 1p, except where noted otherwise). ] ;W[fq] (;B[nd]LB[dj:A]C[Here, too, traditional go theory would have black play on the left side, at around A, where two corner enclosures are facing each other. Black instead finishes the capture of white's upper-right corner stone with o16. ] ;W[cj]C[From a human point of view, this opening looks more interesting for white, who is developing more quickly around the board. ] ;B[qm]LB[qh:B]LB[qi:A]LB[qj:C]C[This early invasion is presumably to prevent white from playing something like A-B-C, which would be an efficient way to start forming territory on the right side (and to overconcentrate black's upper-right corner formation). R7 can be called a territory-oriented move. ] (;W[rk]LB[qh:D]LB[pi:B]LB[ol:A]LB[qm:C]C[The traditional shape would have been A instead – that way, too, the black B and C stones would be unable to connect. However, A forms a bit of a slow shape and makes it possible for black to consider sacrificing C. S9, by comparison, makes it possible for white to capture C on a larger scale, and black therefore needs to move his stone out. Simultaneously, white also aims at D on the right side. ] ;B[om] ;W[nk] ;B[ml]LB[nl:A]C[This keima seems to invite a white cut at A, but... ] (;W[mk]LB[nl:A]C[...white doesn't go for it. See variation for what might have happened with A. ] ;B[ll] ;W[lk] ;B[kl] (;W[kk]LB[jj:A]LB[kj:D]LB[lj:E]LB[oj:B]LB[ok:C]C[White would have liked to play the A keima immediately, but then there might technically have been a cutting point with black's B-C-l9-D-E. ] ;B[jl] ;W[ij]LB[nd:A]LB[qm:B]C[A peculiar, if straightforward exchange concludes. Both players form their respective walls in the centre, but black's looks more useful than white's: black's is facing the more open lower side, while white's is facing the upper side where black A is already forming a finished shape in the corner. Furthermore, during this sequence black has managed to settle his originally weak B group in an easy fashion. ] ;B[mp]LB[qm:A]LB[mn:B]C[Black decides to further reinforce the A group, while also preventing white from forming territory on the lower side. If this move is omitted, white would have a good peeping move at around B. ] ;W[np] (;B[no]LB[ll:B]LB[mo:A]C[The A extension instead is more often seen, but because of the presence of the B wall black is not scared of white's A cut here. ] (;W[oo]LB[lp:B]LB[mq:A]C[If I were white, I would have forced A-B first before this move. ] ;B[mq]C[Now black's n3 block seals in white's corner efficiently. ] ;W[mr]LB[ok:E]LB[ol:D]LB[mn:F]LB[nn:A]LB[on:C]LB[mo:B]C[Note that white's A-B-C sequence is not scary; black can connect his shape with D-E-F. ] ;B[lr] ;W[nr] ;B[mo]LB[qq:A]C[Finally, both players form beautiful shape in this corner, but personally I would like black's result because of his strong outer wall. Also, there may still be something in the corner with the A attachment later. ] ;W[qf]LB[qc:B]LB[rc:A]LB[rd:C]C[This white