(;FF[4]
CA[UTF-8]
GM[1]
GN[KataGo 7 komi self-play - game 74]
PC[https://online-go.com/review/773159]
PB[KataGo]
PW[KataGo]
BR[9p]
WR[9p]
OT[Error: time control missing]
RE[?]
SZ[19]
KM[7]
RU[NZ]

(;B[pd]
;W[pp]
;B[qn]
;W[dp]
;B[dd]
;W[nq]
;B[qf]
;W[cf]
;B[ch]
(;W[fq]C[Playing away is rather difficult to do as a human (all comments on this game by Tang Jie, CHN 5d).
]

;B[ce]
;W[df]
;B[fd]
(;W[dh]
(;B[di]C[Because why would you want to hane here?
]

;W[eh]
(;B[cn]C[The game move is kind of lackluster. Black plays away instead of playing more severely.
]

(;W[hc]C[But playing here as white instead is also fine.
]

(;B[ei]C[But black pushes up.
]

(;W[ci]TR[ch]TR[di]TR[ei]C[In the game, white takes the opportunity to make some exchanges first and making black's shape here thin.
]

;B[cj]
;W[bi]
;B[dk]
;W[bj]
;B[bk]
;W[cl]
;B[ck]
(;W[fi]
;B[bh]
(;W[be]C[There's aji in the corner, so white is definitely preserving the aji with the hane here instead of allowing black to connect underneath.
]

;B[bd]
(;W[gg]
(;B[gb]C[This is kind of a lackluster move in my opinion. It's a very human way of playing.
]

(;W[ql]
;B[he]
(;W[jd]
(;B[lc]
;W[if]
(;B[qj]
(;W[nc]C[This is a very AI thing to do, I think.
]

;B[oc]
;W[ej]
;B[dj]
;W[bg]
;B[aj]
;W[nd]
;B[le]
(;W[ng]C[So in the game, we have white doing a two-space, which kind of avoids the issue. This is brilliant.
]

(;B[on]TR[lc]TR[oc]TR[pd]TR[le]TR[qf]TR[qj]SQ[qn]C[So in the actual game, black ignores all the aji and plays this move. Black is kind of balancing the fight between the two triangle groups and bringing the square group out. AI likes to make the game very difficult for itself, so instead of settling a position and then starting elsewhere, it's increasing the danger here.
]

;W[hb]
;B[gc]
;W[hd]
;B[fe]
(;W[lg]TR[lc]TR[le]C[So in the game, white covered to attack these stones. The aji in the top left still isn't gone, there are some variables in there.
]

;B[nb]
;W[mb]
;B[ob]
;W[mc]
;B[rp]
(;W[pg]C[This is kind of a major move in the game, I think. White lost a few percentage points here. (Comment by Maharani: according to high playout KataGo, this move lost 0.13 pp for white.) White playing here is actually a very big move. White is protecting a threat that doesn't seem very bad, but I also think that it played it because of the policy network. (Comment by Maharani: the policy network evaluated Q13 at 5 % and R3 at 38 %.) White was ahead, so now it's playing safe. This is one of the moves I disagree with.
]

(;B[gq]C[This is kind of more of an AI thing.
]

;W[fp]
;B[of]
(;W[oe]TR[nf]C[So white dodges this issue just by playing here first and then connecting at O14. Very nice technique from both players.
]

;B[pf]
;W[nf]
;B[lq]
;W[qq]
;B[rq]
;W[no]
;B[oo]
;W[op]
;B[mp]
;W[np]
(;B[iq]C[Instead, black defends the group on the bottom, which does make sense. I think it's a great move.
]

;W[ol]
;B[qr]
(;W[rn]C[But because of this move, black responded. S6 isn't a very good move in terms of aji-keshi.
]

;B[qo]
(;W[nn]
(;B[pi]XX[pg]TR[qj]SQ[pr]C[In the game, black just plays here directly and it baffles me. It feels like the wrong direction for me, just because the R10 stone is actually really resilient in terms of fighting here. Maybe I'm not strong enough to judge something like that, but I feel that Q2 would definitely be the right direction for black, forcing white to connect. There's a lot going on in the local position, but I think Q11 is too slow, and allowing white to finally take Q2  kind of made the game harder for black; but black does gain in terms of points around Q13, which black can potentially capture.
]

;W[pr]
;B[qs]
;W[bc]
;B[ae]
;W[af]
;B[bf]
;W[pe]
;B[qe]
;W[be]
;B[ie]
;W[je]
;B[bf]
;W[qd]
;B[qc]
;W[be]
;B[ad]
;W[cc]
;B[cd]
;W[db]
;B[bf]
;W[kp]
(;B[kq]
;W[be]
;B[eb]
;W[ec]
;B[fc]
;W[ea]
;B[fb]
;W[ac]
(;B[dc]C[So black taking the eyespace here is very big, because instead of making the whole thing a ko, potentially black could live locally, but also force the shape.
]

;W[bb]
;B[ca]
;W[da]
;B[ba]
;W[aa]
;B[ab]
;W[rd]
;B[od]
;W[ne]
(;B[qg]C[So you have to respond this way.
]

;W[aa]
;B[bf]
;W[cg]
;B[ai]
(;W[oj]
;B[ln]
;W[lo]
(;B[oi]C[Black pushed up, which makes sense to me.
]

;W[ip]
;B[hr]
;W[rm]
;B[sr]
(;W[hp]
;B[mo]
;W[hq]
;B[jr]
;W[gr]
;B[gs]
;W[fr]
;B[mn]
;W[nm]
;B[ll]
;W[rc]
;B[qb]
;W[rb]
;B[ra]
;W[re]
;B[rf]
;W[na]
;B[se]
;W[og]
;B[ab]
;W[ps]
(;B[qh]C[Black is getting rid of ko threats at this point and then starts the ko.
]

(;W[aa]
(;B[qp]
;W[mr]
;B[ab]
(;W[jk]C[Potentially there just weren't enough ko threats, so a move like this ko threat wasn't very good.
]

(;B[lj]C[But the idea is that black is now surrounding the right side.
]

;W[jp]
;B[lr]
;W[jq]
;B[nj]
(;W[nk]
;B[mj]
;W[pq]
;B[pm]
;W[qk]
;B[cb]
(;W[dn]
;B[ir]
(;W[cm]C[This is an absolutely terrible move. There's no reason to play like this. Black makes a forcing move but can just descend, and now the aji in the corner is just surreal.
]

;B[fj]
;W[gi]
;B[bn]
;W[bm]
;B[cq]
;W[cp]
;B[bp]
;W[br]
;B[am]
;W[al]
;B[an]
;W[co]
;B[bo]
;W[bl]
;B[dm]
;W[dl]
;B[em]
;W[el]
(;B[fl]C[Probably the hardest part in the game is understanding that this hane is required. It makes sense.
]

;W[fm]
;B[en]
(;W[fn]
;B[eo]
(;W[pa]C[At this position, the AI graph jumps. Ultimately, the policy network kind of crashed and burned at this point. (Comment by Maharani: High-playout KataGo gives white -2.3 pp for 198. Q19. The "crash and burn" happened after move 154. Q1, which lost 25 pp for white.)
]

;B[pb]
;W[po]
;B[gl]
;W[rs]
;B[pk]
;W[pl]
;B[bq]
;W[fo]
;B[ek]
;W[cr]
;B[gf]
;W[hg]
;B[id]
;W[ic]
;B[jc]
;W[kc]
;B[kh]
;W[ki]
;B[fg]
;W[fh]
;B[lh]
;W[li]
;B[mh]
;W[mi]
;B[jb]
;W[ep]
;B[kd]
;W[do]
;B[jg]
;W[sa]
;B[sb]C[Black 98 % (KataGo)

-- chat --
Maharani: Video review by Tang Jie, CHN 5d: https://drive.google.com/file/d/1oAA4hhl65ApoR3250G4TxM-wpuCtQm2f/view?usp=sharing

]

)(;W[fk]C[White should definitely play the atari. It's needed.
]

(;B[gl]
;W[ek]
;B[fo]
(;W[hm]C[We need to jump out.
]

;B[hl]
;W[hn]
;B[gj]
;W[gk]
;B[hj]
;W[hk]
;B[ik]C[Now black can capture the left side group, so at this point, white has messed up terribly.
]

)(;W[do]C[I wanna say you can technically connect here as white.
]

;B[go]C[The possible issue is, black can play here.
]

;W[bq]
;B[ep]
;W[dq]
;B[eq]
(;W[dr]
;B[er]
;W[es]
;B[gp]C[This becomes an atari.
]

;W[fs]C[And you can't connect.
]

;B[ds]
)(;W[cr]C[So white can capture the corner, but you lose almost everything.
]

;B[dr]
;W[cq]
;B[er]C[So for example, we play here.
]

;W[gp]TR[lo]TR[fp]TR[hp]TR[ip]TR[jp]TR[kp]TR[fq]TR[hq]TR[jq]TR[fr]TR[gr]C[And then you capture? That might work, but also this group is now in danger.
]

)))(;B[bq]
;W[dr]
;B[cr]
;W[ni]
;B[ks]
;W[km]
;B[ml]
;W[be]
;B[ek]
;W[gl]
;B[bi]
;W[cs]
;B[dq]
;W[rs]
;B[rr]
;W[ep]
;B[bs]
;W[as]
;B[aq]
;W[ki]
;B[gj]C[KataGo gives white -0.3 pp for 198. F9
]

)))(;W[fk]C[Because now, white needs to play this.
]

