Murka 3.0 released

[Adam Hair]

Fantastic job Jesus!

[Werner]

Fantastic job Jesus!

that´s true!
thanks a lot!
PS
1035 Murka 2.0 x64 2276
I made some games against 2300 program:
1 Murka 3 x64 X 33.0 33.0 / 33 0.00 100.00%
...I did not believe the 2900 - now I must.

regards
Werner

[IWB]

Hello Werner,

Unfortunately the engine doesn't ponder and is only giving the first move (which is not wrong but ugly).
So no real chess possible for the IPON

Bye
Ingo

[Graham Banks]

Fantastic job Jesus!

that´s true!
thanks a lot!
PS
1035 Murka 2.0 x64 2276
I made some games against 2300 program:
1 Murka 3 x64 X 33.0 33.0 / 33 0.00 100.00%
...I did not believe the 2900 - now I must.

regards
Werner

Hi Werner,

am I right in understanding you saying that Murka 3 is around the 2900 ELO mark in CEGT ratings?

Graham.

[Ajedrecista]

Hello Graham:

Fantastic job Jesus!

that´s true!
thanks a lot!
PS
1035 Murka 2.0 x64 2276
I made some games against 2300 program:
1 Murka 3 x64 X 33.0 33.0 / 33 0.00 100.00%
...I did not believe the 2900 - now I must.

regards
Werner

Hi Werner,

am I right in understanding you saying that Murka 3 is around the 2900 ELO mark in CEGT ratings?

Graham.

The aim of the author (Igor Korshunov, user WildCat at ImmortalChess Forum) was reach 2900 CCRL level (he said that back in November 2012, link here).

Murka 3 performs very good (please see post #1094 and following ones) according to the author and other members of ImmortalChess. User Rom77 noted Werner's request of tranlation and user Neurocore (Nikita Kurgin, author of engine Eia) translated the file but I did not have access to it (in fact, I have just noted that ImmortalChess members were aware of Werner's request of translation; anyway, I can not have access to attached files in ImmortalChess). It looks that user Alexander Garde is waiting for the activation of his account in TalkChess for posting the good translation of the file.

Writing a little about the author: he is not a complete newcomer. He challenged Martin Fierz (author of checkers engine Cake and CheckerBoard GUI) to a short match of suicidal checkers (engine vs. engine). Kallisto won against Suicidal Cake convincingly (report). I can not claim anything about the originality of Murka, but one thing is sure: there are not too many open source suicidal checkers engines, if any! So I am waiting for results. Good testing!

Regards from Spain.

Ajedrecista.

[Werner]

Hi Graham,
I have just started some games against Gull II w32 1 CPU, which is around 2900.
My posting meant: it must be a few 100 Points more than 2300 - perhaps near 2900.

At the Moment (20 games) the engine is 180 Points behind Gull.
Such a result (2700-2800) would be +500 and more above Murka 2.0.

and yes - no pondering!
Werner

[Graham Banks]

Hi Graham,
I have just started some games against Gull II w32 1 CPU, which is around 2900.
My posting meant: it must be a few 100 Points more than 2300 - perhaps near 2900.

At the Moment (20 games) the engine is 180 Points behind Gull.
Such a result (2700-2800) would be +500 and more above Murka 2.0.

and yes - no pondering!
Werner

Thanks Werner.

[Dariusz Orzechowski]

This is the mentioned translation made by user Neurocore on immortal forum:

Code: Select all

Murka 3


COMMAND LINE PARAMETERS
-----------------------

-h N (size of transposition table in Mb, default 128)
-l (log on)


PROTOCOLS
---------

- UCI
- XBoard


STRUCTURES
----------

- magic bitboards


MOVE GENERATORS
---------------

- all moves
- captures and promotions to queen
- evasions from checks
- quiet
- checks (has mask for exception captures that has already been generated)


SEARCH
------

- fail soft PVS
- sorting: hash, good captures (SEE>=0) by MVV/LVA, 2 killers, history, bad captures
- transposition table (strategy of simple replacement)
- null move
- checks extension, good recaptures extension in first line
- position repetition control
- incremental move generator
- mate distance pruning
- special case of quiescence search when check is
- pruning on small depths (in too good and too bad positions)
- reduction of last moves with verification
- searching first move on less depth in first line (IID) 


TIME MANAGEMENT
---------------

- evident move
- if there was no a significant evaluation drop at the last iteration
- if the best move wasn't changed
- if the best move wasn't changed on last iteration


QS
--

- consider good captures and queen promotions only
- checks are in inlet QS
- alpha-beta pruning
- MVV/LVA sorting


EVALUATION
----------

- material with adjustment
- PST
- recognizing draws
- rooks on open/semiopen lines
- mobility
- attack to king
- pawn shield
- castling rights
- open/blocked pawns
- isolated pawns
- backward pawns
- passers
- no friend piece on passer stop square
- no enemy piece on passer stop square
- no attacked by enemy only squares on passer stop square
- distance between passer and kings
- candidate passer
- trapped bishop
- closed bishop
- closed rook
- bishops of opposite colors
- tempo


CONTROLS
--------

- depth
- time to move
- time for several moves
- time to game
- incremental
- infinite analysis
- positions by move (for tuning)

[Werner]

Ah, fine,
thanks too Dariusz!

I will test the engine now - results then in our Forum or here in about 1 Week.

[Laskos]

Hi Graham,
I have just started some games against Gull II w32 1 CPU, which is around 2900.
My posting meant: it must be a few 100 Points more than 2300 - perhaps near 2900.

At the Moment (20 games) the engine is 180 Points behind Gull.
Such a result (2700-2800) would be +500 and more above Murka 2.0.

and yes - no pondering!
Werner

Seems an original engine stronger than Rybka 1.0.