Not logged inCSS-Forum
Forum CSS-Online Help Search Login
CSS-Shop Impressum Datenschutz
Up Topic Hauptforen / CSS-Forum / Critter 1.4 against Komodo 4 (locked)
- - By Orlando de la Vega Date 2012-01-04 11:13
Critter 1.4 against Komodo 4

PlaceEngineAuthorCountryPointsSingle results
1Critter 1.4Richard Vida62.0/100½½11½0½½½½-½1½1½1½101-½1½½101½½0-11½½½0½1½½-½1½½11½½½½-
½½11110½11-01½1010½11-1101½½1111-½00½1½½11½-½00110½10½
2Komodo 4D. Dailey, L. Kaufman38.0/100½½00½1½½½½-½0½0½0½010-½0½½010½½1-00½½½1½0½½-½0½½00½½½½-
½½00001½00-10½0101½00-0010½½0000-½11½0½½00½-½11001½01½





100 of 100 matches played.
Critter 1.4 v/s Komodo 4 (100: + 40,= 44,- 16)

Start of tournament: ......... 2012.01.03, 13:57:26
Town/ Country: .............. Dax, France
Level: ............................. Blitz 40/1' repeated
Hardware: ...................... AMD Phenom(TM) 9750 Quad-Core
Engines: ......................... Critter 1.4 / Komodo 4 both 64-bit using 1 thread
Hash: ............................. 128 MB for each engine
GUI: ............................... Shredder Classic 3
OS: ................................ Windows 7 Ultimate
Ponder OFF
Nalimov TB ..................... 3+4+5, Gaviota for Critter
Book: ............................. Shredder book, stopped using after 5 moves.
Openings: ....................... 50 different openings randomly chosen by GUI, large mode.



Best regards,
Orlando
Parent - - By Thorsten Eckhardt Date 2012-01-04 16:09
Hi Orlando,

is this a test with 40 moves in 1 minute?
Parent - - By Orlando de la Vega Date 2012-01-04 16:59
40 moves in 1 minute, repeated ! 
Parent - - By Thorsten Eckhardt Date 2012-01-04 22:51
Oh, Orlando... I think this is to short...
Why do you not play with 5, 10 or 20 minutes per game?
40 moves in 1 minute is like a motorrace over 50 meters...
Sorry... 
Parent - - By Orlando de la Vega Date 2012-01-05 11:28
Hi Benno and Thorsten, thanks for your comments

Benno, my tournaments have an average duration of 5 hours.

Thorsten, I agree with you. And besides, personally, I prefer the tournaments with long time control and multi-core use (such as those of Sedat Canbaz or Pavel Háse). But there are two reasons why I make these little tournaments : first because there are many private tournaments, accessible on the Net. And if everybody does the same conditions, it would become boring. And secondly because it allows a first and quick overview of the strength of an engine, which is impossible in conditions of very long duration, where you have to wait several days.

Best,
Orlando
Parent - - By Michael Scheidl Date 2012-01-05 11:51
That makes sense and I think that 100 games under good technical circumstances, are certainly not meaningless. Match configuration details are partially matter of taste and always debatable. For example, I wonder if you couldn't switch to ponder on (except all other 3 cores of the quad are busy with something else), which would improve the game quality without requiring more total time.
Parent - By Michael Scheidl Date 2012-01-05 11:52
(oops)
Parent - By Orlando de la Vega Date 2012-01-05 12:38
Yes, Ponder On is not a bad idea. I'll think about it in future tournament.

Best,
Orlando
Parent - By jonpall Date 2012-01-05 12:43
Komodo`s strength is higher than your match indicates.
GM Larry Kaufman who co-authored Komodo wrote that the Ippo-related programs (he included also Critter as it uses almost all Ippo ideas) are better at bullet chess but scale worse than the rest.
He also stated "So Komodo 4 is 77 Elo behind Houdini 2 at blitz, but only 14 elo behind at 40/20 after almost a thousand games! So we gain 63 elo relative to Houdini with five times as much time. Way beyond sample error."
http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20All%20Versions/rangliste.html

Most of us are interested in thorough long-term analysis wherein Komodo excels.

BTW National 12 is running an engine match (Komodo 4-Houdini 2 40/40. comp at 4 Gh/z ) and keeps posting interim results on rybkaforum, +18=38-18 so far
both 1 core,no tbs,HS MasterBook 4,10 moves,hash 1024.SSE version of Komodo,64 bit.ponder off
Parent - By Benno Hartwig Date 2012-01-05 08:52
Such short time controlls may be accectable if you need several hundreds or thousands of games.
And to me they seem to be usefull if you want to detect the effect of little source- or parameter-changes.
But you cannot get reasonable results with such short times when comparing two different engines.
Since ippolit I knew it very clear, and some other engines might work similar.
(Did anybody have a look on Rybka4.1 - Critter 1.2? They should have a rather similar strength. But look at such very short time controlls...)

Did you really have only 2 hours(!) to do this test?

Benno
Parent - By jonpall Date 2012-01-05 12:45
Komodo`s strength is higher than your match indicates.
GM Larry Kaufman who co-authored Komodo wrote that the Ippo-related programs (he included also Critter as it uses almost all Ippo ideas) are better at bullet chess but scale worse than the rest.
He also stated "So Komodo 4 is 77 Elo behind Houdini 2 at blitz, but only 14 elo behind at 40/20 after almost a thousand games! So we gain 63 elo relative to Houdini with five times as much time. Way beyond sample error."
http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20All%20Versions/rangliste.html

Most of us are interested in thorough long-term analysis wherein Komodo excels.

BTW National 12 is running an engine match (Komodo 4-Houdini 2 40/40. comp at 4 Gh/z ) and keeps posting interim results on rybkaforum, +18=38-18 so far
both 1 core,no tbs,HS MasterBook 4,10 moves,hash 1024.SSE version of Komodo,64 bit.ponder off
Up Topic Hauptforen / CSS-Forum / Critter 1.4 against Komodo 4 (locked)

Powered by mwForum 2.29.3 © 1999-2014 Markus Wichitill