(;B[gl]
;W[ek]C[And then connect.
]

;B[gj]C[Normally, this one would work, but the two stones would end up in atari, so it's not ideal.
]

;W[gk]
(;B[hk]
;W[hj]
)(;B[hl]C[Black would have to extend first.
]

;W[fn]
;B[eo]
;W[fo]
;B[hj]TR[fn]TR[fo]C[And then if white plays this way, black would want to play this one.
]

;W[ep]C[And then white ataris.
]

;B[hk]C[Black plays the atari.
]

;W[do]TR[cf]TR[df]TR[bg]TR[cg]TR[gg]TR[dh]TR[eh]TR[fi]TR[gi]C[White captures and then black goes for an attack on the E12 stones.
]

))(;B[cr]
;W[bq]
;B[bs]
;W[ar]
;B[cs]
;W[km]
;B[ml]
;W[om]
;B[so]
;W[ap]
;B[do]
;W[dq]
;B[as]
;W[aq]
;B[gl]
;W[ek]
;B[go]C[KataGo gives white -3.5 pp for 196. F9
]

)))(;B[en]C[If you play here now, I've made a sequence that I've read out.
]

;W[fl]
;B[cr]
(;W[km]TR[jk]TR[cl]TR[dl]TR[el]TR[fl]XX[gl]XX[hl]XX[il]SQ[jl]XX[fm]XX[gm]XX[hm]XX[im]SQ[jm]XX[fn]XX[gn]XX[hn]XX[in]SQ[jn]XX[fo]XX[go]XX[ho]XX[io]SQ[jo]TR[fp]TR[hp]TR[ip]TR[jp]C[What white has is this tremendous, flabbergasting move here. It's mindblowingly good. I don't think I could find it myself. It's funny how this move affects the lower left corner, and that's just because with the thickness of both groups, L7 creates a wall here. So now black would have to find a second eye in the G5 area.
]

(;B[bq]C[If we protect ourself by playing something like this, white can actually just come and cut.
]

;W[lm]
;B[mm]
;W[ml]
;B[ks]C[And then we need to live on the bottom.
]

;W[ni]C[And then white cuts here.
]

;B[nh]
;W[mi]
;B[mh]
;W[li]
;B[lh]
;W[ki]
;B[kh]
;W[jh]C[What white is doing is forcing black to give up stones.
]

;B[oh]C[Eventually, black will have to connect here.
]

(;W[kg]C[And then white can make this.
]

)(;W[lf]TR[hi]TR[ii]TR[ji]TR[ki]TR[li]TR[mi]TR[hj]TR[mj]TR[hk]TR[mk]TR[hl]TR[il]TR[jl]TR[kl]TR[ll]TR[ml]C[Or play the bamboo, depending on the cut. And then you'll notice that white actually has a very large center here, which could bring the game back to even, but black is just alive everywhere in terms of points.
]

))(;B[ml]C[It would force black to connect here.
]

;W[bq]C[And then here, we have cutting in the bottom left.
]

;B[bs]
;W[ar]
;B[cs]
;W[ap]
;B[do]
;W[dq]
;B[as]
;W[aq]
;B[ao]
;W[aq]C[What white is doing is locally killing the shape.
]

;B[gn]
;W[hm]
;B[fm]
;W[gm]
;B[fo]C[And then it turns into this very large fight.
]

))(;W[bq]C[KataGo gives black -31 pp for 193. E6
]

)))(;W[co]
(;B[cm]
;W[dm]
;B[dl]
;W[bn]
;B[bm]
;W[am]
;B[bl]TR[cl]C[Just give up C8.
]

(;W[bo]C[You can connect.
]

)(;W[fl]C[Or come back and play here.
]

;B[bo]
;W[bp]
;B[an]C[Because now if black takes, you can just roll with it. You can tenuki now.
]

;W[ap]C[Even play here. Black potentially loses another point here.
]

;B[ao]C[Because they can't play this move.
]

;W[bn]
))(;B[ki]
;W[km]
;B[kn]
;W[jm]
;B[jn]
;W[in]C[KataGo gives white -5.8 pp for 174. C5
]

)))(;W[rs]C[Black is alive still.
]

;B[rr]
(;W[sp]
;B[po]C[As I said earlier, there's miai for this eye.
]

)(;W[po]
;B[ro]TR[po]C[And if white plays Q5, there's miai for this eye, so black is alive.
]

)(;W[ni]
;B[mq]
;W[ir]
;B[ki]
;W[mi]
;B[kg]
;W[kf]
;B[nr]
;W[or]
;B[ms]
;W[js]C[KataGo gives white -2.9 pp for 172. S1
]

)))(;W[ir]C[(KataGo gives white -9.8 pp for this move.)
]

(;B[or]TR[ir]C[If white just takes at J2, first of all, there's this, so now white only has half an eye here.
]

;W[qk]
;B[rj]TR[qk]C[And if we come back and fix at R9, there's still not a full eye.
]

;W[pq]
;B[rk]TR[pq]C[And if we connect at Q3, black has to play here.
]

;W[pj]
;B[ok]
;W[pk]
;B[nk]C[So like this.
]

;W[pm]
;B[om]C[And then if we play here, we can get rid of white's false eye.
]

;W[nl]C[Eventually, black will be able to come back and fix.
]

)(;B[pm]C[I think what will most likely happen is that black can make this exchange first.
]

;W[qk]
;B[or]C[Come back and play here.
]

;W[pq]
;B[rk]C[And then hane.
]

;W[rs]
;B[rr]
;W[po]C[Basically, in order to kill, white needs to remove half an eye and play here.
]

;B[ro]C[And black can make this move.
]

;W[pn]
;B[qm]
;W[om]C[So like this, white can capture.
]

;B[rl]TR[qe]TR[of]TR[pf]TR[qf]TR[rf]TR[qg]TR[qh]TR[oi]TR[pi]TR[lj]TR[nj]TR[ll]TR[ln]TR[mn]TR[mo]SQ[dp]SQ[fp]SQ[hp]SQ[ip]SQ[jp]SQ[kp]TR[mp]C[And then black will be to take all the points on the other side. There's still some aji on the bottom, but black has connected the bottom group and now white needs to make use of the wall that they made on the bottom.
]

)))(;B[ko]TR[jk]C[K9 is not even a real ko threat because black can live with a move like this. (KataGo gives black -29 pp for this move.)
]

))(;W[pb]C[Honestly, probably just continuing the ko threats up here would have been ideal.
]

;B[pa]
;W[aa]
;B[pq]
;W[nj]
;B[lj]
;W[cb]
;B[ag]
;W[co]
;B[or]C[KataGo gives white -6.3 pp for 160. Q18
]

))(;B[ro]C[If black wanted to live, they could play here.
]

;W[cb]C[But white would take.
]

;B[ag]C[And then take. This isn't very good for black.
]

;W[co]
;B[bn]
;W[bo]
;B[ki]
;W[jh]
;B[rk]
;W[nj]
;B[pk]
;W[pl]
;B[ml]C[KataGo gives black -31 pp for 157. S5
]

))(;W[rs]C[Technically, white can play this way.
]

;B[rr]
(;W[sp]
;B[ro]
;W[so]
;B[pm]C[However, black has this sente move.
]

;W[po]
;B[pl]TR[po]C[If white plays Q5, black can play here.
]

;W[pk]
;B[qk]SQ[pl]TR[ql]SQ[pm]TR[rm]SQ[on]SQ[qn]TR[rn]SQ[oo]SQ[qo]SQ[ro]TR[so]SQ[rp]TR[sp]SQ[rq]SQ[qr]SQ[rr]SQ[sr]SQ[qs]C[And then cut the group, so black is not completely dead.
]

)(;W[aa]
;B[gf]
;W[fg]
;B[ab]
;W[pb]
;B[pa]
;W[aa]
;B[pm]
;W[cb]
;B[hg]
;W[hf]
;B[ag]
;W[qk]
;B[hh]
;W[jg]
;B[hj]
;W[gj]
;B[hk]
;W[fl]
;B[bp]C[KataGo gives white -1.5 pp for 156. S19
]

)))(;B[ro]
;W[qh]
;B[rh]
;W[sd]
;B[sb]
;W[sf]
;B[rg]
;W[ri]
;B[qi]
;W[sh]
;B[ph]
;W[rk]
;B[rj]
;W[sj]
;B[si]
;W[pa]
;B[pb]
;W[ri]
;B[pl]
;W[si]
;B[pk]
;W[qk]
;B[pj]
;W[sg]
;B[qh]
;W[oa]
;B[mj]
;W[ml]
;B[mk]
;W[lk]
;B[kl]
;W[pm]C[KataGo gives black -26 pp for 155. S5
]

))(;W[ps]TR[rm]C[This is a very good ko threat that black can't tenuki, so it doesn't make sense that white played S7.
]

(;B[bp]
;W[rs]
;B[rr]
;W[sp]
;B[nm]
;W[om]TR[on]TR[qn]TR[oo]TR[qo]TR[rp]TR[rq]TR[qr]TR[rr]TR[sr]TR[qs]C[The black group is dead.
]

(;B[mm]
;W[mo]
;B[kn]C[Black can get out into the center, but with the bottom right dying, it's not ideal.
]

)(;B[mo]C[And if we try to cut white, it's not possible. White can definitely just get out eventually.
]

;W[mn]
))(;B[qp]
;W[hp]
;B[fr]
;W[er]
;B[mo]C[KataGo gives white -30 pp for 134. Q1
]

