(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2]
RU[Chinese]SZ[19]KM[7.50]
PW[AlphaGo]PB[Ke Jie]WR[9p]BR[9p]DT[2017-05-23]EV[Future of Go Summit]RO[1]PC[Wuzhen, China]C[KataGo 30 blocks s2.4G analyzing AlphaGo Master moves in AlphaGo vs Ke Jie match game 1. (notes by lightvector as of March 2020)

It's not clear how strong AlphaGo Master is. There is a non-trivial chance that LZ and/or KataGo have matched or it already, and it seems to me quite likely that some of the Chinese commercial bots have (indeed, I'd be very surprised if Fine Art isn't well past where AlphaGo Master was).

How often does AlphaGo Master's move match KataGo's preference in this game? Very often! It's amazing how often such bots agree almost exactly with each other.

The reason for this could have to do with them being similar in strength, but we also know that even top human pros similar in strength could often prefer very different moves. Plus we know that Master was bootstrapped off of human data, whereas KataGo (despite having some Go-specific inputs) is still zero with respect to its training data.

So it seems also possible that what's going on is simply we are seeing convergence on "correct" play. I'd guess such strong bots are capable of understanding the positions in this "simple" game very well ("simple" by bot standards, at least) , and therefore they often agree because the moves are the correct moves. Maybe there would be more disagreement in "messy" games like you sometimes see in bot vs bot.

And about the differences we see still? Keep in mind that the winrates for black pretty steadily go down in this game, and it's not too long before the game is basically over, so differences in move preferences could simply be differences bot ideas on how to secure a won game, or simply differences in style, or in some cases differences in the order of moves that don't make that big of a difference.

-------------------------------

In the first 122 moves of the game:

AlphaGo move matches Kata's top preference: 48 times
AlphaGo move matches Kata's 2nd preference: 10 times
AlphaGo move matches Kata's 3rd preference: 2 times
AlphaGo move matches Kata's 4rd preference: 1 time
AlphaGo move was not in Kata's top 4: never

Method: Ran analysis in Lizzie until eyeballing the top several moves their visits approximately summed to 100k (sometimes a little more since I didn't catch it in time).

Obviously results could vary on a second run due to randomness.

On each turn, Kata's most notable preferred moves are noted with the approx number of visits each received, as well as something like, e.g.  "others < 2k" indicating each other move received less than 2000 playouts.

]RE[W+0.50]
;B[qd]
;W[pp]LB[dd:A][dp:C][pp:B]C[Second choice.

A 32k 
B 29k
C 19k
Others < 5k]
;B[cc]
;W[cp]LB[cp:A][dp:B][dq:C]C[First choice.

A 80k
B 15k
C 5k
Others < 1k]
;B[nc]
;W[fp]LB[pd:A][ep:C][fp:D][fq:B]C[Fourth choice.

A 80k
B 15k
C 6k
D 2k
Others < 1k]
;B[qq]
;W[pq]LB[qp:A][pq:B]C[Second choice.

A 94k 
B 5k
Others < 2k]
;B[qp]
;W[qn]LB[qn:A][po:B]C[First choice.

A 98k
B 2k
Others < 1k]
;B[qo]
;W[po]LB[po:A]C[First choice.

A 100k
Others < 1k]
;B[rn]
;W[qr]LB[pd:B][dj:C][qr:A]C[First choice.

A 95K
B 10K
C 5K
Others < 4K]
;B[rr]
;W[rm]LB[pd:C][rm:B][ro:A]C[Second choice.

A 94k
B 4k
C 3k
Others < 1k]
;B[pr]
;W[or]LB[pm:A][pn:B][or:C]C[Third choice

A 83k
B 15k
C 9k
Others < 1k]
;B[pn]
;W[qm]LB[qm:A]C[First choice.

A 100k
Others < 1k]
;B[qs]
;W[on]LB[on:A]C[First choice

A 95k
Others < 3k]
;B[dj]
;W[nk]LB[qf:B][ch:C][nk:A]C[First choice.

A 76k
B 21k
C 5k
Others < 2k]
;B[ph]
;W[ch]LB[ch:A]C[First choice.

A 92k
Others < 2k]
;B[cf]
;W[eh]LB[eh:A][ei:B]C[First choice.

A 88k
B 10k
Others < 2k]
;B[ci]
;W[de]LB[dd:B][de:A][pf:D][cg:C]C[First choice.

A 39k
B 32k
C 20k
D 7k
Others < 3k]
;B[df]
;W[dc]LB[dc:A]C[First choice

A 97k
Others < 2k]
;B[cd]
;W[dd]LB[dd:A][ef:B]C[First choice.

