Interesting position from Facebook

[Lyudmil Tsvetkov]

Check again carefully the e5 line.

SF says 130cps white advantage at first, but in the very next plies its score drops to half of that, and then further. What is certain is that SF evaluates the position arising after e5 very wrong - with huge white edge instead of 0.0.

I say e5 is the best move.

I will take the challenge if you will play e5.
kgburcham

Man, and who will mow the lawn?

[Lyudmil Tsvetkov]

I will play e5.

[d]3rk2r/p2n1pbp/1pb3p1/q1p1p2P/3PP3/2PBB3/P2Q1PP1/1R2K1NR w - e6 0 2

For those following with engines, I think the FEN should be
[d]3rk2r/p2n1pbp/1pb3p1/q1p1p2P/3PP3/2PBB3/P2Q1PP1/1R2K1NR w Kk e6 0 2

The OP believes both sides may still castle kingside.

Yeah, I checked the game, both sides can castle.

I do not see why black should have any advantage on the main diagram before Shirov's Ne5: black has better development, but white has pawn on h5 and excellent pawn center. For me, it is completely equal, and engines just like Shirov's move as they more or less play like Shirov.

SF's 85cps black edge after Ne5 does not change even an inch in the course of 20 plies, so this is just an eval assessment, and not a real advantage.

SF's 130cps white edge after e5 goes down to 30cps in the course of the same 20 plies, so again just an eval assessment, and not a real advantage.

The interesting position is this one, which arises after e5 d5 (h6 changes nothing) Bb7:

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

SF says 130cps white edge, and I see it is clear 0.0. Why so?

Well, SF has the following eval terms influencing eval here:

- trapped h8 black rook, mobility lower than 3, and gives 80cps for the mg, or 40cps normal values penalty for Rh8; this rook will remain trapped for quite a while, so SF always penalises it; on the other hand, white Rh1 is not trapped, as mobility is and will be larger than 3, so no penalty here

- d5 passer that is connected pawn also, maybe another 50cps bonus or so, so when you also add some mobility, the white edge rises to 130cps

That is what SF sees, but, I also see the following that SF does not see:

- the situation with the black trapped rook can not be exploited, as the center is closed and white can not beneficially open it, so I do not give penalty to Rh8

- a d5 central connected passer is extremely strong, but, when it is blocked by an enemy piece and the position is relatively closed as above, a blocking piece, like Bd6 or Nd6 completely neutralises any value the passer has, so I also do not give bonus for white's d5 connected passer

When you add up things, my score is 0.0 or very close to it after e5.

So a trapped rook is a very useful and true chess term, as well as connected central passers, but some specific conditions are able to change those.

What I believe is that the position on the main diagram is completely equal, and there are a range of moves leading to a draw, of which Ne5 is just one: e5, f5 etc. also result in a draw. Again, I do not see why black should have any advantage on the main diagram, SF's preference is a pure eval choice.

[Lyudmil Tsvetkov]

Never mind. I see that it is after Lyudmil's pawn push.

Man, do not get angry. I got into the habit of hijacking threads after my posting rights were somewhat reduced...

[Lyudmil Tsvetkov]

And that is the reason why SF (and probably many other engines) prefer Ne5.

In the Ne5 line, after a short while, SF reaches a position like this:

[d]3rkb1r/p3pp1p/1p4pP/q3P3/3Bb3/2Pp1N2/P2Q1PP1/1R2K2R w - - 0 6

and believe it or not, evaluates it with 90cps! black advantage.

Position is completely equal, 0.0.

SF gives 50cps bonus for black d3 passer that will fall in due time, and does not give 50cps bonus for white h6 pawn, a blocked pawn on the 6th rank that is not connected or storming pawn.

But, the h6 pawn equalises.

How many times implementing blocked on 6th was tried in SF, and it fails and fails. Obviously, because of implementation details, but this and related insufficient space advantage in terms of pawns loses SF points.

Of course, after a while, SF sees this line is 0.0, just like the f5 and e5 lines. I see absolutely no distinction in terms of outcome between the 3 lines, so one can not say Ne5 is the best move.

There are simply positions that are equal, and positions that are won.

[MikeGL]

Yeah, I checked the game, both sides can castle.

I do not see why black should have any advantage on the main diagram before Shirov's Ne5: black has better development, but white has pawn on h5 and excellent pawn center. For me, it is completely equal, and engines just like Shirov's move as they more or less play like Shirov.

SF's 85cps black edge after Ne5 does not change even an inch in the course of 20 plies, so this is just an eval assessment, and not a real advantage.

SF's 130cps white edge after e5 goes down to 30cps in the course of the same 20 plies, so again just an eval assessment, and not a real advantage.