)))(;B[mo]C[What happens if black plays this one first?
]

(;W[mn]
;B[ko]
;W[lm]
;B[lp]
(;W[ag]
;B[ab]C[Basically, black will have to take and then the ko starts again.
]

)(;W[jn]TR[hh]TR[ih]TR[jh]TR[kh]TR[lh]TR[hi]TR[li]TR[hj]TR[lj]TR[hk]TR[lk]TR[hl]TR[il]TR[jl]TR[kl]TR[ll]C[But also here, white is actually going to just build the center now. And the ko is still there - black's got to be careful.
]

))(;W[rc]
;B[qb]
;W[mn]
;B[ko]
;W[ag]
;B[ab]
;W[lm]
;B[cb]
;W[kn]
;B[gr]
;W[co]
;B[bn]
;W[jp]
;B[ip]
;W[io]C[KataGo gives black -7.4 pp for 129. N5
]

)))(;W[ip]TR[oj]C[One thing I think white should definitely try to do instead of P10 is to make use of the bottom group.
]

;B[hr]
;W[hp]TR[ag]C[As long as white gets sente and comes back to connect at A13, that's how the game becomes difficult for black because they'll need to win the ko.
]

(;B[oj]SQ[gl]SQ[hl]SQ[il]SQ[jl]SQ[kl]SQ[ll]SQ[ml]TR[ol]TR[ql]SQ[gm]SQ[mm]SQ[gn]SQ[mn]TR[rn]SQ[go]SQ[ho]SQ[io]SQ[jo]SQ[ko]SQ[lo]SQ[mo]C[I played this move just because black is still aiming at cutting the three white stones rather than reducing the marked center area.
]

;W[jp]C[I laid out this position up to here.
]

;B[fr]C[This move was a little difficult for me to find, so I had to use the AI here.
]

;W[er]
;B[mr]
;W[ag]
;B[ab]
;W[rc]
;B[qb]
;W[aa]C[And then white should start this ko again.
]

;B[gf]
;W[hg]
;B[ab]
;W[rb]
;B[ra]
;W[aa]
;B[fg]
;W[fh]C[In this case, you get a very similar situation to the actual game.
]

)(;B[ln]C[When black tries to come out here, white could potentially threaten to kill the whole thing. But also if you try playing this move, now there are a lot more ko threats for white.
]

;W[jp]
;B[mo]C[If black fixes let's say here, what white's trying to do is focus on making this group not alive.
]

(;W[hq]
;B[jr]
;W[jq]
;B[ir]
;W[gr]
;B[mr]
;W[mq]
;B[lp]
;W[nr]C[So like this.
]

;B[ls]C[It looks like the group is alive.
]

(;W[kr]C[But when white throws in, there's really only one eye.
]

;B[ks]
;W[hs]
;B[lr]
;W[is]
)(;W[hs]
;B[kr]C[Actually, it looks like black is alive here. Maybe I missed something? White's corner does become a lot thicker, so the trade-off is there.
]

))(;W[ag]C[So white can come back and connect to start the ko.
]

;B[ab]
;W[hq]C[And then you'll notice that white can start playing all these ko threats.
]

;B[jr]
;W[aa]
;B[gf]
;W[hf]
;B[ab]
;W[jq]
;B[ir]SQ[nm]TR[ln]SQ[nn]TR[mo]TR[mp]TR[gq]TR[iq]TR[kq]TR[lq]TR[hr]TR[jr]C[So the whole black bottom group becomes ko threats. The other thing is that white needs to be careful because the O6 group is not connected.
]

))(;B[fr]
;W[oj]
;B[ln]
;W[jp]
;B[lp]
;W[ag]
;B[ab]
;W[rc]
;B[qb]
;W[aa]
;B[gf]
;W[fg]
;B[ab]
;W[jq]
;B[jr]
;W[aa]
;B[hg]
;W[hf]
;B[ab]
;W[hq]
;B[ir]
;W[rb]
;B[ra]
;W[aa]
;B[hh]
;W[cb]
;B[gh]
;W[fh]
;B[gj]
;W[fj]
;B[gk]
;W[el]
;B[ek]
;W[fk]
;B[fl]
;W[gl]
;B[fm]
;W[gi]
;B[hi]
;W[hj]
;B[hk]C[KataGo gives white -2.2 pp for 126. J4
]

)))(;B[cb]
;W[rc]TR[cb]C[If black just takes, white will play here first.
]

(;B[qb]
;W[rb]C[And then here.
]

;B[ra]
;W[qg]C[And then, I believe, play this move.
]

;B[re]C[Potentially, you lose all your stones.
]

;W[og]
)(;B[rb]
;W[qb]
;B[qa]
;W[pc]
;B[pb]
;W[qg]
;B[rg]
;W[rh]
;B[rf]
;W[qh]C[KataGo gives black -45 pp for 121. C18
]

))(;B[og]C[This leads to more ko threats.
]

(;W[oh]
;B[ph]
;W[qg]C[This doesn't make sense for black to do.
]

)(;W[aa]
;B[gf]
;W[fg]
;B[ab]
;W[oh]
;B[qg]
;W[aa]
;B[bf]
;W[cg]
;B[ai]
;W[oi]
;B[hg]
;W[hf]
;B[hh]
;W[hj]
;B[jf]
;W[ig]
;B[oj]
;W[nj]
;B[mj]
;W[jg]
;B[nk]C[KataGo gives black -4.5 pp for 121. P13
]

)))(;B[da]C[Potentially, you could throw in.
]

;W[ca]
;B[bb]C[And play like this.
]

(;W[dc]C[But white will connect, and it's a bulky five versus the group.
]

)(;W[ed]C[So I think white would try playing here.
]

(;B[dc]C[You can't play the atari.
]

;W[de]C[For obvious reasons.
]

)(;B[bf]C[If you take, I think white just extends.
]

;W[ee]
(;B[dc]C[And now if you cut, you're in a ko, which isn't good.
]

;W[be]
)(;B[ag]C[So you would end up having to take.
]

;W[dc]TR[eb]TR[fb]TR[gb]TR[fc]TR[gc]TR[fd]TR[fe]TR[he]TR[ie]C[And white connects, and then your top group dies.
]

))(;B[ee]C[So you would play the atari.
]

;W[dc]C[Connect.
]

;B[de]C[Connect, and then white can probably just start taking liberties.
]

;W[bf]C[One way is probably just connecting.
]

;B[gf]
;W[hf]
;B[fg]
;W[fh]TR[gf]TR[fg]C[Definitely just take liberties again if black plays like this.
]

;B[fa]C[And eventually when you play this, it just turns into a very large ko.
]

;W[gd]
;B[da]TR[hb]TR[hc]TR[gd]TR[hd]TR[jd]TR[be]TR[je]TR[af]TR[bf]TR[cf]TR[df]TR[hf]TR[if]TR[bg]SQ[fg]TR[gg]SQ[hg]SQ[ig]SQ[jg]SQ[kg]SQ[lg]SQ[mg]SQ[ng]TR[dh]TR[eh]TR[fh]SQ[nh]TR[fi]SQ[ni]SQ[fj]SQ[nj]SQ[fk]SQ[gk]SQ[hk]SQ[ik]SQ[jk]SQ[kk]SQ[lk]SQ[mk]SQ[nk]C[But playing this ko is not worth it, especially if you allow white to make a very solid wall here. That means the whole center area basically becomes white.
]

))(;W[ab]
;B[ba]
;W[ed]
;B[ee]
;W[dc]
;B[bf]
;W[rd]
;B[od]
;W[ne]
;B[og]
;W[be]
;B[de]
;W[ge]
;B[nh]
;W[oh]
;B[oi]
;W[mh]
;B[ni]
;W[bf]
;B[nm]
;W[fa]
;B[ga]
;W[ef]
;B[da]
;W[ea]
;B[fa]
;W[rc]
;B[qb]
;W[ff]
;B[da]
;W[rf]
;B[re]
;W[ea]
;B[ln]C[KataGo gives black -2.3 pp for 111. D19
]

)))(;B[ag]C[It's not good to do this. (KataGo gives black -36 pp for A13.)
]

;W[kq]TR[mp]TR[gq]TR[iq]TR[lq]C[It's kind of devastating once white disconnects the groups.
]

;B[ho]TR[gq]TR[iq]C[These two stones need to run.
]

;W[gn]TR[ho]TR[gq]TR[iq]C[And then white can attack them.
]

))(;B[pr]TR[nm]C[There is no reason not to take this. First of all, your corner group's alive. But also, O7 is a cut. So black should play here in order to create a position where white needs to respond.
]

(;AB[nm]C[If black gets an extra move and is able to cut here, white has to split.
]

;W[om]
;B[nl]TR[qj]SQ[nn]SQ[no]SQ[np]SQ[op]SQ[pp]SQ[nq]SQ[qq]C[But now because of the R10 stone, black is gonna drive this in order to attack white's shape.
]

;W[ok]
;B[nk]
;W[oj]
;B[ln]
)(;W[nm]C[So now, white's forced to play a move like this.
]

(;B[pi]C[And then you if you really need to, you can come back and fix here.
]

)(;B[ae]C[Or in this case, you could potentially fix the aji here.
]

)(;B[ho]SQ[nm]SQ[nn]SQ[no]TR[mp]SQ[np]TR[gq]TR[iq]TR[lq]SQ[nq]C[Or play on this side to make a taller position for the low stones against this white wall.
]

