(;GM[1]FF[4]CA[UTF-8]AP[CGoban:3]ST[2]
RU[Chinese]SZ[19]KM[7.50]
C[Copyright © 2018 弈招围棋 and Daniel Hu
weiqitv.com All rights reserved

Original video:
http://www.weiqitv.com/album/5a79d58628865447198b456a
Translator: Daniel Hu 5d

Note: ... means a break in the talking, looking at other variations before returning to the current position. 

___________

弈招围棋Alphago教学工具之十大恐怖胜率
Alphago teaching tool: Top ten shocking winning percentages part 2

Intro

"Why do standard joseki suffer such criticism?"

"How come sabaki tesujis leave a sudden drop in winning percentage?"

"Why does the star point 3-3 invasion appear so frequently?"

"What is hidden when one move makes a large difference in winning percentage?"

"Has human experience been debunked, or is this a great advance for AI skill?"

Meng Tailing 6p

Happy New Year. This is weiqitv's end of year review on Alphago's teaching tool's top 10 shocking winning percentage's latter part.
]
(;AW[dc][cd][ce][cf][dg][dp]AB[ec][dd][pd][de][df][pq]C[Number 5

The large avalance joseki

There is no question that the large avalanche is one of the most complicated joseki. 

...

Firstly I don't recommend viewers play the large avalanche. For one, the D14 push already loses some winning percentage. But if you like it, you can continue trying it out.]
;B[cc]
;W[db]
;B[bc]
;W[cb]C[From last century, in the 50's, turning inwards became fashionable, invented by Go Seigen sensei.]
;B[cg]C[...

There is certainly no problem with this cut. Moves are born from cuts.]
;W[ch]
(;B[bb]C[Let's take a look at the basic variation. This turn is a vital point

...

But the so called vital point turn. It is this move that is the worst move of all.

Firstly this drops 3.4%.

]
;W[eb]
;B[bg]
;W[bf]C[...

The reason we called the turn vital before was so that white was forced to exchange this block first.]
(;B[bh]
;W[bd]C[This is the starting model for the joseki.]
(;B[ci]C[Our old joseki was atari


]
;W[eg]
;B[fe]
;W[fc]C[but later, we decided that after white ataris white should have a slight advantage.])
(;B[fc]C[So later, it became extending here]
;W[fb]
;B[ci]C[...

But black's three moves turning twice and atari are all mistakes, so black's winning percentage drops to 36%.

And the unforeseen combination means black has clearly made a loss.]
;W[dh]
;B[fe]
;W[di]
;B[cj]C[...

But alphago judges this as black not playing go at all. Black's winning percentage is only around 36%.]
(;W[dj]C[this joseki has occured many many times. We judged the result to be close to even. But in more recent years, there has been muttering that black is ever so slightly worse. So the D14 push of the large avalanche has been seen much less.

But we viewed this as reasonably even.])
(;W[gc]
;B[gd]
;W[ef]
;B[ee]
;W[fd]C[This is a combination that it seems no human thought of before. This really is a good way to play]
(;B[ed]
;W[hc]C[We can see that white's throw in makes black very heavy and short on liberties. I think that if it is played like this, black's winning percentage will slip further again.])
(;B[gf]C[black's best move is to kosumi]
;W[fh]C[so where did black go wrong?


...

So the so-called vital point was the worst move. And treating the left lightly turned out to be an optimistic change of gears.
]))))
(;B[fc]C[Alphago thinks that black shouldn't pull the stone out, but extend first]
;W[fb]
;B[gc]C[and extend again!

It's this move that completely overturns our past understanding of this joseki.

Before we thought that obviously black had to turn to escape as otherwise what would be the point of the cut.

Extending here is clearly giving up on fighting]
;W[bd]C[after white blocks, black is clearly dead in the corner. If black descends, then even if white starts to fill liberties, black is dead, and what's more, white can keep crawling.]
;B[gb]C[so black can pretty much only turn.]
;W[ba]
;B[bh]
;W[ci]C[It looks like black has a real danger of being about to crawl on the second line.]
;B[eg]
;W[dh]
;B[eh]C[but black's best play is to atari and press from above. This keeps the winning percentage around 40%. Although black has lost a little locally, at least the game is playable.

