Leela sacrificed 6 pawns vs 10% handicapped SF ( 4 GPU vs 88 cores CPU)
Sort:
8 pawns vs pawnless is extremely rare to see!!
bong711 wrote:
Do you have a game collection of Best engine vs engine games?
No.
You can download those games from TCEC and chess.com websites .
Mar 11, 2020
0
#6
I mean Best 50 or Best 100 Games. I see only games collection by engines, by ECO, by tournament.
Mar 11, 2020
0
#8
I wish Top engine can be a guest participant in Super GM Tournaments. Unrated results ofc. I Want to see chess brilliancy against Super GMs.
greatswindler wrote:
Very nonhuman play by both sides
Doesnt matter who play it as there is no magic in chess.
The way Leela played is very advanced technique, h5!!!, with the idea of h6!!, the plan is to break on f6 block .
Another theory Leela used is " enemy pawns are better than own pawns if you can block",See how Bishop and Kt beautifully get outpost in front of enemy pawns. If there is no black pawns , Stockfish can attack lonely unsafe Leela king.
bong711 wrote:
Carlsen and Kasparov would never try this sacrifices.
How about Tal?
Mar 12, 2020
0
#12
SeniorPatzer wrote:
bong711 wrote:
Carlsen and Kasparov would never try this sacrifices.
How about Tal?
Tal's opponents are weak during 1950 to 1980. Against 2600s, Tal have difficulty.
Prometheus_Fuschs wrote:
What do you mean by 10% hadicapped?
90 mins vs 9 mins, in testing. 10% timed handicapped.
bong711 wrote:
Do you have a game collection of Best engine vs engine games?
You can download Leela games and watch them.
I always like the games when Leela beat Stockfish because SF is a monster in tactics and also very deep in seeing from opening to ending in her prinicpal variations. The only ways leela can beat SF is
1. Positional Squeeze
2. material imbalance exchange (e.g minor +2P vs R)
3. Or massive give material give away party like previous game.
In this game, Leela Squeeze Stockfish positionally where SF has to allow Kt fork,
At move 44. Positional squeeze is in queen side is too much that SF get Zugzwang and SF has to allow Nb6 fork.
Mar 12, 2020
0
#16
drmrboss wrote:
Prometheus_Fuschs wrote:
What do you mean by 10% hadicapped?
90 mins vs 9 mins, in testing. 10% timed handicapped.
So this is a way to see brilliant play by one engine. In equal time, t Majority of games will end in draw after 80 to 120 moves.
bong711 wrote:
drmrboss wrote:
Prometheus_Fuschs wrote:
What do you mean by 10% hadicapped?
90 mins vs 9 mins, in testing. 10% timed handicapped.
So this is a way to see brilliant play by one engine. In equal time, t Majority of games will end in draw after 80 to 120 moves.
At that TCEC long time control, if there is no opening it will be 90-95% draw rate among LC0 vs SF.
However, they use some unbalanced opening positions to reduce draw rate. TCEC draw rate in LCO vs SF is still 80-82%.
bong711 wrote:
Do you have a game collection of Best engine vs engine games?
Not necessarily best, but here's a list of TCEC engine games where one side was down a lot of material for some time and still won （replace the spaces with periods):
docs google com/spreadsheets/d/1c1Gw-2ZCVV7dNDsTaDDM7d25kA8L0QAIc_KKuNWwaEo/