(;W[qh]
;B[rl]
;W[rm]
;B[qm]
;W[rk]
;B[pl]
;W[qk]
;B[pk]
;W[pj]
;B[oj]
(;W[pi]
;B[ok]TR[qh]SQ[qj]XX[nm]XX[nn]XX[no]XX[np]XX[nq]C[Because again, even if white plays something like R12, black still has the ability to cut through here. White can definitely attack and capture R10, but would have to run with the O7 wall.
]

;W[pm]C[And there's no way for white to play this one.
]

)(;W[pm]
;B[pi]
;W[om]SQ[pi]TR[pm]C[Even if white were to cut first, once black takes, first white needs to connect.
]

;B[rj]C[You can play the atari.
]

;W[sl]
;B[sj]C[Drop down.
]

;W[sk]
(;B[sn]C[You can't play this one.
]

;W[ro]C[Because of this.
]

)(;B[og]
;W[oh]
;B[ph]TR[og]C[But you can push and cut.
]

)))(;W[oj]C[So instead, if white wants to attack R10, they actually need to play here first.
]

;B[hm]
;W[qh]
;B[rl]
;W[rm]
;B[qm]
;W[rk]C[This way, white protects the aji.
]

;B[pl]
;W[qk]
;B[pk]
;W[pj]
;B[ok]
;W[nk]
;B[pm]TR[pk]C[And now when you cut at Q9, there is no way to be able to recuperate the loss.
]

(;W[rj]C[So white can play here, make sure there's nothing shady happening.
]

)(;W[rg]C[Or can play here.
]

;B[rj]
;W[sl]
;B[sj]
;W[sk]
;B[sn]
;W[ro]
;B[qp]
;W[so]C[There's no way for black to actually capture. White doesn't even have to play T5 anymore, white can just tenuki.
]

))))(;W[gr]
;B[nm]
;W[om]
;B[mm]
;W[lo]
;B[mk]
;W[nk]
;B[ln]
;W[ko]
;B[kn]
;W[jo]
;B[mj]
;W[nj]
;B[mo]TR[mj]TR[mk]TR[mm]TR[nm]TR[kn]TR[ln]C[This was kind of the AI sequence. You can make this group in the center and potentially cut white later.
]

)(;W[pi]
;B[pj]
;W[oj]
;B[oi]
;W[qi]
;B[ri]
;W[rh]
;B[ok]
;W[nj]
;B[rj]
;W[nk]
;B[qh]
;W[qg]
;B[ph]
;W[og]
;B[rg]C[What really happens is, there's aji here, so black can just live locally. There's nothing really here for white to do anything to attack this group.
]

;W[rf]
;B[sh]
;W[pe]
;B[rd]
;W[qd]
;B[qc]
;W[re]
)(;W[qh]
;B[rl]TR[qh]C[What I'm showing my student is, even if white took this position, you can play here as black and then cut.
]

;W[rm]
;B[qm]
;W[rk]
;B[pl]
;W[qk]
;B[pk]
;W[pj]
;B[oj]
;W[pi]
;B[ok]SQ[ol]TR[rl]SQ[nn]SQ[no]SQ[np]SQ[op]SQ[pp]SQ[nq]SQ[qq]C[So here, you'd be able to just give up the stone in order to attack this whole group.
]

)(;W[bc]
;B[ae]
;W[af]
;B[bf]
;W[pe]
;B[qe]
;W[be]
;B[ie]
;W[ad]
;B[je]
;W[kd]
;B[id]
;W[jc]
;B[gf]
;W[qg]
;B[og]C[KataGo gives black -3.1 pp for 81. Q2
]

)))(;W[pr]C[You could play here, technically.
]

(;B[nn]
;W[rr]
;B[sr]
;W[qs]TR[nn]C[If black tenukis, you would live locally.
]

)(;B[qs]
;W[nn]TR[qs]C[If white answers, you can come back and play here.
]

)(;B[qp]
;W[pq]
;B[pi]
;W[nn]
;B[rl]
;W[gr]
;B[hr]
;W[gp]
;B[hq]
;W[co]
;B[og]
;W[oh]
;B[oi]
;W[ni]
;B[ph]
;W[nj]
;B[jn]
;W[rk]
;B[pl]C[KataGo gives white -0.5 pp for 80. Q2
]

)))(;W[pr]C[There is no reason not to block here.
]

(;B[qs]C[Force black to drop down first.
]

;W[rn]C[And then play this move.
]

;B[qo]
;W[nn]C[This way, you can come back and fix your shape.
]

)(;B[qp]
;W[pq]
;B[pi]
;W[rn]
;B[po]
;W[rr]
;B[sr]
;W[rs]
;B[nn]
;W[mo]
;B[nl]
;W[ok]
;B[pl]
;W[pk]
;B[rl]
;W[gr]C[KataGo gives white -0.9 pp for 78. Q2
]

))(;W[qo]C[Later, you might be able to press and wedge.
]

(;B[rn]
;W[pn]
)(;B[ro]
;W[nn]
;B[pi]
;W[bc]
;B[ae]
;W[af]
;B[bf]
;W[pe]
;B[qe]
;W[be]
;B[ie]
;W[je]
;B[bf]
;W[qd]
;B[qc]
;W[be]
;B[id]C[KataGo gives white -5 pp for 78. R5
]

))(;W[nn]C[But you also remove the bamboo point.
]

(;B[oj]
;W[om]C[So eventually, this move would become sente.
]

)(;B[fr]
;W[bc]
;B[ae]
;W[af]
;B[bf]
;W[pe]
;B[qe]
;W[be]
;B[ad]
;W[ec]
;B[ed]
;W[er]
;B[ha]
;W[cc]
;B[cd]
;W[dc]
;B[da]C[KataGo gives white -0.7 pp for 78. O6
]

)))(;B[ln]TR[on]TR[qn]TR[pp]TR[rp]TR[lq]TR[nq]TR[qq]TR[rq]C[I want to mention that this technique down here is something you see in the old chinese fuseki style. Usually at this point, you see a keima on either side. (See variation at move 1.)
]

)(;B[mm]C[But in this case, making a keima in the center doesn't make a whole lot of sense.
]

))(;W[pf]TR[lb]TR[pe]C[If white plays here, black is actually going to force at Q15 and M18 first.
]

;B[pe]
(;W[oe]
;B[lb]
(;W[la]
;B[ne]C[And play the atari.
]

;W[od]
;B[md]C[And then of course this works.
]

)(;W[ma]C[If white descends, black would still play the atari.
]

;B[ne]
;W[me]
;B[od]C[Force this.
]

;W[mf]
;B[og]
;W[oh]
;B[ph]TR[pf]TR[pg]C[And then black can take the two stones.
]

))(;W[qg]
;B[ne]
;W[rf]
;B[lb]
;W[ol]
;B[rl]
;W[rm]
;B[qm]
;W[rk]
;B[pl]
;W[qk]
;B[pk]
;W[pj]
;B[rj]
;W[sl]
;B[sj]
;W[sk]
;B[rn]
;W[ok]
;B[pm]
;W[qi]
;B[lq]
;W[qq]
;B[mo]
;W[rq]
;B[gp]C[KataGo gives white -7.3 pp for 64. Q14
]

)))(;B[qq]
(;W[jp]
;B[ae]SQ[jp]TR[qq]C[I think more humans would play at R3, allow white to take the bottom, and then come back and fix the aji. Playing like this would still be an even game, and it's mostly black playing safe. Basically black is saying, "I'm gonna win the game on endgame."
]

)(;W[kp]
;B[of]
;W[oe]
;B[pf]
;W[nf]
;B[pi]
;W[pq]
;B[ol]
;W[co]
;B[dn]
;W[bn]
;B[bm]
;W[bc]
;B[ae]
;W[cg]
;B[ai]
;W[cc]
;B[bf]
;W[eb]
;B[ag]
;W[go]
;B[ni]
;W[mm]
;B[gq]
;W[fp]
;B[iq]
;W[kq]
;B[io]
;W[im]C[KataGo gives black -1.5 pp for 61. R3
]

)))(;W[qq]
;B[lb]TR[qq]C[If we just normally take the corner here as white, what black is wanting to do here is force this technique.
]

;W[la]
;B[ne]
;W[oe]
;B[nf]
(;W[of]
;B[od]C[If white tries to come out, black will play here and the squeeze no longer works for white.
]

;W[me]
;B[mf]
;W[lf]
;B[md]
;W[mg]
;B[ma]C[So this is what black is hoping to do.
]

)(;W[od]C[But what white will do instead is actually connect.
]

;B[of]
;W[pe]C[This turns into this really difficult position where black needs to come back and fix the shape here.
]

(;B[qc]
;W[pf]
;B[og]
;W[pg]
;B[oh]
;W[rd]
;B[qd]
;W[qe]C[But white will be able to push out and play moves like this.
]

;B[rc]TR[qa]TR[ra]TR[sa]SQ[lb]TR[qb]TR[rb]TR[sb]SQ[lc]TR[pc]TR[qc]TR[sc]TR[pd]TR[qd]TR[rd]TR[sd]SQ[le]SQ[ne]SQ[nf]SQ[of]TR[qf]TR[rf]TR[sf]SQ[og]TR[qg]TR[rg]TR[sg]SQ[oh]TR[qh]TR[rh]TR[sh]C[So black can live, but it comes at the cost of losing corner points here and side territory. Black gets to save the outside stones, and they're not exactly strong.
]

(;W[ph]C[So eventually white will be able to continue pushing.
]

;B[oi]
;W[pi]
;B[pj]
;W[oj]
;B[nj]
;W[ok]C[This push and cut could become a serious threat.
]