This is the way alphago recommends as the best variation.]))
(;B[bg]C[because if black directly descends, ]
;W[bd]C[white can block]
;B[bb]
;W[ba]C[and black is dead

And after that we can see that the two black stones and the three on the outside can't both be saved.]
(;B[bh]C[if black crawls on the second line, then so what, white happy to let black do that.]
;W[ci])
(;B[ci]C[so there is pretty much only the clamp]
;W[eg]
;B[bh]
;W[fe]C[but after the keima the 3 stones are also in deep trouble.])
(;B[fe]C[alphago's answer is that black's best move is to turn at B18 and then jump. The winning percentage doesn't change much, it is still over 40%.

Obviously we don't have time to discuss the variations here today. But in the "开局库" series (Master's opening book) we will continue with a very detailed analysis. Let us await another of our end of year review videos.])))
(;AW[dd][dp]AB[pd][op][qp]C[Number 4

Slow moves are bad moves.

This opening kept appearing in the 2000s Nongshim cup. I remember the first was Lee Changho as black against Wang Xi. In the final match between the captains of China and Korea.]
;B[]
;W[qj]
;B[qh]
;W[qm]C[...

Firstly the R10 wedge wasn't the best and neither was R7, but white barely played any correct move. But these are only slightly worse.]
;B[pl]
(;W[ql]
;B[pj]C[This is a very classic shape, and I trust that viewers have played it many times.


...

The problems start here]
(;W[pk]
;B[ok]
;W[qk]
;B[pi]C[...

after wedging, black is already slightly better]
;W[ol]
;B[nk]
;W[om]C[...

But obviously it is this move, an extension, that drops white 14% from 52 to 66.

This so-called slightly slow move suddenly makes the winning percentage unbearable.

But of course black has to watch an important detail.

]
(;B[fc]C[Regarding the right side shape, we don't see it so much these days. People consider white's play to be slightly slow, but only slightly.

But we do feel that black coming around on top is very thick and will have a big influence on the rest of the game. So black is good.

And alphago?

Alphago thinks that black's winning percentage has already reached 66%! In other words from the opening 45% or so, white has already lost 20 percentage points.

so where was the mistake? In many places.]
;W[pq]C[because if black doesn't peep, then when white peeps, ]
;B[pp]C[black can only connect. This exchange is clearly good for white because ...])
(;B[pn]C[When white extends, black should exchange the peep, and then black can choose where to play next.]
;W[pm]C[...

So we can see that after this whole sequence, white only succeeded in capturing a single stone. Letting black get so thick. It is clear that white isn't happy, but no one imagined that white's winning percentage could already have reached a shocking 66%.]
(;B[fc]
;W[pq]C[if white still tries to peep, ]
;B[oq]C[black can resist]
;W[pp]
;B[qq]C[B doesn't need to block and can clamp])
(;B[nc])
(;B[jc])))
(;W[pm]C[The wedge loses 3% compared to turning out.]
;B[pk]
;W[qi]
;B[pi]
;W[rh]
;B[rg]
;W[nm]C[white exchanges two crude moves to get sente to jump out. The winning percentage is 50-50 so black has overcome the disadvantage of the komi.]))
(;W[qf]C[alphago thinks that white should invade, but this difference is also somewhat small.]))
(;AW[fc][dd][qi][pm][qn][po][dp]AB[oc][qd][qg][qo][qp][nq][pq]TR[oi]C[Number 3

Crawling on the first and second lines

This shape on the right is very common. Black has a shimari in the upper right, and when white high approached the lower right, black attached underneath.

When black pressured with R13, white tenukied to enclose the corner. Of course, white could have also jumped to defend.]
;W[]C[...

Starting from here, white's winning percentage is already better that at the start and black's winning percentage is only just over 40%.]
;B[ql]C[There is a classic side joseki here.]
;W[rm]
(;B[pj]C[so normally everyone plays keima


...


And black's invasion and keima don't seem to be a very good way to play either.]
;W[qj]C[...


