Today's Featured Match: IvanHoe B47e v Rybka 4.1

[geots]

Again, this version was a bit unknown to me. I am trying, as I said, to run many different versions of IvanHoe against Rybka 4.1. I am not saying that 50 games will give you a foolproof answer, but it will help anyone who is interested in IvanHoe versions, and at least which they would most likely need if they want the strongest. And when the versions are multiplying faster than rabbits, everyone interested needs help.


12/7/2011

1CPU/32bit
128MB hash
Bases for Ivanhoe B47e=NONE
Bases for Rybka 4.1=3,4,5 man Nalimov
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


IvanHoe B47e......+14/-12/=24 -- 26/50
Rybka 4.1............+12/-14/=24 -- 24/50


Again, PGNs available from my email address on this page.



george

[kranium]

Again, this version was a bit unknown to me. I am trying, as I said, to run many different versions of IvanHoe against Rybka 4.1. I am not saying that 50 games will give you a foolproof answer, but it will help anyone who is interested in IvanHoe versions, and at least which they would most likely need if they want the strongest. And when the versions are multiplying faster than rabbits, everyone interested needs help.


12/7/2011

1CPU/32bit
128MB hash
Bases for Ivanhoe B47e=NONE
Bases for Rybka 4.1=3,4,5 man Nalimov
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


IvanHoe B47e......+14/-12/=24 -- 26/50
Rybka 4.1............+12/-14/=24 -- 24/50

Again, PGNs available from my email address on this page.

george

Hey George!
great job...

i don't mean to correct, but there appears to be much confusion in this regard...
the CCRL has even used it as yet another lame excuse 'not' to test...

the situation is relatively simple...
to clarify:
there are 28 'official' ippolit.wikispaces.com releases (versions) of IvanHoe
(all public beta versions intended for testing)
but, there are many hundreds of so-called 'releases'....

i.e. official versions are not 'multiplying faster than rabbits'

anything else (for ex: IvanHoe B47e) is not really a 'version', it's simply an alternate compile of the official source code...
(i.e. individuals trying to add some speed via compiler optimizations, or perhaps changing piece values, etc.)
and these releases are indeed 'multiplying' at an astonishing rate...

i strongly suggest you avoid the confusing quagmire of releases (many of which have been consequently/unintentionally weakened) ,
and save your CPU time and test the original 'pure' official source code,
exactly as released by the ippolit.wikispaces.com authors...

of course this is completely your choice...

Regards-
Norm

[kranium]

These are the 'official' ippolit.wikispaces.com releases of IvanHoe, beginning Jan. 2010
they have been compiled for Windows and Linux (32bit and 64-bit) by KLO
the source code is exactly as released by the engine authors...
all are available for download from his FTP site:
http://chess.cygnitec.com/engine/

IvanHoe999946a
IvanHoe999946e
IvanHoe999946f
IvanHoe999947c
IvanHoe999948i
IvanHoe999949j
IvanHoe999950h
IvanHoe999950k
IvanHoe999950r
IvanHoe999950t
IvanHoe999952a
IvanHoe999953
IvanHoe999954
IvanHoe999955m
IvanHoe999957a
IvanHoe999958p
IvanHoe999963
IvanHoe999964
IvanHoe999965
IvanHoe999966
IvanHoe999967
IvanHoe999968
IvanHoe999970
IvanHoe999973
IvanHoe999976
IvanHoe999981
IvanHoe999984
IvanHoe999987

sometimes, KLO's compiles may not be as fast as one or another individuals 'release',
but they are 'pure', exactly as intended by the authors...
fairly important when assessing engines, IMHO

[kranium]

piece value changes in PP's B47e

UCI_White_Bishops_Scale = 45;
UCI_White_Pawn_Scale = 100;
UCI_White_Knight_Scale = 300;
UCI_White_Light_Scale = 320;
UCI_White_Dark_Scale = 310;
UCI_White_Rook_Scale = 500;
UCI_White_Queen_Scale = 950;