)(;W[ol]SQ[lb]SQ[lc]SQ[le]SQ[ne]SQ[nf]SQ[of]SQ[og]SQ[oh]TR[oj]SQ[on]SQ[qn]C[Next, white would jump and you'll notice that the push and cut at P10 works, so now black is separated on two sides.
]

))(;B[og]C[But there's definitely still aji that comes from this.
]

;W[qd]
;B[kg]SQ[qd]XX[lg]XX[mg]XX[ng]TR[og]XX[pg]XX[qg]XX[rg]XX[sg]XX[kh]XX[sh]XX[ki]XX[si]XX[kj]XX[sj]XX[kk]XX[sk]XX[kl]XX[sl]XX[km]XX[sm]XX[kn]XX[ln]XX[mn]XX[nn]XX[on]XX[pn]XX[qn]XX[rn]XX[sn]C[For example, if black just takes the outside and allows white to eat the corner, there is potential for black to go for a big box on the outside. This is kind of a big gamble, but it's probably something black doesn't wanna do just because it's a big risk.
]

)(;B[qe]
;W[kc]
;B[pf]
;W[ld]
;B[gq]
;W[gp]
;B[hq]
;W[fr]
;B[lq]
;W[pj]
;B[rq]
;W[no]
;B[hp]
;W[bc]
;B[ad]
;W[qi]
;B[qk]
;W[pk]
;B[rl]
;W[cc]
;B[cd]
;W[ec]
;B[eb]
;W[db]
;B[fc]
;W[ea]
;B[fb]
;W[ed]
;B[ee]C[KataGo gives white -0.5 pp for 59. R3
]

))))(;W[ec]TR[fe]C[F15 basically just protects against this cut.
]

)(;W[cc]C[Earlier, we could do this type of thing.
]

(;B[cd]
;W[bc]
;B[ae]
;W[af]
;B[bf]
;W[cg]
;B[ai]
;W[be]
;B[ad]
;W[db]
;B[eb]
;W[ec]C[We still have this issue.
]

;B[fb]
;W[ed]
;B[ee]
;W[dc]
;B[de]TR[fb]SQ[gd]C[I guess even if you connect at F18, white's still looking to kind of just force the cut at G16 and potentially reduce the liberties on the outside.
]

;W[gd]
;B[fc]
;W[ge]
;B[bb]
;W[ab]
(;B[ac]
;W[ba]C[On the inside, black might have to play the ko.
]

)(;B[ba]
;W[ac]
;B[da]C[Or get stuck with something like this, but it looks like it should be fine for black.
]

))(;B[bc]
;W[pg]
;B[ie]
;W[cd]
;B[de]
;W[je]
;B[hf]
;W[hg]
;B[fg]
;W[ff]
;B[ef]
;W[fh]
;B[gf]
;W[jg]
;B[nb]
;W[mb]
;B[ob]
;W[mc]
;B[lb]
;W[la]
;B[nf]
;W[mf]
;B[me]
;W[ne]
;B[of]
;W[mg]
;B[og]
;W[oh]
;B[oi]
;W[ph]
;B[nh]
;W[pi]
;B[mh]
;W[lg]
;B[pj]
;W[ni]
;B[oj]
;W[mi]
;B[ka]C[KataGo gives white -1.2 pp for 54. C17
]

)))(;B[lg]C[So now when you play like this, white can still make exchanges and then just jump out to play lightly. I think just continuing this attack would have been better for black.
]

(;W[pg]
;B[qg]
(;W[pi]
)(;W[jc]
;B[jg]
;W[jf]
;B[gc]
;W[nb]
;B[li]
;W[nj]
;B[qp]
;W[qq]
;B[rq]
;W[po]
;B[qo]
;W[pn]
;B[mj]
;W[qm]
;B[qr]
;W[pq]
;B[ro]
;W[ni]
;B[km]
;W[co]C[KataGo gives black -4.1 pp for 49. M13
]

))(;W[ni]
;B[li]
;W[cc]TR[li]TR[ni]C[But also potentially when white jumps and black jumps, white will start triggering this aji right now.
]

;B[cd]
;W[bc]
;B[ae]
;W[af]
;B[bf]
;W[cg]
;B[ai]
;W[be]C[White wants to do is kind of make this aji happen.
]

;B[ad]
;W[db]
;B[eb]
;W[ec]
(;B[fb]C[You can no longer try playing this way.
]

;W[dc]
;B[ed]
;W[fc]
;B[gc]
;W[gd]
;B[hd]
;W[ge]
;B[gf]
;W[fe]
(;B[ef]
;W[ff]TR[gd]C[The cut actually works for white now.
]

)(;B[ff]
;W[ee]
;B[ef]
;W[de]
;B[bf]
;W[ag]TR[ad]TR[bd]TR[cd]TR[dd]TR[ed]TR[fd]TR[ae]TR[ce]TR[bf]C[If you play all the way, white can connect underneath and your stones are dead.
]

))(;B[fc]C[So we have to actually block here now.
]

;W[ea]C[Now there's aji, so play the atari.
]

;B[fb]
;W[ac]C[And then potentially make eyespace with something like this.
]

;B[ba]C[I believe the aji is like this.
]

;W[bb]
;B[ca]C[And then we can throw in.
]

;W[aa]
;B[ab]
;W[da]
(;B[aa]TR[da]C[So here if white connects and for example you connect, what will happen is white can throw in and live with a squeeze here.
]

;W[cb]
;B[ba]
;W[aa]
;B[ab]
;W[ca]
;B[fa]
;W[aa]
)(;B[dc]C[So we have to be careful on how we approach this. We actually have to play the ko here.
]

;W[aa]C[This would be the ko.
]

)))))(;W[nf]C[I've thought about this and made a few variations of my own. The biggest idea is that if we do a running jump, black is going to make a wall here.
]

(;B[lg]TR[hc]TR[nc]TR[jd]TR[nd]TR[if]TR[nf]C[To attack both of white's groups. It's also a timing thing.
]

;W[nh]
(;B[li]
;W[nj]
;B[nb]
;W[mb]
;B[ob]C[If we continue jumping, eventually black will play here.
]

(;W[mc]
;B[lb]TR[mc]C[If you connect, black gets this forcing move and then gets to come back and cut.
]

;W[ma]
(;B[ie]
)(;B[jc]
;W[id]
;B[hb]C[Or potentially just connect underneath outright.
]

))(;W[lb]TR[nb]C[Depending on when black hanes, white can actually play here.
]

;B[mc]
;W[kc]
;B[me]
;W[ne]
;B[mk]TR[kb]C[But the L18 cut can be kind of severe when black is going for an attack.
]

;W[mj]
;B[lj]
;W[lk]
;B[mg]
;W[ng]C[So if we're not careful, for example say something like this...
]

;B[kk]
;W[ll]
;B[kl]
;W[lm]
;B[ie]
;W[je]
;B[jf]SQ[lb]SQ[mb]SQ[hc]SQ[kc]SQ[jd]SQ[be]SQ[je]SQ[cf]SQ[df]SQ[if]SQ[bg]SQ[gg]SQ[dh]SQ[eh]SQ[fi]TR[mk]C[Potentially black gives up N9 and then pushes and cuts. The white top and side groups are under attack. They're disconnected and they don't have two eyes.
]

;W[cg]
;B[ai]
(;W[af]C[Potentially, this group can get two eyes through a ko.
]

;B[ad]
;W[ae]
;B[ag]
;W[ah]C[So that's one and a half eyes.
]

)(;W[ff]
;B[eg]TR[ff]C[If you play at F14 first, it's still one and a half eyes, so we need to be careful.
]

)(;W[jb]TR[lb]TR[mb]TR[hc]TR[kc]TR[jd]SQ[be]TR[je]SQ[cf]SQ[df]SQ[if]SQ[bg]SQ[cg]SQ[gg]SQ[dh]SQ[eh]SQ[fi]C[But the L17 group needs to live first. So probably what will happen is black is going to attack either one of these groups.
]

)))(;B[nb]C[Even here, black could still play this way.
]

;W[mb]
;B[ob]
(;W[mc]
;B[lb]
;W[ma]
;B[li]SQ[hc]TR[mc]SQ[jd]SQ[be]SQ[cf]SQ[df]SQ[if]SQ[bg]SQ[gg]SQ[dh]SQ[eh]SQ[fi]SQ[ej]C[And if you connect, black is gonna be able to jump out and attack directly.
]

)(;W[lb]
;B[mc]
;W[kc]
;B[me]
;W[ne]TR[mb]C[If you try to go underneath, it's the same issue.
]

;B[mg]
;W[ng]
;B[li]C[Black will most likely play stuff like this and force you to push out.
]

;W[mj]
;B[mi]
;W[ni]
;B[lj]
;W[mk]
;B[lk]
;W[ml]
(;B[oj]C[Maybe make an exchange first, maybe not.
]

)(;B[ie]
;W[je]
;B[jf]C[And then push and cut. So it's the same issue, different stone. 
]

;W[qg]SQ[lb]SQ[mb]SQ[hc]SQ[kc]TR[nc]SQ[jd]TR[nd]SQ[be]SQ[je]TR[ne]SQ[cf]SQ[df]SQ[if]TR[nf]SQ[bg]SQ[gg]TR[ng]SQ[dh]SQ[eh]TR[nh]SQ[fi]TR[ni]SQ[ej]TR[mj]SQ[qj]TR[mk]TR[ml]SQ[qn]C[So white gets this gigantic wall, and potentially later you can separate the right side stones, but the greater risk is one of the top left groups dying.
]

