New release Pedone 1.1

[Fabio Gobbato]

I've just released a new version of my engine. It should be about 80ELO stronger than 1.0

Changes:
Now it's compatible with all the GUIs
Rewrite the book probe code
Bugs fixes
Evaluation:
- Column bonus for backward, isolated and doubled pawns
- Use different bonus for the squares near the king under attack
- Reduce the score for bishops of opposite colors endings
Search:
- Generate checks only in the fist Quiescence level
- Use of killers from ply-2
- Reduce killer moves
- Retune reductions
- Use of static evaluation for pruning decisions (before it was used the material score)
Parallel search:
- Limit the number of threads for every splitpoint
- Now every thread copy the position from the splitpoint and not from the master thread

https://sites.google.com/site/pedoneche ... ne_1.1.zip

[Graham Banks]

Good timing! Just about to start 4CPU testing with Pedone shortly.
Thanks Fabio.

[Sylwy]

THANK YOU !

SilvianR

[Frank Quisinsky]

Hi Fabio,

waiting of it, after I am reading that you are very active in programmers selection.

In my FCT1 I have a free place because Hannibal goes a number higher in 2900+ selection. So I need a new engine for my (for me most interesting) FCT1 tourney.

With other words ...
I have to do a small Reorganisation and should start Pedone 1.1 directly.

In around 2 hours Pedone 1.1 will run for the next 4 days!
Results can be found in tourney selection.

Thanks and good luck!

Best
Frank

Programmer is thinking a bit Crackers or crazy the Quisinsky guy.
I think it's truth!
Important only is that the others will not have this information!
Must be a secret between you and me.

[TShackel]

I've just released a new version of my engine. It should be about 80ELO stronger than 1.0

Changes:
Now it's compatible with all the GUIs
Rewrite the book probe code
Bugs fixes
Evaluation:
- Column bonus for backward, isolated and doubled pawns
- Use different bonus for the squares near the king under attack
- Reduce the score for bishops of opposite colors endings
Search:
- Generate checks only in the fist Quiescence level
- Use of killers from ply-2
- Reduce killer moves
- Retune reductions
- Use of static evaluation for pruning decisions (before it was used the material score)
Parallel search:
- Limit the number of threads for every splitpoint
- Now every thread copy the position from the splitpoint and not from the master thread

https://sites.google.com/site/pedoneche ... ne_1.1.zip

Hi,

How many elo is it on rating lists? Is it a top engine, or is it kind of mid-level and on its way up?

Sincerely,

Tim.

[Frank Quisinsky]

Hi Tim,

you should follow the results ...
CEGT tested 1.0 ... after I read at last time around 2.620
+ 80 for the new version ...

Around 2.700 Elo.

For my testing in the for me most important Group of engines ... it's great to have this engine in my list.

Best
Frank

[Graham Banks]

Hi,

How many elo is it on rating lists? Is it a top engine, or is it kind of mid-level and on its way up?

Sincerely,

Tim.

Code: Select all

CCRL 40/40 Rating List - Custom engine selection
570918 games played by 1598 programs, run by 19 testers
Ponder off, General books (up to 12 moves), 3-4-5 piece EGTB
Time control: Equivalent to 40 moves in 40 minutes on Athlon 64 X2 4600+ (2.4 GHz)
Computed on March 21, 2015 with Bayeselo based on 570'918 games
Tested by CCRL team, 2005-2015, http://computerchess.org.uk/ccrl/4040/

Rank                Engine                 Elo   +    -   Score  AvOp  Games
1 Pedone 1.0 64-bit                    2735  +29  -28  53.8%  -28.5   406
  Pedone 0.5 64-bit                    2620  +31  -31  46.9%  +25.4   381
  Pedone 0.4 64-bit                    2420  +31  -31  53.0%  -21.8   365
  Pedone 0.3 64-bit                    2302  +36  -36  48.0%  +13.3   255

[Lyudmil Tsvetkov]

Hi Tim,

you should follow the results ...
CEGT tested 1.0 ... after I read at last time around 2.620
+ 80 for the new version ...

Around 2.700 Elo.

For my testing in the for me most important Group of engines ... it's great to have this engine in my list.

Best
Frank

So, a 10 days break in general, 1 month delay for Pawel, and express service for Fabio.

Seems like very fair testing conditions.

[Frank Quisinsky]

Hi Lyudmil,

can't test Rodent at the moment! Nothing to do with Pawel or Rodent.

Because I have only 19 participants after Hannibal switched the group. The reason I test Pedone directly and others can follow. Same I do with Atlas after Texel switched the group. I tested Atlas directly.

At first Pedone = test field completly
At second Rodent

Documented in my ToDo calendar!
http://www.amateurschach.de/main/_todo.htm
http://www.amateurschach.de/main/_todo-history.htm

And to my 10-days break ...
I am my own boss and creating a rating list = producing confusion! Believe me, all the years I am the master of desaster in producing confusion. So, my 10-days break is delayed. More interesting = producing confusion!

It seems you follow what I do.
If you really follow ... maybe you are the first one which understand my test system. If so ... congratulation!

Best
Frank

PS: A bit joking ...
Each new engine I can added will get priority. Reason is easy ... I must know the playing style directly. Personal interest should be in forground in all things we do.

[Frank Quisinsky]

Hi Graham,

1.300 games for 4 versions isn't not enough. OK, better as to have nothing. I prefer CEGT with more games and better to compare (Shredder = 2.800 Elo). CCRL Elo goes an own way and is clearly to high. Not easy to compare CEGT results with the results from others.

In the time I started with SWCR I made an interview with GM George Mayer (2.675 Elo at this time, number second in Germany at this time). George plays from time to time vs. Computer chess engines. He gave the information with 40 and 10 ... 2.800 Elo for Shredder is realistic. Same opinion other GMs have and I am starting with 2.800 Elo for Shredder. Today I used other engines as Shredder because Shredder isn't consistancy enough because stronger in endgames and not good for using as main engine in calculation Elo for others. Furthermore Shredder have big holes in tactic, lost a lot of games very fast. Hiarcs is better ... more consistancy (playing phases).

Today I set Hiarcs to 2.825!
Around the same as Shredder to 2.800!

We discuss about it TalkChess.
IPON do it, CEGT do it ... CCRL goes an other way.

Again CCRL Elo to compare with others is a bit complicated. Furthermore, for a lot of engines, not enough games ... calculating for others will be more critical.

Nothing against your work ...
I prefer not my own work, more CEGT and here I am looking if I test new engines. Furthermore, CCRL are testing to many clones. To many of the same engines are also bad in calculation Elo of all the others.

My opinion only ... should be not a declaration of war.

Best
Frank