There is nothing wrong with the crawl.]
(;B[qk]C[...

in fact alphago thinks black's bump is a mistake.

But from this bump, white played two extremely ridiculous moves.

At this point, black's winning percentage is only 40-41%]
(;W[rk]C[...

but after the hane, black's winning percentage suddenly jumps to 53-55%]
;B[rl]
(;W[sl]C[...

connecting underneath lost another 5%. Black's winning percentage is already going towards 60%.

Just these two standard moves connecting underneath

...

But these two moves which have been played for decades lose 20%. Black goes from clearly behind to having a large advantage.

]
;B[rj]
;W[sk]
;B[pi]
;W[ri]
(;B[qh]
;W[rh]C[I remember they used to say when I was a kid that white turning gains 2 points.]
;B[rg]C[the shape is settled locally ]
;W[ol]C[but we can also say white kosumis, a very big move.

This means that when black invades, black settles this in sente, and it is even. That was the judgement then.

But when I grew up and looked back at this shape, I felt that this seemed very strange, and one side played very badly. Starting from the S7 kosumi to connect underneath, to the turn, it seemed like white was just playing on the first and second lines. It looked very submissive, but I wasn't clear about it. It didn't seem great, but given that it was a joseki, then you accept the reality that it is a joseki.

])
(;B[rh]C[But in recent years due to human progress, we've discovered this kosumi. Compared to letting white gain two points in sente, why can't black play here first?]
(;W[qh]C[unless white squeezes]
;B[ph]
;W[sj]
;B[rg]C[but the squeeze helps black a lot in thickness. Of course, you prefer to have your stones on top.])
(;W[sj]C[Normally white captures

...

and at this point alphago tells us that black's winning percentage has reached 60%. White's winning percentage has suffered a severe slip.]
(;B[cn]C[and then we discovered that black could actually tenuki here!

And returning to white connecting underneath. White's only 3rd line stone was R10. All other moves were on the first and second line.])
(;B[ph]C[alphago thinks that black only needs to defend to be perfectly happy.]))))
(;W[pi]C[even insteading of continuing with the plan to connect underneath, alphago thinks white should cut losses and turn.]))
(;W[pi]C[In other words, white's best move is to come out]
;B[ok]C[and white's best move is still not to connect underneath but ]
(;W[nj])
(;W[oj]C[black's winning percentage is only around 40%])))
(;B[pi]C[but this is black's best move]
;W[qk]
;B[qh]
;W[pk]C[But this is already much better for white than before.]))
(;B[rj]C[the second line keima is also joseki]
;W[pk]C[I remember when I came across this many times when I was younger]
;B[pj]
;W[qj]
;B[qk]
;W[pl]
;B[ri]
;W[qh]
;B[rh]
;W[oj]C[and our teachers would tell us black running at Q11 was critical. Whether black tried to escape or not would directly impact on who was better in this joseki. If black can't run out then clearly black is dissatisfied.]))
(;AW[dd][cn][qn][dp]AB[pd][ip][fq][pq]C[Number 2

The most shocking joseki.


This is the kobayashi opening.]
;B[po]
;W[rp]
;B[pl]C[Even in the 21st century, this is a very common move

...

Firstly this move isn't great]
;W[pn]
;B[on]
;W[om]
;B[pm]C[But the biggest mistake is this cut. It seems surprising even now. If even the cut could be a mistake, then this offers evidence that the Q8 pincer isn't great.

If you pincer and don't cut, you can't justify it, so no-one thought that this cut could be the problem, and the worst mistake, dropping from 40% to 30%.]
;W[oo]
;B[nn]
;W[pp]C[I remember being told as a kid that whether this joseki was good or not depended on the position of J4.]
(;B[op]
;W[qo]C[In recent years, people do still play this shape.

In the Changqi cup, I played this against Tuo Jiaxi.
]
(;B[no]C[...


And this atari drops another 5%]
;W[oq]
;B[po]
;W[pr]
(;B[qp]C[capturing is the final bad move.]
;W[qq]
;B[pp]
;W[np]
;B[oo]C[It's when black gets ataried into a big dumpling that the winning margin is largest.]
;W[rr]
;B[nq]
;W[or]
;B[mp]C[Black's shape is painfully bad, and this joseki can no longer be called joseki.