))))(;B[nb]
;W[cc]
;B[cd]
;W[mb]
;B[ob]
;W[eb]
;B[gc]
;W[bc]
;B[ae]
;W[dc]
;B[hd]
;W[ke]
;B[kd]
;W[lf]
;B[je]
;W[kf]
;B[jc]
;W[qo]
;B[pn]
;W[ol]
;B[nn]
;W[lp]
;B[rn]
;W[pi]
;B[pj]
;W[oj]
;B[oi]
;W[nj]
;B[ph]
;W[ni]C[KataGo gives white -1.8 pp for 48. O14
]

)))(;W[ol]C[I do have a sequence here. I think white should jump here, that's my personal opinion. Usually, when you have a pincered stone on the side, we don't start another fight.
]

;B[qp]
;W[qq]
;B[rq]
(;W[pq]
;B[ro]
;W[po]
;B[qm]
;W[pl]
;B[jp]TR[qm]TR[qn]TR[ro]TR[qp]TR[rq]C[I figured something like this would happen, and then black could play here. It's actually kind of sente for white to attack the corner group.
]

;W[qo]
;B[rp]
;W[rr]
;B[rl]
;W[rk]
;B[sl]
;W[qk]
;B[gp]XX[nc]SQ[qk]SQ[rk]SQ[ol]SQ[pl]SQ[ql]TR[rl]TR[sl]TR[qm]TR[qn]TR[ro]TR[qp]TR[rp]TR[rq]C[In attacking the corner group, white was able to get this exchange, which I think would actually be very useful for attacking the top side.
]

)(;W[rr]
;B[ro]
;W[sq]
;B[po]
;W[rp]
;B[op]
;W[pq]
;B[rl]
;W[nc]
;B[oc]
;W[ej]
;B[dj]
;W[bg]
;B[aj]
;W[nd]
;B[le]
;W[nf]
;B[lg]
;W[nh]
;B[li]
;W[nb]
;B[me]C[KataGo gives white -0.5 pp for 40. P8
]

))(;W[ej]
;B[dj]
(;W[bg]C[Honestly, I think you can make these exchanges at any time.
]

)(;W[pj]
;B[pi]
;W[bg]
;B[aj]
;W[oj]
;B[qk]
;W[pl]
;B[le]
;W[kf]
;B[gq]
;W[fp]
;B[jp]
;W[rl]
;B[qp]
;W[qq]
;B[rq]
;W[rp]
;B[qo]
;W[rr]
;B[pq]
;W[qr]
;B[op]
;W[np]
;B[on]
;W[mn]
;B[nm]
;W[qi]C[KataGo gives white -0.1 pp for 40. E10
]

)))(;B[bf]
;W[bg]
;B[ae]
(;W[cg]TR[bh]TR[ch]C[If black really wants to get rid of the aji, they can sacrifice two stones here.
]

)(;W[pm]TR[cg]C[But potentially white doesn't answer at C13 and can come back and play here.
]

;B[cg]TR[bd]TR[ae]TR[bf]TR[ql]TR[pm]TR[pp]C[You can take this, but white has spent moves in the lower right while you have spent a couple of moves on the first and second line in order to get rid of aji while white is growing the center and the corner.
]

)))(;B[mc]
;W[if]
;B[qj]
(;W[qc]
(;B[pc]TR[ol]C[One reason I think jumping at P8 is important for white is that we can live locally in the corner.
]

(;W[qd]
;B[qe]
;W[pb]
;B[ob]
;W[pa]
(;B[ra]C[The way to attack would be this move.
]

;W[nb]
;B[nc]
;W[oa]
;B[mb]
;W[oc]C[White can take.
]

;B[od]
;W[rb]C[And then here, we're just alive now.
]

)(;B[oc]C[What black just should do though is potentially just fix the weakness.
]

)(;B[nc]C[Either like that or like this.
]

;W[rb]C[Then white needs to come back and live.
]

;B[pl]C[This way, black can play here.
]

))(;W[pb]C[But also, sometimes white can just make a ko.
]

;B[ob]
;W[rb]
;B[pa]
;W[qa]C[Making this ko would be big.
]

))(;B[ob]C[If what white's really worried about is black attacking with a move like this (which I'm gonna be honest, I don't know full well), potentially there might be some stuff here for white.
]

;W[qd]
(;B[qe]
;W[pb]
;B[pc]
(;W[pa]C[For example, this can kind of revert back.
]

;B[nc]XX[pa]CR[ob]XX[pb]TR[hc]CR[mc]CR[pc]XX[qc]TR[jd]CR[pd]XX[qd]SQ[ie]SQ[je]CR[qe]TR[if]CR[qf]TR[gg]C[White probably doesn't want to get into a situation where, if we do get to live in the corner, we're giving our opponent this large thickness on the outside. You'll notice the outside white stones are all knight's moves, so what happens there influences what will happen around O17. And vice versa, if black's able to cut white's shape on the outside, potentially the white corner group can die.
]

)(;W[rb]
;B[pa]
;W[qa]C[Or it could revert to a ko.
]

))(;B[pc]
;W[qe]
;B[pf]
;W[rf]
;B[rg]TR[pc]C[If black descends, white could just come back and make shape here.
]

;W[se]
;B[qb]
;W[rb]
(;B[ra]TR[se]C[If you play the tiger's mouth, you're going to die.
]

;W[rd]
;B[sb]
;W[sc]C[Or at least, this would become a ko.
]

)(;B[sc]C[So here, black could definitely play probably something like this.
]

;W[rd]
;B[sg]C[Threatening to poke out white's eye.
]

;W[sb]
;B[sf]TR[sb]C[If white defends at T18, black would come back here.
]

;W[ra]
;B[re]
;W[sd]C[You can make an eye like this.
]

))))(;W[pj]
;B[pi]
;W[oj]
;B[qk]
;W[pl]
;B[pk]
;W[ok]
;B[ol]
;W[pn]
;B[nl]
;W[pm]
;B[qp]
;W[qq]
;B[po]
;W[oo]
;B[qo]
;W[nn]
;B[pq]
;W[op]C[KataGo gives black -0.8 pp for 37. N17
]

)))(;W[nc]
;B[oc]
(;W[nd]
;B[jc]C[This is a similar sequence where black gets the outside and white is kind of pincered here, but this is totally playable, I think.
]

)(;W[jd]TR[nc]C[We'd surrender O17.
]

;B[lc]
;W[nd]TR[lc]C[Now if black pincers, potentially we can come out.
]

;B[le]C[This becomes another fight.
]

;W[jf]C[White is preserving the aji in the corner.
]

;B[lg]SQ[hc]TR[nc]SQ[jd]TR[nd]SQ[be]SQ[cf]SQ[df]SQ[jf]SQ[gg]XX[ng]XX[pg]C[While black is surrounding the two stones; but you'll notice that white is actually surrounding the corner. White can still jump out or play the shoulder hit in order to come out here.
]

)(;W[ob]
;B[pb]
;W[od]
;B[pc]
;W[kc]
;B[qp]
;W[qq]
;B[rq]
;W[rr]
;B[ro]
;W[sq]
;B[po]
;W[rp]
;B[op]C[KataGo gives white -1.3 pp for 36. O17
]

)))(;W[nc]C[It allows white to play something like this.
]

(;B[he]
;W[pc]
;B[qc]
;W[qb]
;B[oc]
;W[pb]
;B[od]
;W[ob]
(;B[nd]TR[he]C[Even if black comes back and fixes at H15, I'd be happy as white to just give black the outside.
]

)(;B[rc]
;W[lc]TR[rc]C[And if black plays S17, we can play here.
]

;B[jc]TR[gb]SQ[hc]TR[fd]TR[he]C[Like this, black spends four moves to surround H17.
]

;W[ql]TR[gb]TR[jc]TR[fd]TR[he]C[Which means now, as white I can come back and play some larger moves. However, black getting those four stones means that the aji is different in the corner.
]

;B[on]
;W[cc]
;B[cd]C[Let's say for example like this.
]

;W[bc]
(;B[bf]
(;W[eb]C[We'd like to play this move, but at this point, it probably doesn't matter.
]

)(;W[bg]
;B[ae]
;W[cg]TR[ae]C[Of course, if black plays A15, eventually white will get this.
]

))(;B[ad]C[So it's possible that black just descends.
]

;W[db]C[And we can play here.
]

;B[eb]
;W[ec]
(;B[fb]C[Like this, black can connect.
]

;W[dc]C[If we connect, black connects.
]

;B[ed]
;W[ea]C[We don't have the atari anymore, which will come into play in the game.
]

)(;B[fc]
;W[ea]
;B[fb]
;W[dc]C[Normally, this is kind of what would happen.
]

;B[ed]
(;W[bb]TR[ed]C[When black connects, we can kind of force a ko shape on top.
]

)(;W[ac]TR[ed]C[Or take a liberty.
]

;B[da]
;W[ba]
(;B[ab]
;W[ca]
;B[bb]TR[hc]C[We can kind of force black to give us a ko later. But because H17 was surrounded, black's outside is very thick.
]

)(;B[ca]
;W[cb]
;B[ab]
;W[da]C[Like this, we can obviously live because it's miai.
]