move is bigger than it looks, as it is aiming for the A placement in the corner. A is valuable because of its double threat of B and C, linking up to one white stone no matter what. ] ;B[pf] ;W[rc] ;B[qc] ;W[qb] (;B[rd]LB[ob:C]LB[pb:A]LB[rb:B]C[Locally black could consider A-B-C-s16 instead, too, but s16 is arguably more powerful because the white corner is not quite 100% alive. ] (;W[rb]LB[ra:A]LB[sb:C]LB[sc:B]C[For life and death, white A instead looks correct; then, after black B and white C, a ko fight for white's life results. ] ;B[nb] (;W[ob] (;B[qj] ;W[qk] (;B[ni]LB[pc:A]C[Black could technically kill the white corner group with A, so it is interesting why it doesn't do so. See variation. ] ;W[kg]LB[ih:C]LB[ki:A]LB[jk:B]C[L13 is a nice, light move, mainly preventing black from building the upper side with A-B-C next. ] ;B[lg]LB[kg:A]C[M13 seems to be an attempt to try for force white A into bad shape. It is extremely difficult for humans to find these moves. (Comment by Antti Törmänen, JPN 1p) So in a nutshell: P18 is a must for white to avoid black playing P18 himself and removing any defect. And killing white after P18 is slow because it does not kill properly (you 'll need to get back to it when the outside evolves). After that it's normal to have strange moves like M13 because its linked to this idea: allowing black to kill the corner properly, cleanly, by removing the aji outside. (it's easy to see the link between moves like M13 and the aji follow up I ve shown) (Comment by Clydevil, ~2d) ] ;W[lf] ;B[kf] ;W[oh] ;B[ph] ;W[lh] ;B[mg] ;W[jf] ;B[ke] ;W[je] ;B[kd] ;W[oa]LB[kc:A]LB[jg:B]C[White's last sequence seems to have been played with the idea that, since white can no longer get the A pincer, white made sure to get a good timing to play p19 to live in the corner. Of course, from a human point of view, the B cut still looks worrying. ] (;B[kh]LB[jg:A]C[Black, surprisingly, cuts from the other side. See variation for A. ] ;W[jg] ;B[li] ;W[jh] ;B[ki] ;W[jk]C[White now forms ideal shape in the centre – it feels a bit like black helped white, here. ] ;B[ji] ;W[ii] ;B[ik] ;W[jj] ;B[ic]LB[kc:B]LB[jd:A]C[This keima is extremely big. If omitted, white would be happy to get the A push either in sente, or in gote but with the B follow-up in sente. ] ;W[il]C[White then initiates the next fight in the centre. ] (;B[hk]LB[jj:C]LB[hm:B]LB[im:A]C[Black could probably also have sacrificed the stone with A-h9-B instead for an easier line of play. The loss of one stone wouldn't be very expensive, as then white's C stone would also turn out to be unnecessary. ] ;W[im] ;B[po]LB[im:A]LB[nm:C]LB[on:B]LB[lq:D]C[White A is removing some of black group's eye shape, so black promptly plays q5 to reinforce. If omitted, white might start an attack with something like B-C-D. ] ;W[pp] ;B[on] ;W[nl]LB[ok:A]LB[qk:B]C[This turn is both attacking and defending – if omitted, black could wedge at A to capture the B