UCI_Black_Bishops_Scale = 45;
UCI_Black_Pawn_Scale = 100;
UCI_Black_Knight_Scale = 300;
UCI_Black_Light_Scale = 310;
UCI_Black_Dark_Scale = 320;
UCI_Black_Rook_Scale = 500;
UCI_Black_Queen_Scale = 950;

i.e. king bishops are more valuable than queen bishops...
and:

QueenMatVal=1000;
RookMatVal=500;
BishopMatVal=365;
KnightMatVal=345;
PawnMatVal=100;
BishopPairMatVal=45;

i.e. the MatVal (overall material value) of each piece has been scaled in the proportions indicated above

this aside, there are also other significant changes which may affect testing results

[geots]

Again, this version was a bit unknown to me. I am trying, as I said, to run many different versions of IvanHoe against Rybka 4.1. I am not saying that 50 games will give you a foolproof answer, but it will help anyone who is interested in IvanHoe versions, and at least which they would most likely need if they want the strongest. And when the versions are multiplying faster than rabbits, everyone interested needs help.


12/7/2011

1CPU/32bit
128MB hash
Bases for Ivanhoe B47e=NONE
Bases for Rybka 4.1=3,4,5 man Nalimov
Ponder_Learning=OFF
Perfect 12.32 book w/12-move limit
40/3 Repeating
Match=50 games


IvanHoe B47e......+14/-12/=24 -- 26/50
Rybka 4.1............+12/-14/=24 -- 24/50

Again, PGNs available from my email address on this page.

george

Hey George!
great job...

i don't mean to correct, but there appears to be much confusion in this regard...
the CCRL has even used it as yet another lame excuse 'not' to test...

the situation is relatively simple...
to clarify:
there are 28 'official' ippolit.wikispaces.com releases (versions) of IvanHoe
(all public beta versions intended for testing)
but, there are many hundreds of so-called 'releases'....

i.e. official versions are not 'multiplying faster than rabbits'

anything else (for ex: IvanHoe B47e) is not really a 'version', it's simply an alternate compile of the official source code...
(i.e. individuals trying to add some speed via compiler optimizations, or perhaps changing piece values, etc.)
and these releases are indeed 'multiplying' at an astonishing rate...

i strongly suggest you avoid the confusing quagmire of releases (many of which have been consequently/unintentionally weakened) ,
and save your CPU time and test the original 'pure' official source code,
exactly as released by the ippolit.wikispaces.com authors...

of course this is completely your choice...

Regards-
Norm

Norm, constructive criticism never bothers me- especially from a friend. I appreciate it. And I agree. It is confusing. Problem is I can't really tell the releases from the ones that changes are made and put back out there. You don't know how much- if you don't mind- I would appreciate it if you would mail me a list of all the "pure and original" releases of IvanHoe. Really only ones necessary, as I have yours, and RobboLito is really academic as it is merging with IvanHoe. Off course if there is an exception- a pure version that has really been improved- I would want its name also.

Be looking for it,

george

[geots]

These are the 'official' ippolit.wikispaces.com releases of IvanHoe, beginning Jan. 2010
they have been compiled for Windows and Linux (32bit and 64-bit) by KLO
the source code is exactly as released by the engine authors...
all are available for download from his FTP site:
http://chess.cygnitec.com/engine/

IvanHoe999946a
IvanHoe999946e
IvanHoe999946f
IvanHoe999947c
IvanHoe999948i
IvanHoe999949j
IvanHoe999950h
IvanHoe999950k
IvanHoe999950r
IvanHoe999950t
IvanHoe999952a
IvanHoe999953
IvanHoe999954
IvanHoe999955m
IvanHoe999957a
IvanHoe999958p
IvanHoe999963
IvanHoe999964
IvanHoe999965
IvanHoe999966
IvanHoe999967
IvanHoe999968
IvanHoe999970
IvanHoe999973
IvanHoe999976
IvanHoe999981
IvanHoe999984
IvanHoe999987

sometimes, KLO's compiles may not be as fast as one or another individuals 'release',
but they are 'pure', exactly as intended by the authors...
fairly important when assessing engines, IMHO

Damn Norm, I wish I had read this thread first. I wasted time on one asking you to please do what you have already done here. But thanks for taking the time to make the list for me.