))))))(;B[qp]
;W[qq]
;B[rq]
;W[pr]
;B[ro]
;W[po]
;B[pl]
;W[pc]
;B[qc]
;W[qb]
;B[oc]
;W[pb]
;B[od]
;W[qd]
;B[ob]
;W[rc]
;B[kd]
;W[he]
;B[gq]
;W[fp]
;B[jq]C[KataGo gives white -3 pp for 34. O17
]

)))(;B[hd]C[I think white is more attackable in this position.
]

;W[id]
;B[he]C[So I think I would have much preferred black just attaching and really coming out.
]

(;W[gb]TR[be]TR[cf]TR[df]TR[gg]TR[dh]TR[eh]TR[fi]C[The other thing is that black wants to be able to attack the C14 stones, just because they don't really have a base.
]

(;B[eb]TR[hd]C[Black can directly protect like this.
]

;W[ec]C[This way, even if white plays here, there is some aji but we can definitely attack this stone.
]

(;B[fc]
;W[cc]
(;B[cd]
(;W[dc]
(;B[fb]
;W[ed]
;B[ee]
;W[de]
(;B[bc]
;W[fe]
;B[cb]
;W[ef]
;B[db]C[We can squeeze later.
]

)(;B[ef]TR[be]C[Or, if we have the liberties as black, for example if we've taken B15 already, we can definitely extend.
]

))(;B[ed]
;W[fb]C[White's aji also potentially works later if this becomes a forcing move.
]

))(;W[ed]
;B[dc]TR[ed]C[White's aji really only works if white can handle something like this with E16 as a forcing move.
]

))(;B[dc]C[Of course, there's always just this, so we can still capture.
]

;W[cd]
;B[de]
;W[bc]
;B[bf]
;W[ad]
;B[bg]C[If white plays like this, it's not game over for black. Potentially we could force a ko or attack the outside.
]

))(;B[dc]
;W[fc]C[So the aji for white is there, but it's not complete. This whole sequence would allow black to just preserve the corner.
]

;B[bf]
;W[bg]
;B[ae]
;W[cg]TR[be]C[If black ever takes the stone at B15, white can obviously take here, so that's not ideal unless it's a matter of endgame or life-and-death.
]

))(;B[ie]
;W[je]
;B[ig]
;W[bg]
;B[ai]
;W[cg]C[The only way for white's group to get a base is to play this.
]

;B[aj]
(;W[af]C[Like this, white's group can get an eye.
]

(;B[ff]C[But I don't know if white can get two eyes here because there's still a ko.
]

)(;B[ad]
;W[ae]
;B[ag]C[For example, this a ko that's winnable for black.
]

))(;W[ff]C[If white comes back and fixes, there's still a ko there.
]

;B[eg]C[Black can also definitely just take away the full eye.
]

)))(;W[fc]
;B[fh]
;W[gh]
;B[ic]
;W[gd]
;B[ge]
;W[gc]
;B[ee]
;W[ff]
;B[fe]
;W[jc]
;B[eg]
;W[dg]
;B[fj]
;W[gi]
;B[ib]
;W[jb]
;B[ef]
;W[fg]
;B[ec]
;W[eb]
;B[db]
;W[fa]
;B[bf]
;W[ej]
;B[dj]
;W[bg]C[KataGo gives black -1.3 pp for 33. H16
]

)))(;W[bg]C[It's an issue for black that this move is sente.
]

;B[ai]C[So we'll need to come back and play here.
]

)(;W[ej]C[But first, white would probably play the atari.
]

;B[dj]
(;W[bg]C[And then play this.
]

;B[aj]C[You'd have to play the atari.
]

(;W[bc]C[Now white can come back and attach.
]

;B[ae]
;W[af]C[So this would become a ko.
]

)(;W[cc]C[White can also play here.
]

;B[cd]
;W[eb]C[And then lean.
]

;B[fb]
;W[ec]
;B[fc]
;W[bc]C[So we're looking to play stuff like this.
]

;B[ae]
;W[af]
;B[ad]
;W[ed]
;B[ee]
(;W[de]C[And then potentially cut.
]

)(;W[dc]TR[de]SQ[fe]C[Or if you can't cut at D15, you can connect and you'd have the cut at F15.
]

)))(;W[gg]
;B[hd]
;W[gd]
;B[gc]
;W[ge]
;B[fc]
;W[hb]
;B[fh]
;W[gh]
;B[fe]
;W[gf]
;B[fj]C[KataGo gives white -7.1 pp for 32. E10
]

)))(;W[ej]SQ[dd]SQ[fd]SQ[ce]TR[bi]TR[ci]TR[bj]C[You can play the atari later, totally fine. Essentially, white is giving up the three stones to basically attack the D16 group.
]

;B[dj]
))(;W[bh]
;B[fh]SQ[cf]SQ[df]TR[bh]SQ[dh]SQ[eh]SQ[bi]SQ[ci]SQ[bj]C[If white ataris, black can hane and threaten the whole group.
]

;W[kc]
;B[fg]
;W[ql]
;B[qp]
;W[qq]
;B[rq]
;W[rr]
;B[ro]
;W[sq]
;B[po]
;W[rp]
;B[op]
;W[pq]
;B[qj]
;W[pk]
;B[pj]C[KataGo gives white -3 pp for 28. B12
]

))(;W[fh]C[I thought about white just extending.
]

;B[ck]C[Then black can play two-space.
]

;W[eb]SQ[dd]SQ[fd]SQ[ce]TR[di]TR[ei]TR[ck]C[Now white can change the mode of attack. Instead of making the D11 shape weak, we're actually making the D16 shape weak, which would change the game drastically. I think it might just be easier to play this way.
]

(;B[hd]
;W[id]
;B[he]
;W[gc]
;B[ie]
;W[jd]
;B[ig]C[The issue with playing like this is, let's say black just escapes and eventually comes out.
]

;W[ci]C[If you come back now, black has the ladder.
]

;B[bi]
;W[cj]
(;B[dj]
;W[bj]
;B[bk]C[So there's this. Chances are that this ladder to the side of the board is definitely never gonna be bad for black.
]

)(;B[bj]C[And of course, there is always this, too.
]

;W[dj]
))(;B[qp]
;W[qq]
;B[rq]
;W[rr]
;B[rp]
;W[pq]
;B[pk]
;W[cc]
;B[be]
;W[fc]
;B[bg]
;W[nc]
;B[hd]
;W[pc]
;B[qc]
;W[qb]
;B[oc]
;W[pb]
;B[nd]
;W[qd]C[KataGo gives white -2.3 pp for 20. F12
]

)))(;B[ci]C[Ideally, I wouldn't like H17 as white, just because we can come back as black and fix the shape like this.
]

;W[ei]
(;B[ej]
;W[fj]
;B[ek]
;W[fk]
;B[el]C[Possibly, it reverts back.
]

)(;B[dk]
;W[fb]
;B[db]
;W[ge]
;B[kc]
;W[je]
;B[ld]
;W[ql]C[KataGo gives black -5.7 pp for 19. C11
]

))(;B[cj]C[Personally, I think here would be a lot better than the game move.
]

;W[ei]
(;B[ej]
;W[fj]
;B[ek]
;W[fk]
;B[el]
(;W[fl]
;B[hd]TR[fl]C[Because now, when white comes back and plays at F8, we can take sente and play here.
]

)(;W[he]
;B[eb]TR[he]C[If white jumps, we could come back and fix.
]

)(;W[cc]
;B[be]TR[cc]C[If white plays like this, we have the connection underneath, so I think we're perfectly fine in this department.
]

;W[ec]
;B[ed]
;W[fc]
(;B[dc]
;W[db]
;B[cb]C[We'd play like this.
]

(;W[bb]
;B[eb]
;W[ca]
;B[gc]
;W[fb]
;B[gb]
;W[ea]
;B[hd]C[So we would get the outside if white wants to take the corner.
]

)(;W[eb]
;B[bb]TR[eb]C[If white plays at E18, we can definitely take this.
]

))(;B[gc]C[If you don't think you can give away the corner just yet, you can play like this.
]

;W[gb]
;B[gd]
;W[dc]
(;B[bf]C[You can still attack because you can connect underneath.
]

)(;B[ie]TR[fi]C[But potentially you can also cut at F11 later, so being able to play a ladderbreaker is essentially good.
]

)(;B[fi]
;W[gi]
;B[fh]
;W[gg]
;B[fg]
;W[ff]TR[fi]C[One thing you do need to be worried about if you cut directly is that white may try to do this.
]

;B[dg]C[But it seems like it's fine for black.
]

))))(;B[qp]
;W[qq]
;B[rq]
;W[pr]
;B[ro]
;W[eb]
;B[ec]
;W[fb]
;B[db]
;W[po]
;B[bf]
;W[dk]
;B[cg]
;W[kc]
;B[pn]C[KataGo gives black -4.1 pp for 19. C10
]

)))(;W[ci]C[In my opinion, this is a sharper move than the game move, just to show for reference.
]

;B[cj]
;W[bi]
(;B[dj]
(;W[bj]
;B[bk]
;W[bh]TR[ck]C[When it's like this, white would want to play here. It leaves a weakness at C9, which is valuable.
]

;B[bp]
;W[cl]
;B[ck]
;W[bl]C[Because when we play here and force black to either descend or hane, we can force their shape where we would actually get the outside.
]

(;B[ak]
;W[dl]
)(;B[dl]
(;W[dm]
;B[cm]
;W[el]
(;B[al]
;W[dk]
;B[bm]
;W[fj]
)(;B[dk]
;W[ak]TR[dk]C[Or if black connects, we can connect underneath.
]