A 93k
B 5k
Others < 2k]
;B[ef]
;W[di]LB[di:A]C[First choice.

A 96k
Others < 2k]
;B[ei]
;W[dh]LB[dh:A]C[First chioce.

A 100k
Others < 1k]
;B[cj]
;W[ce]LB[gd:C][ce:A][ej:B]C[First choice.

A 93k
B 4k
C 3k
Others < 2k]
;B[be]
;W[bf]LB[bf:A]C[First choice.

A 98k
Others < 1k]
;B[bg]
;W[bd]LB[bd:A]C[First choice.

A 100k.
Others < 1k.]
;B[af]
;W[bc]LB[bc:A]C[First choice.

A 100k
Others < 1k]
;B[fi]
;W[cm]LB[pf:B][fh:C][cm:A]C[First choice.

A 93k
B 4k
C 3k
Others < 2k]
;B[hq]
;W[ek]LB[fh:D][ek:A][fr:B][gr:C]C[First choice.

A 55k
B 27k
C 12k
D 5k
Others < 2k]
;B[fh]
;W[gq]LB[ej:A][gq:C][fr:B]C[Third choice

A 86k
B 10k
C 4k
Others < 3k]
;B[hp]
;W[ej]LB[ej:A][jq:B]C[First choice

A 90k
B 4k
Others < 3k]
;B[eq]
;W[gr]LB[gr:A][hr:B]C[First choice.

A 97k
B 3k
Others < 1k]
;B[cq]
;W[dp]LB[bp:B][dp:A]C[First choice.

A 97k
B 3k
Others < 1k]
;B[dq]
;W[ep]LB[fo:B][ep:A]C[First choice.

A 81k
B 16k
Others < 3k]
;B[bp]
;W[bh]LB[bh:A][bo:B]C[First choice.

A 52k
B 48k
Others < 1k]
;B[ah]
;W[bo]LB[fg:A][bi:C][bo:B]C[Second choice

A 19k  (first choice due to LCB)
B 77k
C 3k]
;B[bq]
;W[fg]LB[fg:B][jp:C][kp:A][ds:D]C[Second choice.

A 65k
B 27k
C 6k
D 4k
Others < 3k]
;B[gg]
;W[kp]LB[od:B][kp:A]C[First choice.

A 97k
B 3k
Others < 2k]
;B[ko]
;W[jo]LB[od:C][jo:A][jp:B]C[First choice.

A 86k
B 8k
C 4k
Others < 3k]
;B[jn]
;W[in]LB[in:A][io:B]C[First choice.

A 95k
B 3k
Others < 3k]
;B[jp]
;W[io]LB[io:A]C[First choice.

A 100k
Others < 1k]
;B[lp]
;W[kq]LB[kq:A]C[First choice

A 100k
Others < 1k]
;B[lq]
;W[kr]LB[jq:B][kr:A]C[First choice.

A 93k
B 6k
Others < 2k]
;B[lr]
;W[ir]LB[ir:A]C[First choice

A 98k
Others < 2k]
;B[kn]
;W[il]LB[il:B][im:A]C[Second choice.

A 95k
B 4k
Others < 1k]
;B[oq]
;W[pf]LB[oc:D][pf:B][qf:A][mo:C]C[Second choice

A 65k
B 18k
C 15k
D 7k
Others < 2k]
;B[nh]
;W[rf]LB[nf:C][rf:A][pm:B]C[First choice

A 74k
B 18k
C 4k
Others < 3k]
;B[od]
;W[qi]LB[qi:A]C[First choice

A 99k
Others < 1k]
;B[qg]
;W[rd]LB[rd:A][qe:B]C[First choice

A 97k
B 2k
Others < 1k]
;B[qf]
;W[qe]LB[qe:A][re:B]C[First choice

A 94k
B 6k]
;B[pe]
;W[re]LB[re:A]C[First choice

A 100k
Others < 1k]
;B[qc]
;W[rg]LB[oe:B][rg:A]C[First choice

A 94k
B 6k
Others < 1k]
;B[kh]
;W[ic]LB[hc:C][ic:B][kg:A]C[Second choice

A 25k  (first choice due to LCB)
B 60k
C 16k
Others < 4k]
;B[gc]
;W[kc]LB[kc:A][lc:B]C[First choice