stones. ] ;B[op] ;W[oq] ;B[jn] ;W[jm] ;B[ln] ;W[gk]LB[hk:A]LB[im:B]C[G9 is a flashy move, possibly the highlight of this game. Strategically, white is trying to motivate the black A stones to escape, with the plan of turning the lower-left corner area into territory while black is forced to concentrate on reinforcing A. If black instead chooses to sacrifice the A stones, then the g9 attachment will have weakened them with utmost efficiency. Looking at this position now, black's decision to extend to A looks increasingly mistaken; I would much prefer to have played B instead. ] (;B[cn]LB[hj:A]C[See variation for if black tries to wriggle its way out with A. ] (;W[dd]LB[co:A]C[A human player would no doubt play A here, first profiting from attacking black in the lower-left corner. D16 looks really strangely timed, even if it is a possible move in itself. ] ;B[de] ;W[cc]LB[dc:A]LB[ec:C]LB[ed:B]C[This move can only be played when the A-B-C ladder favours white. ] ;B[dc] ;W[ed] ;B[ee] ;W[ec] ;B[db] ;W[ce] ;B[bd] ;W[eb] ;B[cb] ;W[be] ;B[bc] ;W[fe]LB[ec:A]C[The white A stones cannot be rescued in this game, but white gets a lot of forcing moves on the outside thanks to this cut. ] ;B[ff] ;W[ge] ;B[gf] ;W[he] ;B[gc] ;W[ef] ;B[df] ;W[eg] ;B[dg] ;W[eh] ;B[dh] ;W[ei]LB[be:D]LB[bg:B]LB[hk:A]LB[co:C]C[Locally this result does not look very interesting for white, as white loses sente and also gives black a big corner territory. However, in a global sense, now white has formed some centre territory and the A stones have also been neutralised. Furthermore, now white has an ideal whole-board miai position; white is happy to get either B or C next. (B saves the D stones in a very efficient way because of black's shortage of liberties.) ] (;B[bp]LB[bh:A]LB[co:B]C[Territory-wise, black would really like to jump to A, but that is 'only' endgame; for now, the difference between black b4 and white B is bigger. ] (;W[cm]LB[bg:A]C[White takes some sente profit before moving on to A. ] ;B[bn] ;W[cq] ;B[dm] ;W[cl] ;B[bq] ;W[bg]LB[eb:G]LB[cf:F]LB[cg:E]LB[bh:A]LB[ch:B]LB[ci:C]LB[di:D]C[Black cannot cut white with A, as then white captures black with B-C-D-E-F. However, at the same time white's G stones are already captured, so black doesn't need to respond to b13. ] (;B[gm]LB[hl:A]C[Black tries to force white A in sente, but... ] ;W[br] ;B[do] ;W[fn]LB[hl:A]LB[cn:B]C[...white refuses in a really active way. If black pulls out with A, white looks to start a double attack between A and B. ] (;B[rp]LB[kq:D]LB[lq:B]LB[kr:A]LB[ms:C]C[This early black endgame seems to revolve around whether white will get to force A-B in sente later. If black can get a stone at C in sente, then white A will be less efficient because of black's response of D. ] ;W[ro] ;B[qo] ;W[rq] ;B[pq] (;W[oo]LB[qq:A]LB[pr:B]LB[qr:C]LB[ms:D]LB[ps:E]C[White can just capture black with A-B-C-D-E instead, but then black has achieved its goal of getting D in sente. White will not give a single point away