22% is the lowest I have seen in any joseki.])
(;B[qq]C[It should be extend.]
;W[oo]
;B[np]
;W[qr]C[Sacrificing an extra stone allows black to maintain outside thickness and get sente.]))
(;B[np]C[But I pulled back, and thought black was good.

I remember Qiu Jun giving a tewari analysis once that this was good for black and very efficient.

So there has been a while where this has been considered good for black.

]
;W[qq]C[But even like this, black's percentage is 30%. The Q7 cut isn't good, and choosing the outside and letting white get this tiger's mouth means the game is already difficult.

Alphago teaches us how important the points in the corner is, and how valuable the thickness is, when the outside is thin.]))
(;AE[ip]AB[jp]C[If the stone was at the star point, then it wasn't good for black.]
;B[op]
;W[qo]
;B[no]
;W[oq]
;B[po]
;W[pr]
;B[qp]
;W[qq]
;B[pp]
;W[np]
;B[oo]
;W[rr]
;B[nq]
;W[or]
;B[mp]C[Teachers and books said that if the stone was at the star point, then this joseki wasn't great for black.

]
;AE[jp]AB[ip]C[But shift the stone and black was ok.

But I've never played as black myself. It feels bad for black to be ataried into this sort of shape. But joseki is joseki, and this has been passed down for decades. Although no-one plays black these days, this has still been passed down as joseki.

How would alphago view this?

Black's winning percentage is 22%! From the starting 45%, black has lost over 20 percent!

Black has lost a lot, more than you will ever meet. It's impossible to find a shape where you lose this much.

Let's see where black went wrong.]))
(;AW[db][dc][cd][od][ce][cf][cg][dp]AB[ec][dd][ld][qd][de][df][dg][dh][pp]C[Number 2

The magic sword is no longer the magic sword.

The magic sword arises with the two space high pincer M16.

This shape comes from the Master series played against three world champions Jiang Weijie, Kim Jiseok and Chen Yaoye.]
;B[]
;W[pg]
;B[oe]
;W[ne]
;B[of]
;W[pd]
;B[qf]
(;W[qc]
;B[qe]
;W[nf]
;B[og]
;W[rc]C[This is a very fierce and complicated joseki.]
;B[nd]
;W[md]
;B[nc]
;W[mc]
;B[ob]
;W[oc]
;B[nb]
;W[mb]C[B's cut looks like a fight, but actually it is a sacrifice.]
;B[pb]
;W[pc]
;B[me]
;W[kb]
;B[qb]
;W[rb]
;B[kc]
;W[lb]C[Black's stones are dead, but black can play a ladder breaker.]
;B[ep]C[This has been called joseki and has many decades of history. It can be called roughly even.

And alphago?

Alphago believes this really is about even. But in the process, black and white made many mistakes.])
(;W[pe]C[In all of the 3 Master games, it captured the corner stone.]
;B[pf]
;W[qe]C[There was no variation like this before, because we considered these moves too crude.]
;B[re]
;W[rd]
;B[rf]
;W[qc]
;B[nd]C[Black gets to cut]
;W[nc]
;B[md]C[Black has a thickness from the right side to the upper side and white only has the corner.

We used to think that this was clearly better for black.

But alphago thinks that black's winning percentage is only 26%. Considering that white gained a little in the upper left, white profits 15% in this joseki.

This is a huge overturning!

Alphago thinks that this can't even be called joseki.]
;W[pj]C[Black played the same in all 3 games.]
;B[pm]
;W[mi]C[White reduces the influence. Then black doesn't have anything left on the outside. And has already lost all the points in the corner.

This was a shape we had never thought of, but after alphago, it has become a classic shape. We can't explain this fully with just this variation. But expect to see more in the Master opening book series.

So this episode goes up to here. I'm sure that with this, the magic sword joseki will rapidly go extinct.

Above was all the content. Thanks to everybody for watching! Goodbye.])))