The interesting position is this one, which arises after e5 d5 (h6 changes nothing) Bb7:

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

SF says 130cps white edge, and I see it is clear 0.0. Why so?

Well, SF has the following eval terms influencing eval here:

- trapped h8 black rook, mobility lower than 3, and gives 80cps for the mg, or 40cps normal values penalty for Rh8; this rook will remain trapped for quite a while, so SF always penalises it; on the other hand, white Rh1 is not trapped, as mobility is and will be larger than 3, so no penalty here

- d5 passer that is connected pawn also, maybe another 50cps bonus or so, so when you also add some mobility, the white edge rises to 130cps

That is what SF sees, but, I also see the following that SF does not see:

- the situation with the black trapped rook can not be exploited, as the center is closed and white can not beneficially open it, so I do not give penalty to Rh8

- a d5 central connected passer is extremely strong, but, when it is blocked by an enemy piece and the position is relatively closed as above, a blocking piece, like Bd6 or Nd6 completely neutralises any value the passer has, so I also do not give bonus for white's d5 connected passer

When you add up things, my score is 0.0 or very close to it after e5.

So a trapped rook is a very useful and true chess term, as well as connected central passers, but some specific conditions are able to change those.

What I believe is that the position on the main diagram is completely equal, and there are a range of moves leading to a draw, of which Ne5 is just one: e5, f5 etc. also result in a draw. Again, I do not see why black should have any advantage on the main diagram, SF's preference is a pure eval choice.

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

I also checked with different engines, it seems they all show +1 point something, although those were shallow searches. Probably saw a pawn up down the line, and also following the line seems to end up with a bad bishop for black due to Be3 vs Bg7, which is a plus for white due to no activity for Bg7. Strange engines, because as what I understood those engines will only show 100cp or more if it is 1 pawn up. Will try to run it later on a different machine if I have time.

[Lyudmil Tsvetkov]

Yeah, I checked the game, both sides can castle.

I do not see why black should have any advantage on the main diagram before Shirov's Ne5: black has better development, but white has pawn on h5 and excellent pawn center. For me, it is completely equal, and engines just like Shirov's move as they more or less play like Shirov.

SF's 85cps black edge after Ne5 does not change even an inch in the course of 20 plies, so this is just an eval assessment, and not a real advantage.

SF's 130cps white edge after e5 goes down to 30cps in the course of the same 20 plies, so again just an eval assessment, and not a real advantage.

The interesting position is this one, which arises after e5 d5 (h6 changes nothing) Bb7:

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

SF says 130cps white edge, and I see it is clear 0.0. Why so?

Well, SF has the following eval terms influencing eval here:

- trapped h8 black rook, mobility lower than 3, and gives 80cps for the mg, or 40cps normal values penalty for Rh8; this rook will remain trapped for quite a while, so SF always penalises it; on the other hand, white Rh1 is not trapped, as mobility is and will be larger than 3, so no penalty here

- d5 passer that is connected pawn also, maybe another 50cps bonus or so, so when you also add some mobility, the white edge rises to 130cps

That is what SF sees, but, I also see the following that SF does not see:

- the situation with the black trapped rook can not be exploited, as the center is closed and white can not beneficially open it, so I do not give penalty to Rh8

- a d5 central connected passer is extremely strong, but, when it is blocked by an enemy piece and the position is relatively closed as above, a blocking piece, like Bd6 or Nd6 completely neutralises any value the passer has, so I also do not give bonus for white's d5 connected passer

When you add up things, my score is 0.0 or very close to it after e5.

So a trapped rook is a very useful and true chess term, as well as connected central passers, but some specific conditions are able to change those.

What I believe is that the position on the main diagram is completely equal, and there are a range of moves leading to a draw, of which Ne5 is just one: e5, f5 etc. also result in a draw. Again, I do not see why black should have any advantage on the main diagram, SF's preference is a pure eval choice.

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

I also checked with different engines, it seems they all show +1 point something, although those were shallow searches. Probably saw a pawn up down the line, and also following the line seems to end up with a bad bishop for black due to Be3 vs Bg7, which is a plus for white due to no activity for Bg7. Strange engines, because as what I understood those engines will only show 100cp or more if it is 1 pawn up. Will try to run it later on a different machine if I have time.

One of the main lines is: e5 d5 Bb7 f3 Ba6 c4 (Ba6 is even weaker) Qd2 Bd2 0-0 (so castling rights are important here indeed) Ne2 f5, and after a while SF reaches a position like that:

[d]r7/p2nk3/1p3bp1/1PpPp3/2P1Pp2/2KB1P2/5BP1/R7 w - - 0 15

Here I look anxiously at SF score, well, this engine will not scare me once more, will it?, but it does! - 100cps white advantage.
I really got scared...

well, it is obvious the above simple ending is nothing but a draw. SF assesses it with 100cps white edge, for me it is 0.0, as I have much more refined eval.