freely. ] ;B[eo] ;W[gn] ;B[op] ;W[sp]LB[ch:C]LB[oo:B]LB[fp:D]LB[pr:A]C[Now black A starts a real ko for white's five stones, but white has no doubt read out that it can win the ko at B. White's ko threats at around C and D are the key to this. ] (;B[gp] ;W[gq] ;B[fm] ;W[in] ;B[en]LB[do:A]C[Now white no longer has ko threats against the A group. ] ;W[el] ;B[em] ;W[jq]LB[kr:B]LB[pr:A]C[Note that this move defends against black's A ko fighting move, as white can connect under with B. ] (;B[hq]LB[hp:D]LB[eq:G]LB[iq:C]LB[gr:F]LB[hr:A]LB[ir:B]LB[jr:E]C[Black aims for A-B-C-D, after which white has a hard time choosing between E and F – either way will leave a weakness for black to exploit, either at E or G. ] ;W[kr] ;B[ip] ;W[lq] ;B[hr] ;W[jo]LB[mo:A]C[This move is sente for now because black A is not completely alive. ] ;B[nm] ;W[dr] ;B[ep] (;W[er]LB[eq:A]C[I would have played A without thinking, but apparently e2 is a slightly more optimal way of forming life. ] ;B[hl] ;W[ko] ;B[qq] ;W[rp] ;B[eq] ;W[gr] ;B[gs] ;W[fs] ;B[hs] ;W[ar] ;B[rn] ;W[cp] ;B[co] ;W[ri]C[From here on, it seems smaller endgame starts. The game is still dead even. ] ;B[fb]LB[rr:A]C[This move looks strange, but it is considerably reducing white's ko threats in this area. Black is aiming to start trouble with A in the lower-right corner. ] ;W[od] ;B[rr] ;W[oo] ;B[di] ;W[dj] ;B[op] ;W[sr] ;B[pr] ;W[oo] ;B[fp] ;W[fr] ;B[op] ;W[ho] ;B[hp] ;W[oo] ;B[dq] ;W[cr] ;B[op] ;W[rs] ;B[hj] ;W[gi] ;B[qs] ;W[oo] ;B[rj] ;W[sj] (;B[op]LB[sq:B]LB[ss:A]C[This capture is strange – black is in a double ko with p4 and A, but black should prefer to capture A instead because of its prospect of living in ko with B. P4 merely loses a ko threat. ] ;W[qr] ;B[dl] ;W[qg] ;B[pg] ;W[pl] ;B[pm] ;W[qn] ;B[pn] ;W[rr]C[White has apparently calculated that it can afford to secure the lower-right corner territory. ] ;B[rh] ;W[jd] ;B[jc] ;W[qh] (;B[ck] ;W[qi] ;B[dk]LB[qi:B]LB[ck:A]LB[dk:C]C[The move order with A-B-C is peculiar – probably all humans would play black B before black A, but presumably KataGo has read out that either way is possible. ] ;W[ci] ;B[oe] ;W[ch] ;B[fc] ;W[bk] ;B[kn] ;W[io] ;B[bl] ;W[pj] ;B[oi] ;W[ek] ;B[nc]LB[ma:C]LB[lb:B]LB[mb:A]C[This connection is bigger than it may seem, as it prevents white's endgame of A-B-C. ] ;W[re] ;B[ie] ;W[id] ;B[bj] ;W[bi] ;B[gj] ;W[fj] ;B[hd] ;W[if] ;B[hf] ;W[hh] (;B[sc] ;W[sb] ;B[ak]LB[sb:B]LB[sc:A]C[A-B-a9 is also a strange order of moves, but doesn't make a difference for the final score. ] ;W[sd] ;B[se] ;W[sf] ;B[aq] ;W[js] ;B[ih] ;W[ig] ;B[pc] ;W[pb] ;B[hg] ;W[gg] ;B[gd] ;W[fg] ;B[gl] ;W[km] ;B[lm] ;W[na] ;B[ma] ;W[fk] ;B[ae] ;W[af] ;B[ad] ;W[ir] ;B[ai] ;W[rl] ;B[rm] ;W[ea] ;B[fa] ;W[ah] ;B[aj] ;W[oo] ;B[lo] ;W[mi] ;B[lp] ;W[kq] ;B[mj] ;W[lj] ;B[mh] ;W[nj] ;B[mi] ;W[qe] ;B[pd] ;W[cf] ;B[da] ;W[cg] ;B[sl] ;W[oj] ;B[ol] (;W[is]LB[op:A]C[White should connect at A instead. ] ;B[op] ;W[bo] ;B[ao] ;W[oo] ;B[iq] ;W[kc] ;B[kb] ;W[lc] ;B[ld] ;W[md] ;B[lb] ;W[la] ;B[mb] ;W[kp] ;B[hm] ;W[hn]C[ -- chat -- GoEtLent: P18. From move 0: R16 D4 C16 R4 F16 P17 Q15 O3 N17 Q9 Q11 F3 O16 C10 R7 S9 P7 O9 N8 N9 M8 M9 L8 L9 K8 J10 N4 O4 O5 P5 N3 N2 M2 O2 N5 R14 Q14 S17 R17 R18 S16 S18 O18 P18 R10 R9 O11 L13 M13 M14 L14 P12 Q12 M12 N13 K14 L15 K15 L16 P19 L12 K13 M11 K12 L11 K9 K11 J11 J9 K10 J17 J8 H9 J7 Q5 Q4 P6 O8 P4 P3 K6 K7 M6 G9 C6 D16 D15 C17 D17 E16 E15 E17 D18 C15 B16 E18 C18 B15 B17 F15 F14 G15 G14 H15 G17 E14 D14 E13 D13 E12 D12 E11 B4 C7 B6 C3 D7 C8 B3 B13 G7 B2 D5 F6 S4 S5 R5 S3 Q3 P5 E5 G6 P4 T4 G4 G3 F7 J6 E6 E8 E7 K3 H3 L2 J4 M3 H2 K5 O7 D2 E4 E2 H8 L5 R3 S4 E3 G2 G1 F1 H1 A2 S6 C4 C5 S11 F18 P16 S2 P5 D11 D10 P4 T2 Q2 P5 F4 F2 P4 H5 H4 P5 D3 C2 P4 S1 H10 G11 R1 P5 S10 T10 P4 R2 D8 R13 Q13 Q8 Q7 R6 Q6 S2 S12 K16 K17 R12 C9 R11 D9 C11 P15 C12 F17 B9 L6 J5 B8 Q10 P11 E9 O17 S15 J15 J16 B10 B11 G10 F10 H16 J14 H14 H12 T17 T18 A9 T16 T15 T14 A3 K1 J12 J13 Q17 Q18 H13 G13 G16 F13 G8 L7 M7 O19 N19 F9 A15 A14 A16 J2 A11 S8 S7 E19 F19 A12 A10 P5 M5 N11 M4 L3 N10 M10 N12 O10 N11 R15 Q16 C14 D19 C13 T8 P10 P8 J1 P4 B5 A5 P5 J3 L17 L18 M17 M16 N16 M18 M19 N18 L4 H7 H6 GoEtLent: P18 GoEtLent: M13 ] ;B[hi]LB[op:A]C[KataGo 50 % White, however, finally has more ko threats than black and so wins the A ko. The game record ends here, although New Zealand rules are used and so the neutral intersections are also worth points. What follows is my assumption of what should happen. ] ;W[op]C[White doesn't have enough ko threats to continue fighting the ko (while also trying to get the last dame), so it has to connect. ] ;B[sn] ;W[sk] ;B[sm] ;W[jp] ;B[so] ;W[go] ;B[fo] ;W[ql] ;B[ok] ;W[kj] ;B[fl]LB[dd:D]LB[bg:E]LB[gk:B]LB[hk:A]LB[cn:C]LB[cp:F]C[The last dame is black, with komi 7 points under area scoring; this means a tie, as expected. While there were many interesting moments early on in this game, to me the main highlight was the play at around A/B. Black's A extension may originally have been a bit questionable, and it gave way to white's brilliant B attachment for starting to take hold the lower-left corner. The following play was interesting, too; in human go, black's C invasion would get severely attacked right away, but white instead started settling the game with the D attachment, eventually creating the miai of E and F. Strategically, at that point it felt to me like white had won the game; but as expected of the AI, black's endgame play was sharp, too, and finally it seems like a draw was unavoidable. ] )(;W[op] ;B[is] ;W[kc] ;B[kb] ;W[lc] ;B[ld] ;W[iq] ;B[kp] ;W[jp] ;B[sn] ;W[fo] ;B[go] ;W[hn] ;B[hi] ;W[fl] ;B[so] ;W[hm]C[KataGo gives white -1.5 pp for 284. P4 ] ))(;B[ak] ;W[sd] ;B[sc] ;W[sb] ;B[aq] ;W[js] ;B[gl] ;W[na] ;B[ma] ;W[fk] ;B[ae] ;W[af] ;B[pc] ;W[pb] ;B[ad] ;W[km] ;B[lm] ;W[rl] ;B[rm] ;W[ir] ;B[ai] ;W[ah] ;B[aj] ;W[cf] ;B[ea] ;W[cg] ;B[gd] ;W[ie] ;B[oo] ;W[sl]C[KataGo gives black -0.1 pp for 233. A9 ] ))(;B[qi] ;W[si] ;B[ck] ;W[dk] ;B[bk] ;W[bj] ;B[ak] ;W[na] ;B[kn] ;W[io] ;B[ma] ;W[nc] ;B[mb] ;W[ld] ;B[le] ;W[md] ;B[lc] ;W[ne] ;B[nf] ;W[ci] ;B[gj] ;W[fj] ;B[hi] ;W[hh] ;B[me]C[KataGo gives black -3.5 pp for 207. R11 ] ))(;B[ss] ;W[op] ;B[sq] ;W[os] ;B[or] ;W[sr] ;B[ie] ;W[jd] ;B[sq] ;W[ns] ;B[ps] ;W[sr] ;B[jc] ;W[id] ;B[sq] ;W[ng] ;B[nf] ;W[sr] ;B[hd] ;W[hf] ;B[sq] ;W[mf] ;B[mh] ;W[sr] ;B[hi]C[KataGo gives black -5.5 pp for 193. T1 ] ))(;W[eq] ;B[cp] ;W[fs] ;B[hl] ;W[ko] ;B[go] ;W[rn] ;B[ar] ;W[bs] ;B[dq] ;W[cr] ;B[fc] ;W[ri] ;B[rh] ;W[od] ;B[oe] ;W[di] ;B[ch] ;W[bh] ;B[kn] ;W[io] ;B[hj] ;W[hi] ;B[qn] ;W[gj] ;B[rj] ;W[oi] ;B[oj] ;W[pj] ;B[nj] ;W[qi] ;B[qh]C[KataGo gives white -2.5 pp for 152. E3 ] ))(;B[pr] ;W[kr] ;B[lq] ;W[ls] ;B[qq] ;W[oo]C[KataGo gives black -37 pp for 143. Q2 ] ))(;B[pr] ;W[kr] ;B[hn] ;W[di] ;B[fb] ;W[ch] ;B[ea] ;W[lq]C[KataGo gives black -14 pp for 135. Q2 ] ))(;W[qq] ;B[eo] ;W[di] ;B[ch] ;W[bh] ;B[fb] ;W[fm] ;B[gn] ;W[fo] ;B[ep] ;W[eq] ;B[fp] ;W[go] ;B[gp] ;W[gq] ;B[hq] ;W[hr] ;B[ir] ;W[iq] ;B[hp] ;W[jr] ;B[dr] ;W[dq] ;B[cr] ;W[is] ;B[hn] ;W[gl] ;B[bs] ;W[rn] ;B[in] ;W[hl] ;B[el] ;W[fl] ;B[ek] ;W[od] ;B[oe] ;W[ri]C[KataGo gives white -4.5 pp for 130. R3 ] ))(;B[hl] ;W[hm] ;B[gl] ;W[gn] ;B[el] ;W[en] ;B[dn] ;W[fk] ;B[em] ;W[io] ;B[cp] ;W[dq] ;B[ar] ;W[cr] ;B[bm] ;W[bl] ;B[am] ;W[jq] ;B[qo] ;W[rp] ;B[kp] ;W[kr]C[KataGo gives black -12 pp for 125. H8 ] ))(;B[bh] ;W[ch] ;B[di] ;W[dj] ;B[fb] ;W[ci] ;B[ea] ;W[do] ;B[cr] ;W[dr] ;B[br]C[KataGo gives black -51 pp for 121. B12 ] ))(;W[bg] ;B[cq] ;W[dq] ;B[cl] ;W[kq] ;B[lq] ;W[iq]C[KataGo gives white -37 pp for 114. B13 ] ))(;B[bh] ;W[cp] ;B[gl] ;W[di] ;B[ci] ;W[bj] ;B[fk] ;W[gj] ;B[dk]C[KataGo gives black -31 pp for 113. B12 ] ))(;W[co] ;B[hl] ;W[in] ;B[bo] ;W[bp] ;B[hj] ;W[gh] ;B[do] ;W[cp] ;B[fi] ;W[ip] ;B[el] ;W[qo] ;B[jq] ;W[fm] ;B[gj] ;W[iq] ;B[fh] ;W[nm] ;B[jo]C[KataGo gives white -1 pp for 86. C5 ] ))(;B[hj] ;W[fi]C[White keeps on heaping the pressure for now. ] (;B[hl] ;W[in]LB[il:A]LB[mo:B]C[This turn is one of the moves white is looking forward to playing if black starts fighting; this reinforces the white A group while weakening black B. ] ;B[gm] ;W[ip] ;B[fj] ;W[gh] ;B[gj] ;W[dl]LB[hk:A]C[Something like this might follow. The black centre group would no doubt survive, but the rest of the lower-left corner area would turn into white territory. Compared to this, it is probably more interesting for black to sacrifice the A stones and live closer to the lower-left corner instead. ] )(;B[gj] ;W[gm] ;B[fk] ;W[gh] ;B[dk] ;W[di] ;B[dm] ;W[fm] ;B[cn] ;W[bo] ;B[ck] ;W[bn] ;B[eo] ;W[do]C[KataGo gives black -5.5 pp for 85. H10 ] )))(;B[im] ;W[hk] ;B[hm] ;W[cf] ;B[ef] ;W[hd] ;B[ch] ;W[dg] ;B[ce] ;W[bf] ;B[hf] ;W[fc] ;B[ec] ;W[jc] ;B[dj] ;W[ci] ;B[di] ;W[ck] ;B[dh] ;W[bh] ;B[jd] ;W[id] ;B[ih] ;W[kc]C[KataGo gives black -1 pp for 73. J7 ] ))(;B[jg] ;W[kh] ;B[ig] (;W[jd] ;B[jc] ;W[ic] ;B[jb] ;W[ge] ;B[id] ;W[ie] ;B[hd] ;W[he] ;B[gd] ;W[gg]C[Something like this might follow. To my eyes, black's upper-side territory looks