An Engine-Engine Test-Preparing For Some Serious Testing!

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
geots
Posts: 4790
Joined: Fri Mar 10, 2006 11:42 pm

An Engine-Engine Test-Preparing For Some Serious Testing!

Post by geots » Sun Aug 25, 2013 2:43 am

Please no one get their drawers in a wad over this match. Accurate it will be, but not enough games to get excited about. Of course with a rating list I am considering- the control would have to be much faster to get anything done. But this control would be for strictly the playoff matches (best of 12- possibly all 4-Core but most likely 6-Core) and the championship match (best of 24, ala Fischer-Spaasky in '72), being all games 60'+15". And I want to check closely and get an average per game time length.

So no wins prove much of anything- and no winner takes over anything from the loser. And this test will be "First to 6 Games" between Houdini 3 and Stockfish 4.Here are other controls:





Alienware AURORA_R4
Intel i7 w/6 True Cores

Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
60'+15"
Games= "1st To 6"





Tho this match is not the "be all and end all" of everything- I will still keep you posted along the way.


(I finally got the new 6-Core intel i7 Alienware loaded and ready to go.) I am tired.

gts

User avatar
geots
Posts: 4790
Joined: Fri Mar 10, 2006 11:42 pm

Re: An Engine-Engine Test-Preparing For Some Serious Testing

Post by geots » Tue Aug 27, 2013 6:02 am

geots wrote:Please no one get their drawers in a wad over this match. Accurate it will be, but not enough games to get excited about. Of course with a rating list I am considering- the control would have to be much faster to get anything done. But this control would be for strictly the playoff matches (best of 12- possibly all 4-Core but most likely 6-Core) and the championship match (best of 24, ala Fischer-Spaasky in '72), being all games 60'+15". And I want to check closely and get an average per game time length.

So no wins prove much of anything- and no winner takes over anything from the loser. And this test will be "First to 6 Games" between Houdini 3 and Stockfish 4.Here are other controls:





Alienware AURORA_R4
Intel i7 w/6 True Cores

Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
60'+15"
Games= "1st To 6"





Tho this match is not the "be all and end all" of everything- I will still keep you posted along the way.


(I finally got the new 6-Core intel i7 Alienware loaded and ready to go.) I am tired.

gts





Well, this match is being played as "First To 6 Games". And it looked as if Houdini 3 was in the catbird seat. With 9 games played Houdini had a 5.5 to 3.5 lead.

But Stockfish 4 (official release) has come back and won game 10. In the mid-50 moves of Game 11 Stockfish 4 has a +7.25 lead over Houdini- and since Houdini agrees- you can most likely turn the lights out on this one.

So now we have 1 game remaining with Houdini getting white. A draw makes a drawn match- either one wins game 12- they win the match 6.5 to 5.5.

I am quite sure it will be after sleep and later in the day before I see any results. So far I have been quite pleased with 60'+15"- we shall see.






Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
60'+15"
Match= "First To 6 Games"

User avatar
geots
Posts: 4790
Joined: Fri Mar 10, 2006 11:42 pm

Houdini 3 v Stockfish 4 Match Results Are In And Ready!

Post by geots » Tue Aug 27, 2013 10:28 pm

I hope no one missed the intro on this match. If so, read the above 2 threads before asking a question. Quite possibly the answer to your question lies there. This match was billed as "First To 6 Games"- with a rather longish time control- 60'+15". This was my first 60+15, and I may go with it as the "top match level controls" for longer games. We shall see soon. I used Stockfish 4 because that is the "official" version that Marco requested. I mention this in passing only because I don't want people lining up to ask me why I did not use one of the dev. versions that has come out since version 4.

Certainly no big deal- but kind of humorous that Fritz 11 gui still refers to this control as "blitz". Whatever floats your boat.

I mentioned that Stockfish was behind 5.5 to 3.5 after 9 games. He was able to run off 2 consecutive wins and it was all knotted up at 5.5 each with one game to play. Likely a bit much to expect an engine to run the table in the last 3 games and win the match. ESPECIALLY when the other engine happens to be Houdini 3. But Stockfish damn near did it.

Again, look at the scores below and you can add 12 more to the couple hundred "long time control" games that Stockfish dev. versions and official releases have played ag. Houdini 3 and never allowed Houdini 3 to put together back to back wins. Not even ONCE! Quite telling to me. Also, the first match win Houdini 3 has over the dev. versions.







Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
60'+15"
Match = "1st To 6"



Code: Select all

                         1  2   3   4   5   6   7   8   9   0   1   2  
 Houdini 3 x64           0 1/2 1/2  1  1/2 1/2  1  1/2  1   0   0   1 = 6.5/12   
 Stockfish 4 64 SSE4.2   1 1/2 1/2  0  1/2 1/2  0  1/2  0   1   1   0 = 5.5/12


More to follow- sooner than later-

Post Reply