How well do you know KLO- I know him as well as you can someone whose work you have seen, but never had a conversation with. I can't deal much with his compiles- I don't have the patience. He puts out a bugfix version, changes the letter on the end, but doesn't change it in the exe, and you have to adjust the name after you load it. I thought it was just a few- but he NEVER takes the time to go back and do it right. I could make the name change easier and much cleaner if I could use the parameters. But they are so long, hiding the taskbar at the bottom and changing the screen resolution won't get you close to the bottom where the "save", "load" and "ok" are located.

Could you possibly mention this to him. I has caused me to pass up a few versions because I was tired of fooling with it.


Best,

george

[geots]

piece value changes in PP's B47e

UCI_White_Bishops_Scale = 45;
UCI_White_Pawn_Scale = 100;
UCI_White_Knight_Scale = 300;
UCI_White_Light_Scale = 320;
UCI_White_Dark_Scale = 310;
UCI_White_Rook_Scale = 500;
UCI_White_Queen_Scale = 950;

UCI_Black_Bishops_Scale = 45;
UCI_Black_Pawn_Scale = 100;
UCI_Black_Knight_Scale = 300;
UCI_Black_Light_Scale = 310;
UCI_Black_Dark_Scale = 320;
UCI_Black_Rook_Scale = 500;
UCI_Black_Queen_Scale = 950;

i.e. king bishops are more valuable than queen bishops...
and:

QueenMatVal=1000;
RookMatVal=500;
BishopMatVal=365;
KnightMatVal=345;
PawnMatVal=100;
BishopPairMatVal=45;

i.e. the MatVal (overall material value) of each piece has been scaled in the proportions indicated above

this aside, there are also other significant changes which may affect testing results

Norm, do you know if 47e has been tested any at all with the above changes? But again, I would have to have the version sent to me with the values already changed- as I told you the bottom of the parameters I cannot get close to.


george

[kranium]

Norm, do you know if 47e has been tested any at all with the above changes?
george

yes,
Misha has tested it extensively, results are on chess2u and immortal

[kranium]

I can't deal much with his compiles- I don't have the patience. He puts out a bugfix version, changes the letter on the end, but doesn't change it in the exe, and you have to adjust the name after you load it. I thought it was just a few- but he NEVER takes the time to go back and do it right.

i'm not aware of any issue with version names...
(he's been maintaining that site from the very beginning, even with Robbolito)
there has been a fix or two over the years but he always updates his FTP site with the the updated/fixed download

he's active on open chess...you just need to post your question there and he'll address it
http://www.open-chess.org/viewtopic.php?f=7&t=1686
user: kingliveson

[Frank Quisinsky]

Hi Norman,

at the moment I am testing a bit for the PeterPan versions. In around one week IvanHoe by PeterPan will start in SWCR because I have a longer time no add a new IvanHoe version in SWCR.

Your settings are the topic for myself.
Looks great I think. I believe, not sure, Ahmed used in his older compiled versions of IvanHoe such settings too. Not the same but comparable.

Perhaps I should try out this settings for IvanHoe "PeterPan" in around week. What do you think?

Hint:
In SWCR I used PeterPan versions only. I am sure KLO versions are good too but I can't test both and already I tested three of the PeterPan versions without any problems.

Could be interesting Norman:
PeterPan's compile, I made some changes in standard settings (only for a better ponder = on mode, hash-settings) and from yourself comes the other settings

I saw that Fire is not longer available.
I hope that you will works on Fire again, sources must not be free. Everyone know that the code cleanings comes from yourself, so you can work on it and you can make only newer executables available (without sources of Fire).

I have no interest to lost your work.

Best
Frank

I am not a fan from free sources after all what is happen and in all the years before. I try to gave this information in interview I made with so many of our programmers. Look in the Fruit interview, available for around 6 years, I don't know. The result of strong open sources can be see with Houdini. Programmers should hold it's own work secret. OK, I like the Stockfish project too and can understand what the programmers says to this topic in the interview. But I don't believe that open sources in times today are right. Persons which have interest to try out own ideas in open sources can used other, not so strong sources for his work. They must not have for each good ideas a solution. More interesting is to bring the own brain in computer chess position.