28 July 2019

A deeper, too long look into Lc0 v. Stockfish

In the previous post I took a look at the odd beginning and ending of a game between Lc0 and Stockfish. At the end white is clearly superior, so the b5 blunder doesn’t really matter.

Let’s scroll the whole game to see and spot other things that even an amateur could classify as mistakes and play better — allegedly.

Pointless check?

aabbccddeeffgghh1122334455667788
  1. … ♛h4+
  2. g3 ♛d8

Check and retreat exactly where it was. Stockfish loses tempo and allows Lc0 to advance a pawn controlling f4. Almost all the pieces are in their squares. They are mostly playing with pawns.

After the white made the black Queen flight back, it’s the bishop’s turn to be threatened and retreat:

  1. a3 ♝f8

And all black’s pieces are back in their initial squares!

At least white has a knight out.

aabbccddeeffgghh1122334455667788

Dominate the file

aabbccddeeffgghh1122334455667788
  1. … ♛c7

It seems to me like if Stockfish obeyed to a basic principle: dominate the file. But white has a different idea for that file, and a Queen is easily scared and she flies back to where it came:

  1. ♖c1 ♛d8

Then white exchanges the rooks and the double black domination of the c-file is over.

I am not hungry

aabbccddeeffgghh1122334455667788

The black knight in c4 is in danger, so black must save it:

  1. … ♞a5

Then white doesn’t need its Queen there anymore, and maybe it wants to avoid an exchange. So:

  1. ♕d2

Black maybe has an opportunity here. Maybe g×f6 isn’t that good because of the hole in the king fortress, and then white would play e×f6, and that advanced white pawn there would be a pain. I would have taken the chance — but I am a poor amateur player, and that’s why Stockfish play

  1. … ♞c4

The white Queen is in danger, so you can see why this move was done, with no rush to get the white knight in f6.

White repeats the same threat as before and black puts back its knight again, as somebody could have foreseen, and then white changes its move:

  1. ♕c1 ♞a5
  2. ♗×g6

Now black can eat at least one of a white bishop or knight. But this would weaken the king’s fortress and it would be very dangerous with the Queen ready to do ♕×h6+.

Let’s eliminate the queens from the equation first:

  1. … ♛×c1
  2. ♖×c1

Now black can gain back a knight:

  1. … g×f6

All this seems well played. Then why white has the advantage?

Do not let the pawns go

Black had to eat back the knight, but now white can play e×f6. And later black let white play d×e5 (black exchanges knights).

Instead of

  1. … ♚g8

I would have played e5. I know I am not good at all with the pawn structure, but here I would have thought that freeing the wasteful black bishop could have been more useful. My defeat would have been at least quickier!

This variation shows another (quickier!) path to final black defeat:

  1. … e5
  2. ♘×e5 ♞×e5
  3. d×e5 ♜e8
  4. ♗×f7 ♜×e5
  5. h4 ♝g4
  6. ♖c6 d4
  7. ♗c4

It makes clear again why I am an amateur player and if you go on why black played ♚g8… nonetheless, black’s situation seems doomed and its agony could last only until move 49, which is a good improvement with respect to the 118 moves remaining to the checkmate.

aabbccddeeffgghh1122334455667788

White pawns in f6 and e5 shouldn’t be there, but they are. The pawn in f6 in particular limits king’s freedom. And what about that totally wasted black bishop?

Free the bishop… woops!

After exchanging the last rooks, black tries to give some room to its bishop who stayed in the back mire for so long.

  1. … d4?

But white has an easy move to dominate the longest diagonal:

  1. ♗f3
aabbccddeeffgghh1122334455667788

Black has a plan:

  1. … h5

It offers the pawn to the bishop. Black takes it and black plays ♝b7.

Then it tries to push the pawn in the d-file to promotion, but it has a hard time also because its king can’t help if it tries to go in the white field through c and d-file (because of pawns in b5 and e5).

A passage could have been g8, h7, h6, g5… That’s before black put its bishop in h7 and let the white pawns in g- and h-file advance while moving pointlessly the king in the back-left mire.

Later it seems it tries to find the east passage, but it’s late and there are other pressuring matters…

aabbccddeeffgghh1122334455667788
  1. ♔e3

And the bishop must go back to h7 (obstructing the east passage). Black pawn can’t make it, especially now that white manages to bring its bishop in c4.

White pawns dominate and black can’t stop all of them.

Balls

From here kings and bishops dance around… Nothing interesting happens.

aabbccddeeffgghh1122334455667788

At last white decide to push a pawn forward:

  1. g6

The ballet started at move 69, after white caught the advanced black pawn, which had no hope.

Stuck in the role

aabbccddeeffgghh1122334455667788

The black bishop in g6 is stuck in its role of controlling the final rush of the white f- and h-file pawn. Black king is cut off from the scene because of the white bishop d5 (the west journey is too slow to matter). Black g-file pawn is lost.

At the move 106 white has three pawns in the sixth rank. Clearly Black can’t stop them all.

Final words

It’s hard to say where strong engines get it wrong. I’ve spotted just a couple of lost tempo opportunity for the black. Were all the problems there?

According to Stockfish itself, played here, things started to go less well for black from around the move 25.

From move 33 on, almost all the inaccuracies are black’s.

An interesting move marked as inaccuracy is

  1. … ♞a5

Black saved the knight but didn’t see ♞c×e5… Which makes black lose the knight but, well, … it must be because its position was already compromised…?

No comments:

Post a Comment

Be polite and possibly on topic.