All About Chess Evaluation!

Comments on https://lichess.org/@/bigbrainchess_08/blog/all-about-chess-evaluation/IO3573MJ

in static equilibrium you said theoretical draw , boring position and fortress
but engines cant detect fortresses right?
FEN 4r3/3k4/2p5/ppPp4/3P4/PP3PP1/3P1K2/8 w - - 0 1
in this position White to move plays 1.¢f1 and the position
is a draw because his king can, from the squares
f1, f2, g1 and g2, control attempts by the rook
to enter through the e- and h-files. The pawns
on f3 and g3 are effective barriers to the black
king. The queenside is in effect blocked; on
...a4 there comes b3-b4, and on ...b4 comes
a3-a4.
All computer programs evaluate this position
as easily won for Black. Why is this?
The engine looks for a line of finite length
with the best moves for both sides, and then
evaluates the final position according to many
criteria (material, the activity of the pieces, the
position of the kings, the pawn structures and
so on). The fortress is, however, something
permanent, something ‘eternal’. The position
in the above diagram doesn’t really change in
a hundred or a thousand moves. Programs
always analyse only to a certain final depth,
and so cannot recognize the fortress.

"the equivalent advantage if a player were up that many pawns"

• That is no longer the case; now +1 is defined as the engine wins 50% of games from this position against itself.

@tpr said ^

"the equivalent advantage if a player were up that many pawns"

• That is no longer the case; now +1 is defined as the engine wins 50% of games from this position against itself.

Yes, this is the formal definition, although I do prefer to mention the previous definition just so that it is somewhat easier to understand. When playing a live game or analysing through the engine, it is a rough reference point with which to evaluate the strength of your advantage.

@vi_haa_20 said ^

in static equilibrium you said theoretical draw , boring position and fortress
but engines cant detect fortresses right?
FEN 4r3/3k4/2p5/ppPp4/3P4/PP3PP1/3P1K2/8 w - - 0 1
in this position White to move plays 1.¢f1 and the position
is a draw because his king can, from the squares
f1, f2, g1 and g2, control attempts by the rook
to enter through the e- and h-files. The pawns
on f3 and g3 are effective barriers to the black
king. The queenside is in effect blocked; on
...a4 there comes b3-b4, and on ...b4 comes
a3-a4.
All computer programs evaluate this position
as easily won for Black. Why is this?
The engine looks for a line of finite length
with the best moves for both sides, and then
evaluates the final position according to many
criteria (material, the activity of the pieces, the
position of the kings, the pawn structures and
so on). The fortress is, however, something
permanent, something ‘eternal’. The position
in the above diagram doesn’t really change in
a hundred or a thousand moves. Programs
always analyse only to a certain final depth,
and so cannot recognize the fortress.

I get your point, and it's mainly because just like you said engines have a finite depth capability. Typically for lichess SF 18 it is 99 half moves, so it typically cannot even detect the 50 move rule. However, fortress has a broad definition, so it really depends on complex fortress versus fortresses that even the engine depicts as advantageous for a side but tablebase indicating it as a draw. So yes there are some limitations but it really depends. For example the wrong coloured bishop endgame the engine gives +/- 0.9 or so, but tablebase indicates a draw.

@BigBrainChess_08 said ^

in static equilibrium you said theoretical draw , boring position and fortress
but engines cant detect fortresses right?
FEN 4r3/3k4/2p5/ppPp4/3P4/PP3PP1/3P1K2/8 w - - 0 1
in this position White to move plays 1.¢f1 and the position
is a draw because his king can, from the squares
f1, f2, g1 and g2, control attempts by the rook
to enter through the e- and h-files. The pawns
on f3 and g3 are effective barriers to the black
king. The queenside is in effect blocked; on
...a4 there comes b3-b4, and on ...b4 comes
a3-a4.
All computer programs evaluate this position
as easily won for Black. Why is this?
The engine looks for a line of finite length
with the best moves for both sides, and then
evaluates the final position according to many
criteria (material, the activity of the pieces, the
position of the kings, the pawn structures and
so on). The fortress is, however, something
permanent, something ‘eternal’. The position
in the above diagram doesn’t really change in
a hundred or a thousand moves. Programs
always analyse only to a certain final depth,
and so cannot recognize the fortress.

I get your point, and it's mainly because just like you said engines have a finite depth capability. Typically for lichess SF 18 it is 99 half moves, so it typically cannot even detect the 50 move rule. However, fortress has a broad definition, so it really depends on complex fortress versus fortresses that even the engine depicts as advantageous for a side but tablebase indicating it as a draw. So yes there are some limitations but it really depends. For example the wrong coloured bishop endgame the engine gives +/- 0.9 or so, but tablebase indicates a draw.

i agree with you but engines cant detect almost 99% of fortresses as mentioned above " The fortress is, however, something

permanent, something ‘eternal’ "

"the strength of your advantage"

• +1 is enough to win a game.
  'The winning of a pawn among good players of even strength often means the winning of the game.' - Capablanca
  An engine wins 50% of games from a +1 position against itself and draws the other 50%, so a human should win 50% of games against a fellow human.