Good luck to SF and you winning this.

[Dann Corbit]

Never mind. I see that it is after Lyudmil's pawn push.

Man, do not get angry. I got into the habit of hijacking threads after my posting rights were somewhat reduced...

I would never get angry at you. You are one of the most interesting posters around. I read everything that you write.

I just got confused a little.

[MikeGL]

Yeah, I checked the game, both sides can castle.

I do not see why black should have any advantage on the main diagram before Shirov's Ne5: black has better development, but white has pawn on h5 and excellent pawn center. For me, it is completely equal, and engines just like Shirov's move as they more or less play like Shirov.

SF's 85cps black edge after Ne5 does not change even an inch in the course of 20 plies, so this is just an eval assessment, and not a real advantage.

SF's 130cps white edge after e5 goes down to 30cps in the course of the same 20 plies, so again just an eval assessment, and not a real advantage.

The interesting position is this one, which arises after e5 d5 (h6 changes nothing) Bb7:

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

SF says 130cps white edge, and I see it is clear 0.0. Why so?

Well, SF has the following eval terms influencing eval here:

- trapped h8 black rook, mobility lower than 3, and gives 80cps for the mg, or 40cps normal values penalty for Rh8; this rook will remain trapped for quite a while, so SF always penalises it; on the other hand, white Rh1 is not trapped, as mobility is and will be larger than 3, so no penalty here

- d5 passer that is connected pawn also, maybe another 50cps bonus or so, so when you also add some mobility, the white edge rises to 130cps

That is what SF sees, but, I also see the following that SF does not see:

- the situation with the black trapped rook can not be exploited, as the center is closed and white can not beneficially open it, so I do not give penalty to Rh8

- a d5 central connected passer is extremely strong, but, when it is blocked by an enemy piece and the position is relatively closed as above, a blocking piece, like Bd6 or Nd6 completely neutralises any value the passer has, so I also do not give bonus for white's d5 connected passer

When you add up things, my score is 0.0 or very close to it after e5.

So a trapped rook is a very useful and true chess term, as well as connected central passers, but some specific conditions are able to change those.

What I believe is that the position on the main diagram is completely equal, and there are a range of moves leading to a draw, of which Ne5 is just one: e5, f5 etc. also result in a draw. Again, I do not see why black should have any advantage on the main diagram, SF's preference is a pure eval choice.

[d]3rk2r/pb1n1pbp/1p4p1/q1pPp2P/4P3/2PBB3/P2Q1PP1/1R2K1NR w - - 0 3

I also checked with different engines, it seems they all show +1 point something, although those were shallow searches. Probably saw a pawn up down the line, and also following the line seems to end up with a bad bishop for black due to Be3 vs Bg7, which is a plus for white due to no activity for Bg7. Strange engines, because as what I understood those engines will only show 100cp or more if it is 1 pawn up. Will try to run it later on a different machine if I have time.

One of the main lines is: e5 d5 Bb7 f3 Ba6 c4 (Ba6 is even weaker) Qd2 Bd2 0-0 (so castling rights are important here indeed) Ne2 f5, and after a while SF reaches a position like that:

[d]r7/p2nk3/1p3bp1/1PpPp3/2P1Pp2/2KB1P2/5BP1/R7 w - - 0 15

Here I look anxiously at SF score, well, this engine will not scare me once more, will it?, but it does! - 100cps white advantage.
I really got scared...

well, it is obvious the above simple ending is nothing but a draw. SF assesses it with 100cps white edge, for me it is 0.0, as I have much more refined eval.

Good luck to SF and you winning this.

What I mean is, after Ba6 the LS.Bishop will be exchanged with Bd3xa6, hence avoiding a bad bishop at Bd3. Then white pushes the c4 pawn after removing his own LSBishop, giving white a very good Be3 compared to Bg7.

[Lyudmil Tsvetkov]

This is the line where SF score drops steeply in a range of lines.

[d]3rk2r/p2n1pbp/qp4p1/2pPp2P/4P3/2P1BP2/P2Q2P1/1R2K1NR w Kk - 0 4

And how you win this?

For me, this is fully equal, neither side has sufficient relative advantages to win. In order to win, you need a bit of this and a bit of that, but above this is not present, relative advantages are only compensated.

white has nice g2-d5 chain, nice d5 passer, nice h5 lever, black's bishop is with many pawns of its colour, but, at the same time, a2 is isolated, active black queen, Ng1 still has not developed, white king not castled, d5 passer can be easily blocked, etc.

So, as I see it, fully equal, neither side has sufficient relative advantages to claim a win.

Of course, you need a single mistake to win, which often happens in human games, but here we talk what the objective outcome of the position is.