big, but the AI clearly also values white's centre influence highly. ] )(;W[ge] ;B[hh] ;W[gd] ;B[fc] ;W[jd] ;B[jc] ;W[ic] ;B[jb] ;W[fe] ;B[fh] ;W[de] ;B[dd] ;W[dh] ;B[eg] ;W[df] ;B[dg] ;W[cg] ;B[ch] ;W[bf] ;B[bg] ;W[cf] ;B[di] ;W[bh] ;B[ci] ;W[bi] ;B[dj] ;W[ck] ;B[dk] ;W[mh]C[KataGo gives black -2.5 pp for 61. K13 ] )))(;B[pc] ;W[pb] (;B[sc] ;W[sb] ;B[oa] ;W[pa] ;B[ra]C[White only has one eye locally, and his group cannot escape either. ] ;W[kc]LB[mb:B]LB[nc:A]LB[md:C]C[Here, the AI plays the l17 pincer to aim for the A-B-C cut next. Black has to respond. ] ;B[ld] ;W[hc] ;B[gc] ;W[hd] ;B[ni]C[The game might continue like this, with the white corner group finally captured. ] ;W[lb]C[Note that this peep is white's sente at any time. ] ;B[mb]LB[lb:D]LB[mb:E]TR[nb]LB[gc:G]LB[hc:F]LB[kc:B]LB[hd:H]LB[ld:C]LB[pf:A]LB[ni:K]LB[qj:I]LB[qk:J]C[White n18 would link under, so black has to connect.The previous scenario is close to if A-B-C-D-E-F-G-H-I-J-K got played in this order – the only difference is black's o18 stone, which black got for free, preventing white's sente endgame at n19. Even with the extra stone, white can generally be happy with the B-C-D-E exchange, which 'probably' makes up for black's getting to enclose a big corner territory in the upper-right corner. This seems to be why white doesn't 'need' to defend his upper-right corner for now. ] )(;B[ni] ;W[kh] ;B[sc] ;W[sb] ;B[oa] ;W[qh] ;B[ph]C[KataGo gives black -12 pp for 47. Q17 ] )))(;B[sc] ;W[sb] (;B[pc] ;W[pb] ;B[oa] ;W[pa] ;B[ra] ;W[qh] ;B[qi] ;W[ri] ;B[ph] ;W[qg] ;B[qj] ;W[rj] ;B[pg]C[Now, if black answers P18 it leaves bad aji even if it's dead. (Comment by Clydevil, ~2d) ] ;W[kc] ;B[ic] ;W[pl] ;B[po] ;W[od] ;B[ne] ;W[oe] ;B[nf] ;W[of] ;B[nh] ;W[ng] ;B[mg] ;W[og] ;B[mi] ;W[mf]C[Here is an example of bad aji if we consider black plays freely outside, the corner will wake up and bad stuff will happen. (sometimes it's just shibori, but bad enough for black) (Comment by Clydevil, ~2d) ] )(;B[oa] ;W[pa] ;B[pc] ;W[pb] ;B[ra] ;W[kc] ;B[kd] ;W[jd] ;B[ke] ;W[qh] ;B[ph] ;W[qi]C[KataGo gives black -6 pp for 45. T17 ] )))(;W[qh] (;B[qi] ;W[ri] ;B[ph] ;W[qg] ;B[qj] ;W[rj] ;B[qk]LB[ob:A]LB[kh:B]C[If white plays outside first with this sequence then he cannot play P18 after (now in the picture). If it plays P18 now (instead of R12), black will ignore the corner and severely attack the center around B. If white plays around B (instead of R12), black will remove any agi with A. Kata values a lot this black move at A that settles a shape without any defect. (Comment by Clydevil, ~2d) ] )(;B[ob] ;W[cg] ;B[df] ;W[cf] ;B[ce] ;W[dg] ;B[ef] ;W[gc] ;B[fc] ;W[gd] ;B[gb] ;W[hb]C[KataGo gives white -16 pp for 44. R12 ] )))(;W[ra] ;B[ch] ;W[df] ;B[qo] ;W[po] ;B[rp] ;W[qn] ;B[ro] ;W[qq] ;B[ql] ;W[pj] ;B[rq] ;W[pn] ;B[rn] ;W[rr] ;B[ol] ;W[qi]C[KataGo gives white -3.5 pp for 42. S19 ] ))(;B[pb] ;W[qg] ;B[rd] ;W[oh] ;B[ch] ;W[df] ;B[cf] ;W[dd] ;B[ce] ;W[cc] ;B[de] ;W[ed] ;B[ee] ;W[fc] ;B[gc] ;W[fb] ;B[gb] ;W[db] ;B[fa] ;W[eb] ;B[cn] ;W[bp] ;B[dk] ;W[dj]C[KataGo