Four wild handicaps for FM level player.

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

lkaufman
Posts: 6020
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Four wild handicaps for FM level player.

Post by lkaufman »

Laskos wrote:
lkaufman wrote:
Instead of removing the two White pawns, if we remove the b1 knight we get a position that Komodo evaluates as winning or nearly winning for Black. However I think that most grandmasters would choose to play White given the choice of color. We might learn something from playing this position out.
Uri's position (two pawns off) at five times longer time control (5'+3'') is favoring white even more, although engines think it's pretty balanced:

Komodo self-games:

Code: Select all

Games        :     30 (finished)

White Wins   :     23 (76.7 %)
Black Wins   :      3 (10.0 %)
Draws        :      4 (13.3 %)
Unfinished   :      0

White Perf.  : 83.3 %
Black Perf.  : 16.7 %
With knight b1 off, engines (Komodo and SF) evaluate it correctly as almost won for black, in 50 games of Komodo at 1'+0.6'':

Code: Select all

Games        :     50 (finished)

White Wins   :      3 ( 6.0 %)
Black Wins   :     42 (84.0 %)
Draws        :      5 (10.0 %)
Unfinished   :      0

White Perf.  : 11.0 %
Black Perf.  : 89.0 %
Humans, on the other hand, favor white.
Since taking the knight off leaves White losing badly despite king on e5, I tried merely giving the exchange (a1 for b8), when White is only missing a piece that won't help him attack while Black is missing a defender. Many people have commented here that in the opening the Exchange isn't worth so much, but despite this the engine evaluates the position as only very slightly better for White, and when I tried playing it out thousands of times using Rybka 4 Monte Carlo feature at 5 ply (which is really 8 ply), White scored only 51.6%. In this initial situation I think any GM would give big money odds to play White and would want to resign as Black. This disparity is so huge that there must be something to learn from it. If nothing else, it shows that the Exchange handicap is a pretty big one.
Komodo rules!
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: Four wild handicaps for FM level player.

Post by Laskos »

lkaufman wrote:
Since taking the knight off leaves White losing badly despite king on e5, I tried merely giving the exchange (a1 for b8), when White is only missing a piece that won't help him attack while Black is missing a defender. Many people have commented here that in the opening the Exchange isn't worth so much, but despite this the engine evaluates the position as only very slightly better for White, and when I tried playing it out thousands of times using Rybka 4 Monte Carlo feature at 5 ply (which is really 8 ply), White scored only 51.6%. In this initial situation I think any GM would give big money odds to play White and would want to resign as Black. This disparity is so huge that there must be something to learn from it. If nothing else, it shows that the Exchange handicap is a pretty big one.
Interesting. I am not a big fan of Monte Carlo implementation in Fritz, a lot of clearly sub-optimal moves are weeded too slowly to lower priority, some sort of faster minimaxing adaptation of frequencies must be implemented. Anyway, I can confirm in 100 Komodo self-games at 1 min + 0.6s that the position is very balanced and the engines show correct evaluation, unlike GMs:

Code: Select all

Games        :    100 (finished)

White Wins   :     28 (28.0 %)
Black Wins   :     25 (25.0 %)
Draws        :     47 (47.0 %)
Unfinished   :      0

White Perf.  : 51.5 %
Black Perf.  : 48.5 %
I tried murky "awkward kings" positions where both engines and humans get it wrong. For example:
[d]r1bq1b1r/ppp1kppp/2n2n2/3pp3/8/8/PPPPPPPP/RNBQRBNK w - -

Komodo and Stockfish see it as balanced, with Komodo very slightly positive, Stockfish very slightly negative. Humans would most probably favor black. But the result of Komodo self-games at 1 min + 0.6s is that white is significantly bettter:

Code: Select all

Games        :    100 (finished)

White Wins   :     30 (30.0 %)
Black Wins   :     12 (12.0 %)
Draws        :     58 (58.0 %)
Unfinished   :      0

White Perf.  : 59.0 %
Black Perf.  : 41.0 %
lkaufman
Posts: 6020
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Four wild handicaps for FM level player.

Post by lkaufman »

Laskos wrote:
lkaufman wrote:
Since taking the knight off leaves White losing badly despite king on e5, I tried merely giving the exchange (a1 for b8), when White is only missing a piece that won't help him attack while Black is missing a defender. Many people have commented here that in the opening the Exchange isn't worth so much, but despite this the engine evaluates the position as only very slightly better for White, and when I tried playing it out thousands of times using Rybka 4 Monte Carlo feature at 5 ply (which is really 8 ply), White scored only 51.6%. In this initial situation I think any GM would give big money odds to play White and would want to resign as Black. This disparity is so huge that there must be something to learn from it. If nothing else, it shows that the Exchange handicap is a pretty big one.
Interesting. I am not a big fan of Monte Carlo implementation in Fritz, a lot of clearly sub-optimal moves are weeded too slowly to lower priority, some sort of faster minimaxing adaptation of frequencies must be implemented. Anyway, I can confirm in 100 Komodo self-games at 1 min + 0.6s that the position is very balanced and the engines show correct evaluation, unlike GMs:

Code: Select all

Games        :    100 (finished)

White Wins   :     28 (28.0 %)
Black Wins   :     25 (25.0 %)
Draws        :     47 (47.0 %)
Unfinished   :      0

White Perf.  : 51.5 %
Black Perf.  : 48.5 %
I tried murky "awkward kings" positions where both engines and humans get it wrong. For example:
[d]r1bq1b1r/ppp1kppp/2n2n2/3pp3/8/8/PPPPPPPP/RNBQRBNK w - -

Komodo and Stockfish see it as balanced, with Komodo very slightly positive, Stockfish very slightly negative. Humans would most probably favor black. But the result of Komodo self-games at 1 min + 0.6s is that white is significantly bettter:

Code: Select all

Games        :    100 (finished)

White Wins   :     30 (30.0 %)
Black Wins   :     12 (12.0 %)
Draws        :     58 (58.0 %)
Unfinished   :      0

White Perf.  : 59.0 %
Black Perf.  : 41.0 %
Black needs to make five moves to achieve a normal, harmonius position (...g6, Bg7,Re8,Kf8,Kg8). If White makes five "good" moves during that time, development will be equal, and White will be on move, so should be slightly better. But your result is more than slightly better. Perhaps Black is starting fighting without patching up his position first.
Komodo rules!