A few SDK games

[greenshoes]

Recently played these, I am 9k on OGS―though that's residual from an account that's been long inactive so my "true" strength is a mystery to me... I feel my greatest weakness is early game, particularly the transition from fuseki to middle-game really leaves me empty-handed if I have sente. Maybe I am playing too passive?

[jlt]

I think the most important lesson you should learn from game 2 is

• Don't let your opponent hane at the head of your two stones unless you have a very good reason to do so.
• Seriously consider to hane at the head of your opponent's two stones if you have the possibility.

For instance:
At move 14, you played away, but this allows your opponent to play later at D14. This would give White a good shape (tiger's mouth) and Black a bad shape (hane at the head of two stones). Did you consider to play at D14? Were you afraid this would not be sente? Suppose White doesn't respond, did you read what happens if you push at C14 and then cut?

continue at C6.

played away, so you know that C7 is an interesting point for both players. Not to be played immediately but interesting for later.

should have been at N17. You see that hanes at the head of your two stones.

should be at M16.

now you see that your shape is constricted.

it's rarely a good move to cut directly like that, unless you have enough support nearby.

now you see that your two stones have been haned on both sides.

After your group is almost dead so this is almost game over.

For game 1, you clearly outplayed your opponent but you made a few joseki mistakes.

is better at C6.

is usually at J17 or H16.

[Joaz Banbeck]

In game 1, move 25 is the wrong direction. Q17 would work well with your top side stones.

[EdLee]

Hi greenshoes,

game 1, very minor (~2%):

25.png (21.72 KiB) Viewed 8465 times

-15% drop for you, but cut was losing move, -46% for W (W should've taken F13 for himself), and you kept climbing to 99% by .
A few % fluctuations very minor.

[dfan]

These days, AIs and pros often prefer "blocking in the wrong direction" when someone invades underneath the star point. But that's because they understand the joseki followups very well, which often end up blocking in the right direction at the end of the day. Until one has spent some time learning those subtleties, I think it's best to play the old way, where you try to choose the blocking move that works best with the stones that it's facing.

[Bill Spight]

These days, AIs and pros often prefer "blocking in the wrong direction" when someone invades underneath the star point. But that's because they understand the joseki followups very well, which often end up blocking in the right direction at the end of the day. Until one has spent some time learning those subtleties, I think it's best to play the old way, where you try to choose the blocking move that works best with the stones that it's facing.

Besides which, the difference between blocking one way and another, in terms of AI winrate estimates with similar numbers of rollouts per candidate play, is typically less than 1%. I.e, indistinguishable from noise. Call me if it's more than 2%. So, in the mortal words of Aleister Crowley, Do what thou wilt.

Edit; In this case, I see that one option is slightly more than 2% better in terms of winrate estimates, but one option has 27k rollouts and the other has only 1.5k rollouts, which is not similar enough for a good comparison, in my book. Still, the recommended play looks better to me.

[Knotwilg]

These days, AIs and pros often prefer "blocking in the wrong direction" when someone invades underneath the star point. But that's because they understand the joseki followups very well, which often end up blocking in the right direction at the end of the day. Until one has spent some time learning those subtleties, I think it's best to play the old way, where you try to choose the blocking move that works best with the stones that it's facing.

The way I see it, "blocking in the wrong direction" also takes into account the stones present, but in a different way than we traditionally do. Traditionally we take stones present as an extension from the wall that forms when blocking, building a moyo.
The AI way looks at stones present as keeping the stability in check of the ponnuki that forms after the double hane.

The double hane has gained preference over the hane-stretch, because the former claims back the corner, while the latter builds influence. So the question then, instead of "in which direction will I form the best moyo" becomes "in which direction is the ponnuki the easiest to keep in check".