A 97k
B 2k
Others < 2k]
;B[jd]
;W[id]LB[hb:A][id:B]C[Second choice

A 54k
B 44k
Others < 3k]
;B[ge]
;W[hb]LB[hb:A][hg:B]C[First choice

A 78k
B 19k
Others < 2k]
;B[gb]
;W[jf]LB[mb:C][jf:A][hg:B]C[First choice

A 65k
B 23k
C 12k
Others < 4k]
;B[je]
;W[ie]LB[ie:A]C[First choice

A 97k
Others < 3k]
;B[ld]
;W[hg]LB[lc:B][fd:C][hg:A]C[First choice

A 82k
B 14k
C 4k
Others < 3k]
;B[eg]
;W[lc]LB[lc:A][fd:B][kg:D][jh:C]C[First choice

A 74k
B 15k
C 8k
D 5k
Others < 2k]
;B[le]
;W[hf]LB[nb:D][mc:E][fd:B][hf:A][ih:C]C[First choice

A 35k
B 27k
C 15k
D 12k
E 9k
Others < 6k]
;B[qh]
;W[rh]LB[rh:A]C[First choice.

A 100k
Others < 1k]
;B[pi]
;W[qj]LB[pj:B][qj:A]C[First choice

A 94k
B 5k
Others < 2k
]
;B[gk]
;W[fd]LB[mc:D][fd:B][lf:A][kg:C]C[Second choice

A 86k
B 7k
C 5k
D 5k
Others < 4k]
;B[gd]
;W[lf]LB[nb:D][mc:C][lf:A][kg:B]C[First choice

A 85k
B 10k
C 4k
D 4k
Others < 2k

It's just about reached (super-macro) endgame after this. So we can end notes here.

Also KG says black has only 2.5% winrate and is trailing by 11 points. AlphaGo will gradually bleed all these points back to win by 0.5 (the actual game result) whereas KG won't unnecessarily give away points. So if we were to continue, we would expect increasing differences between the two anyways as KG maximizes the win while AlphaGo doesn't care.

End of notes.

]
;B[mf]
;W[lg]
;B[gm]
;W[gn]
;B[fn]
;W[go]
;B[dl]
;W[mo]
;B[oo]
;W[pm]
;B[op]
;W[mg]
;B[nf]
;W[lo]
;B[nn]
;W[lm]
;B[pn]
;W[dk]
;B[ck]
;W[cl]
;B[el]
;W[bk]
;B[bi]
;W[li]
;B[ii]
;W[ds]
;B[dr]
;W[hi]
;B[ik]
;W[jk]
;B[ij]
;W[md]
;B[mc]
;W[ke]
;B[me]
;W[kd]
;B[om]
;W[ls]
;B[ms]
;W[ks]
;B[nr]
;W[ng]
;B[og]
;W[es]
;B[cs]
;W[fr]
;B[er]
;W[fs]
;B[bs]
;W[hl]
;B[pl]
;W[ql]
;B[rc]
;W[ro]
;B[rp]
;W[sn]
;B[hm]
;W[im]
;B[kk]
;W[kj]
;B[lk]
;W[jl]
;B[mj]
;W[mi]
;B[nj]
;W[pk]
;B[fm]
;W[cn]
;B[ol]
;W[ok]
;B[ni]
;W[ih]
;B[ji]
;W[mb]
;B[nb]
;W[lb]
;B[fe]
;W[cb]
;B[mp]
;W[mm]
;B[eb]
;W[na]
;B[oa]
;W[ma]
;B[qb]
;W[bj]
;B[ai]
;W[aj]
;B[ag]
;W[gl]
;B[fk]
;W[bl]
;B[kg]
;W[kf]
;B[ib]
;W[jb]
;B[ga]
;W[ha]
;B[ed]
;W[ec]
;B[fc]
;W[gf]
;B[ff]
;W[gj]
;B[hk]
;W[hh]
;B[fj]
;W[no]
;B[fq]
;W[hr]
;B[kl]
;W[km]
;B[mn]
;W[ln]
;B[nl]
;W[db]
;B[da]
;W[ca]
;B[ea]
;W[np]
;B[nq]
;W[oj]
;B[oi]
;W[en]
;B[em]
;W[eo]
;B[dm]
;W[dn]
;B[sp]
;W[so]
;B[hn]
;W[ho]
;B[hc]
;W[ia]
;B[ao]
;W[an]
;B[ap]
;W[sc]
;B[sb]
;W[sd]
;B[jg]
;W[ad]
;B[gh]
;W[ae]
;B[ee]
;W[ml]
;B[mk]
;W[pj]
;B[bf]
;W[nm]
;B[on]
;W[he]
;B[ig]
;W[ki]
;B[jh]
;W[fl]
;B[jj]
;W[fo]
;B[hj]
;W[gi]
;B[ll]
;W[jm]
;B[lh]
;W[mh]
;B[lj]
;W[if]
;B[hd])