gives black -5 pp for 41. Q18 ] ))(;W[mq] ;B[pp] ;W[pq] ;B[po] ;W[qq] ;B[lp] ;W[mg] ;B[mh] ;W[lh] ;B[lg] ;W[nh] ;B[mi] ;W[ni] ;B[lf] ;W[mf] ;B[me] ;W[li] ;B[cg] ;W[dc]C[KataGo gives white -2 pp for 30. N3 ] ))(;B[mo] ;W[mq] ;B[kp] ;W[cg] ;B[ni] ;W[kh] ;B[lg] ;W[hc] ;B[fb] ;W[jf] ;B[lh] ;W[kg] ;B[ck] ;W[dk] ;B[cl] ;W[cn] ;B[dn]C[KataGo gives black -3 pp for 29. N5 ] ))(;W[jj] ;B[mp] ;W[mq] ;B[pp] ;W[pq] ;B[po] ;W[qq] ;B[lp] ;W[cg] ;B[mh] ;W[hc] ;B[jc] ;W[dc] ;B[dd] ;W[ec] ;B[cc] ;W[fc] ;B[gd] ;W[hd] ;B[ed] ;W[ib]C[KataGo gives white -1 pp for 24. K10 ] ))(;W[nl] ;B[nm] ;W[mm]LB[qh:A]C[A complicated fight breaks out. White, however, in general has an easier time of it because his A move on the right side can help the white group settle if necessary. ] (;B[mk] ;W[mn] ;B[nj] ;W[pl] ;B[pm] ;W[mj] ;B[lj] ;W[mi] ;B[ni] ;W[li] ;B[kk] ;W[rl] ;B[rm] ;W[qh]LB[mk:A]LB[pm:B]C[In general, this fight is busier for black: black has to look after three of his groups, of which A is getting particularly weak, while white only has two weak groups.Instead of trying to fight with A... ] ;B[pp]C[...The AI would apparently instead prefer to start settling the black group by leaning on white's corner. ] ;W[pq] ;B[qq] ;W[rq] ;B[qr] ;W[pr] ;B[qo] ;W[rp] ;B[op] ;W[lm] ;B[mo]LB[ml:A]C[This is one possibility for how the AI might play out the fight. Arguably black has settled his group quickly, but at the same time white can also be happy with having captured the A stone, and for his generally strong shapes. ] )(;B[op] ;W[oq] ;B[pp] ;W[pq] ;B[qo] ;W[rp] ;B[np] ;W[mq] ;B[mp] ;W[lm] ;B[lp] ;W[lj] ;B[ch] ;W[hc] ;B[jc] ;W[ej] ;B[ec] ;W[kc] ;B[kb] ;W[lb] ;B[lc] ;W[kd]C[KataGo gives white -2 pp for 20. O8 ] )))(;W[ol] ;B[om] ;W[nl] ;B[mn] ;W[mp] ;B[lm] ;W[rk] ;B[dj] ;W[dk] ;B[ci] ;W[ck] ;B[di] ;W[fk] ;B[pp]C[KataGo gives white -2 pp for 16. P8 ] ))(;B[dj] ;W[cl] ;B[kc] ;W[oe] ;B[of] ;W[ne] ;B[pc] ;W[kf] ;B[le] ;W[nf] ;B[lf] ;W[og] ;B[pf] ;W[mg]C[KataGo gives black -1.5 pp for 13. D10 ] ))(;B[cq] ;W[dq] ;B[cp] ;W[cn] ;B[co] ;W[do] ;B[bn] ;W[qh] (;B[nd] ;W[cm] ;B[bm] ;W[cl] ;B[jp] (;W[kb] ;B[lb] ;W[lp] ;B[jn] ;W[hp] ;B[oi] ;W[pi] ;B[lm] ;W[hd] ;B[hc] ;W[bd] ;B[rp] ;W[ro] ;B[id] ;W[cc] ;B[dc] ;W[cb] ;B[oj] ;W[qf] ;B[og] ;W[rc] ;B[rd] ;W[qc] ;B[qo] ;W[rn] ;B[oq]C[KataGo gives black -0.2 pp for 11. C3 ] )(;W[lq] ;B[in] ;W[hp] ;B[lp] ;W[mp] ;B[lo] ;W[mo] ;B[kq] ;W[lr] ;B[rp] ;W[qq] ;B[qk] ;W[ql] ;B[qj] ;W[pl] ;B[pj] ;W[oj] ;B[pi] ;W[qc] ;B[rc] ;W[rb] ;B[pc] ;W[qb] ;B[re]C[KataGo gives black -0.2 pp for 11. C3 ] ))(;B[cm] ;W[dn] (;B[bl] ;W[cr] ;B[br] ;W[dm] ;B[dl] ;W[el] ;B[dk] ;W[nd] ;B[pg] ;W[ph] ;B[md] ;W[nf] ;B[me] ;W[ne] ;B[oh] ;W[mf] ;B[qg] ;W[kf] ;B[je] ;W[og] ;B[oi] ;W[pj] ;B[qb]C[KataGo gives black -0.2 pp for 11. C3 ] )(;B[cr] ;W[nd] ;B[md] ;W[nf] ;B[me] ;W[ne] ;B[pg] ;W[ph] ;B[og] ;W[mf] ;B[oh] ;W[mh] ;B[ke] ;W[rf] ;B[pc] ;W[ob] ;B[pb]C[KataGo gives black -0.2 pp for 11. C3 ] ))))