Vvarkey,Rybka 1.1 - Relic,Rybka 1.1
CSS/Pal
Freestyle Final (45+5) round 1
Everybody had to wait for two comps to finish J, but nobody complained – this time the endgame happened to be very rare and quite sophisticated (it was not just a different color bishops with setup blockade J)
_pliki/image003.jpg)
BLACK TO
MOVE
48…e4!
49.R:e4 [49.B:e4 Qf1#] 49...N:e4 50.B:e4
_pliki/image005.jpg)
BLACK TO
PLAY
Is black better in this endgame having Queen for 2 pieces
and 2 pawns? Black is certainly the one who can “try” to win, but I think this
position is a draw. Certainly this one
is a draw: five stones tablebases J
_pliki/image007.jpg)
who ever
to move draw
Ok, a lot of moves passes and white somehow let black to create passed
pawn, anyway it should be still a draw according to my “human evaluation”.
_pliki/image009.jpg)
BLACK TO
MOVE
It looks
like there is no way to push a pawn forward without loosing it…Both engines
evaluate the position the same way - black is much better. If so, (assuming it
is a draw) will they play the best moves to make it? Can engines play the best moves having the wrong evaluation of
position?
Guest
answer (Vasik Rajlich):
Yes, and no. The
important thing is how an engine evaluates the positions in the search tree
relative to each other. If the factor being misevaluated persists throughout
the search tree, the misevaluation is not harmful. More commonly, however, when
an engine misevaluates the base position, relative misevaluations inside the
tree will follow. For example, if black’s position is overestimated, then as
white the engine may volunteer to go into a worse position of another type
rather than keep the current imbalance.
The position here (from
VVarkey-Relic) falls more in the first category – Rybka overvalues the queen,
but does so consistently throughout the search tree, and there are few
plausible ways to exchange the queen. So, in this position, the root
misevaluation should not be very harmful. Still, as the game continuation
proves, it may not be completely harmless, either.
_pliki/image011.jpg)
WHITE TO
MOVE
How is black going
to win if white waits - playing Kh2-g2-h3 when black does not attack f2 and
moving bishop when Black does? 80.Nf5?! Looks like white is trying to
exchange pawns. Qd5+ 81.f3 Qe5 85.Bc6 Qf4 86.Ng2 Qh6+ 87.Kg3 Qg5+ 88.Kh3
Qf5+ 89.Kg3 Qg6+ 90.Kh3 Qf6 91.f4
After about 40
moves using “zugzwang” or “luck” black managed to get into this
position:
_pliki/image013.jpg)
WHITE TO MOVE
128.f5 White
could try to play 128.Kf1 but it looks like it is the losing move: 128..
Kf3 129.Nc3+
a)
129… Ke3! 130.Nd1+ Kd2 131.Nf2 Q:f4 132.Kg1 Qg3+ 133.Kf1
b)
129...Kg3? 130.Ne2+ Kg4 131.b6=;
c)
129...K:f4? 130.Bg2 Qh8 131.Ne4 Qa1+ 132.Kf2 Qb2+ 133.Kg1
Here Rybka 1.1 will "never" take the “b” pawn because it will hit
tablebases with draw evaluation.
133...Q:b5 =); ]
128...Qg3+ 129.Kd2 Qd3+ 130.Ke1
K:f5 Somebody from kibitzers posted that this position according
to six stone tablebase is winning for black…
_pliki/image015.jpg)
WHITE TO
MOVE
…perhaps Kibitzer was right, but the game finish after next 50 move as a draw! Engine could not find a forced win with so little time or Kibitzer was wrong and endgame is drawn anyway, or the position is winning but with more that 50 moves...131.Kf2 Ke5 132.Ne3 Kd4 133.Nd5 Qf5+ 134.Ke2 Qh3 135.Kf2 Qg4 136.Nb6 Qf4+ 137.Kg2 Qe5 138.Nd5 Ke4 139.Nf6+ Kd3 140.Ng4 Qf4 141.Kh3 Kc4 142.Kh4 Kb4 143.Bd7 Kc3 144.Kh3 Kb3 145.Kh4 Kb4 146.Be8 Kc5 147.Bc6 Kd4 148.Kh3 Qd6 149.Kh4 Qc5 150.Kg3 Qf8 151.Kg2 Qd6 152.Kf3 Qb8 153.Kg2 Qd8 154.Kg3 Qc7+ 155.Kf3 Qd6 156.Kg2 Qe6 157.Kg3 Qg6 158.Kf4 Qh5 159.Kf3 Qf5+ 160.Kg3 Qh7 161.Nf6 Qg7+ 162.Ng4 Qg8 163.Kf4 Qg6 164.Kg3 Qf7 165.Kh4 Qe7+ 166.Kg3 Qa3+ 167.Kf4 Qf8+ 168.Kg5 Qe7+ 169.Kf4 Qe2 170.Kf5 Qc2+ 171.Kg5 Qc5+ 172.Kf4 Qc4 173.Kg5 Qf1 174.Nf6 Qd3 175.Nd7 Qc3 176.Nf6 Qe3+ 177.Kg6 Qe7 178.Ne4 Qe6+ 179.Kg7 Qe5+ 180.Nf6 Qe7+ ½–½