;B[bm]
;W[al]
;B[aj]
;W[ai]
;B[am]
;W[aj]
;B[bg]
;W[cg]TR[bh]TR[ai]TR[bi]TR[ci]SQ[di]TR[aj]TR[bj]SQ[cj]SQ[dj]TR[ak]SQ[bk]SQ[ck]SQ[dk]TR[al]TR[bl]TR[cl]SQ[dl]C[With all the throw-ins, black still can't capture the white stones, so the outside stones would be disconnected.
]

))(;W[ak]C[Even if as white you don't think you can play the hane and cut, this is still a valuable experience.
]

;B[dm]
;W[ek]
;B[dk]TR[bl]TR[cl]C[Just because we're taking away black's points here, so there's value in white just playing one-directionally.
]

)))(;W[bh]
;B[bf]
;W[cl]
;B[bg]
;W[cg]
;B[bj]
;W[cm]
;B[dn]
;W[ek]
;B[bo]
;W[bp]
;B[bm]
;W[bl]
;B[fi]
;W[ei]
;B[ej]
;W[fj]
;B[fk]
;W[gj]
;B[dk]
;W[el]
;B[dl]
;W[dm]
;B[em]
;W[fl]
;B[ak]
;W[gl]
;B[fn]
;W[am]
;B[bn]C[KataGo gives white -1.1 pp for 18. C11
]

))(;B[dk]C[If black plays this now, black gets a B2 bomber shape.
]

;W[bj]
;B[bk]
;W[cl]
;B[ck]
;W[ei]
;B[dj]
;W[bh]SQ[hc]TR[ci]XX[di]XX[cj]XX[dj]XX[bk]XX[ck]XX[dk]C[One of the reasons that I think white should have cut at C11 immediately instead of playing at H16 is that we want to give black some weak shape.
]

;B[bp]C[So I imagine black could come back and fix.
]

;W[hc]C[And then if white wanted to, they could play here.
]

)))(;B[cj]TR[cf]TR[df]TR[dh]TR[eh]SQ[cm]C[My first reaction would be that we'd need to respond with a move like this tiger's mouth to make sure that the white group remains unsettled. It doesn't matter if white is able to continue at C7. We want to have our C10 group attacking the C14 stones.
]

;W[ei]C[I have this sequence that I found with AI.
]

(;B[ej]
;W[fj]
;B[ek]
;W[fk]
;B[el]C[I believe up until here, any player could usually find this sequence.
]

;W[fl]
(;B[ic]
(;W[cn]TR[ic]C[If black protects the top side, white would need to come back and play in the corner.
]

;B[dn]
;W[do]
;B[en]TR[cf]TR[df]TR[dh]TR[eh]TR[ei]TR[fj]TR[fk]TR[fl]SQ[cn]SQ[do]SQ[dp]SQ[pp]SQ[fq]SQ[nq]C[Black is still trying to keep the C14 group separated and also reduce the potential bottom after white went for the corner territory.
]

)(;W[em]C[If white plays on top, it's actually not that big of an issue.
]

(;B[dm]
;W[dn]
;B[en]
;W[fm]
;B[cn]
;W[do]
;B[cm]TR[cn]C[Black can connect thanks to the atari.
]

;W[bo]C[Now white really needs to focus on protecting, giving black sente.
]

;B[mp]C[Now black can come back and play down here.
]

)(;B[cm]TR[em]C[White's hane is also not sente, so we can play here.
]

(;W[ck]C[Normally, the technique is to attack black's shape with this move.
]

;B[dm]C[But if we play like this, there is no such shape anymore.
]

;W[dl]
;B[cl]
;W[dk]
;B[dj]
;W[bk]
;B[bj]
)(;W[dl]C[If we play the hane, it's a very similar shape. We don't have the ability to attack the stones.
]

;B[dm]
;W[ck]
;B[cl]
;W[dk]
;B[dj]
;W[bk]
;B[bj]
)(;W[dm]C[So then we play here.
]

;B[cl]
;W[cn]
;B[cq]SQ[ch]SQ[di]SQ[cj]SQ[ej]SQ[ek]SQ[cl]SQ[el]SQ[cm]TR[cn]C[After white hanes, the 3-3 is still open because now, the black group is most definitely alive.
]

))))(;B[em]C[There are kind of diminishing returns if we extend now.
]

;W[hc]C[White's more likely to play here.
]

))(;B[dk]
;W[hc]
;B[hd]
;W[ic]
;B[kc]
;W[eb]
;B[ec]
;W[fb]
;B[db]
;W[cb]
;B[bf]
;W[bg]
;B[be]C[KataGo gives black -2.3 pp for 17. C10
]

))(;B[ci]C[I would also expect something similar like this from humans, but if you just connect, it's kind of less threatening.
]

;W[ei]
(;B[ej]
;W[fj]
;B[ek]
;W[fk]
;B[el]
;W[hc]C[Because now I think white would definitely just tenuki. The only thing with this low move is that black can attach.
]

(;B[cn]C[Even if you do get the shape here.
]

)(;B[hd]
;W[id]
;B[he]
;W[gb]
;B[eb]C[Black is perfectly fine with being able to protect the top side.
]

))(;B[dk]
;W[hc]
;B[hd]
;W[id]
;B[he]
;W[ek]
;B[el]
;W[fk]
;B[dm]
;W[gc]
;B[fc]
;W[fl]
;B[kd]C[KataGo gives black -3.1 pp for 17. C11
]

)))(;B[ci]C[Black descending would also be a possible move. For this one, I kind of had to look at the AI.
]

;W[cm]
(;B[cl]
;W[bl]
;B[dl]
;W[bk]
;B[dj]
(;W[gf]C[If we back off as white, we find a little bit of an issue if we play this way.
]

;B[fh]TR[gf]C[If white jumps out normally, black can actually just take this.
]

)(;W[fh]TR[cf]TR[df]TR[dh]TR[cp]TR[cq]C[So the correct move is actually to just play here directly. It shows that if black doesn't play severely against the three stones, you end up getting attacked on both sides; although there is still aji in the corner, so it's not completely unplayable.
]

))(;B[fi]
;W[fh]
;B[ei]
;W[eh]
;B[di]
;W[gh]
;B[hj]
;W[ih]
;B[id]
;W[ji]C[KataGo gives black -1.9 pp for 15. C11
]

)))(;W[cj]C[You don't have to attach. You could definitely play a pincer.
]

;B[eh]
;W[ej]
;B[gh]C[But there are definitely a lot of issues here for black.
]

(;W[gj]C[For example, we could play here.
]

)(;W[fb]
(;B[gc]
;W[cc]
;B[be]
;W[dc]TR[gc]SQ[cf]SQ[df]C[If black plays on the outside, white is going to sacrifice its outside stones for something like this and would come live in the corner.
]

)(;B[ec]
;W[nc]TR[ec]XX[cf]XX[df]SQ[bh]C[If black attacks us, we'd play all the way over here after forcing black to secure a relatively small corner. We'd also have an attachment at B12 later to connect out our two stones.
]

))(;W[ql]
;B[kc]
;W[jf]
;B[jh]
;W[hj]C[Game 37 continued this way.
]

)))(;W[cc]C[I imagine we'd get into this position as a human.
]

(;B[dc]
;W[cd]
;B[de]
(;W[bf]
;B[eg]C[This is kind of an older joseki.
]

;W[kc]
;B[ic]C[Black would get the top and white could come and split.
]

;W[nc]
;B[oc]
;W[nd]C[We'd have something like this.
]

)(;W[cb]C[But nowadays, even the AI would play like this.
]

;B[df]
;W[cg]
;B[dg]
;W[bh]
;B[bi]
(;W[jc]TR[bi]C[That way, when black double-hanes, white can come back and play on the top side one line closer than before.
]

(;B[lc]C[The opponent would be forced to play here.
]

;W[gc]C[White can make a two-space on either side.
]

)(;B[hc]C[This would obviously be too close to the wall.
]

;W[nc]C[Black would not want to do this.
]

))(;W[kc]
;B[ic]TR[kc]C[If white plays at L17, black is happy to extend this way.
]

;W[nc]
;B[oc]
;W[nd]TR[cb]C[White would be forced to do this, but the reason that this is okay is that now we're actually undercutting black at this point.
]

)))(;B[cd]
;W[dc]
;B[ed]
;W[fb]
;B[qp]
;W[qq]
;B[rq]
;W[qo]
;B[rp]
;W[pn]
;B[ro]
;W[po]
;B[qm]
;W[ol]
;B[cn]
;W[fq]
;B[dk]
;W[gc]
;B[gd]
;W[hd]C[KataGo gives white -0.6 pp for 10. C17
]

)))(;B[pq]
;W[dp]
;B[pd]
;W[dd]
;B[qk]C[For those who don't understand what I mean at move 75, in the old chinese fuseki, players had moves like this (all comments on this game by Tang Jie, CHN 5d).
]

;W[po]C[White could play here.
]

;B[np]C[Black played like this.
]

;W[qq]
;B[qr]
;W[qp]
;B[rr]C[And then here.
]

;W[qm]C[And I believe this was the move. I haven't played the Chinese fuseki in a while.
]

;B[nn]C[I believe it was the jump.
]

;W[om]
;B[on]
;W[pn]
;B[pl]
;W[pm]
(;B[jq]C[And then black can play on the side here.
]

;W[rk]C[And then white would try to live like this.
]

)(;B[nk]C[Or black would play here to force white to live locally.
]

;W[rk]
;B[rj]
;W[rl]C[I believe this is the shape; someone could definitely correct me on that.
]

)))