Where can I get databases of comp games

[Kirill Kryukov]

All CCRL games are downloadable also. See here:
http://computerchess.org.uk/ccrl/4040.live/games.html

Or for more options, here:
http://computerchess.org.uk/ccrl/4040/games.html

Brilliant resource Graham and nicely categorised. Can I assume there are no duplicates in there? (I can ask Chessbase to check for me but I don't always trust it's remove duplicates function)

There shouldn't be any duplicates in there. Kirill would be the one to ask.

Sorry to disagree Graham, but "pgn-extract" shows 147 pairs of duplicates in ccrl 40/40.

Here are the files of the duplicates. Each game in the 1st file (of 147 games) has a duplicate game (with different engines) somewhere in the 2nd file (of 147 games).

http://www.mediafire.com/file/k5m0ygdry ... ates-1.pgn

http://www.mediafire.com/file/qymnyj2zv ... ates-2.pgn

-Norm

Hi Norm,

When two games are played by different pairs of players, we don't consider such games as "duplicates".

Best to you,
Kirill

[Spock]

Hi Norm,

When two games are played by different pairs of players, we don't consider such games as "duplicates".

Best to you,
Kirill

An amazing coincidence, yes. A duplicate - no, not in my view

[Norm Pollock]

Hi Norm,

When two games are played by different pairs of players, we don't consider such games as "duplicates".

Best to you,
Kirill

Hi Kiril,

That's the first time I heard that interpretation. It might be logical when it comes to computers, but I would not be satisfied with that interpretation when it comes to human v human games. Just suppose that by unintentional coincidence Kramnik-Topalov played a game that was exactly the same as a Alekhine-Capablanca game. I think almost everyone would consider that to be a "duplicate" game.

My feeling is that once a game has been played, it has sort of been "copyrighted", and any duplication of it thereafter, by computers or humans, is in my opinion, a "duplicate" game. Even if the duplication is non-intentional.

-Norm

[Kirill Kryukov]

Hi Norm,

When two games are played by different pairs of players, we don't consider such games as "duplicates".

Best to you,
Kirill

Hi Kiril,

That's the first time I heard that interpretation. It might be logical when it comes to computers, but I would not be satisfied with that interpretation when it comes to human v human games. Just suppose that by unintentional coincidence Kramnik-Topalov played a game that was exactly the same as a Alekhine-Capablanca game. I think almost everyone would consider that to be a "duplicate" game.

My feeling is that once a game has been played, it has sort of been "copyrighted", and any duplication of it thereafter, by computers or humans, is in my opinion, a "duplicate" game. Even if the duplication is non-intentional.

-Norm

For those who collect and analyze our games it's very important to distinguish between games played by different engines, even if the moves are identical. Average rating, and rating difference of the two players is an important factor that may decide the way you interpret and use the game. For example, when you build a book, you may want to treat games by strong engines and by weak engines differently.

Also for those authors who analyze their engine's weak points by checking the games, they need every game of their engine, not only those games that were never played before.

Also, for the purpose of comparing different engines and computing the ratings it does not make any sense to remove the duplicates where only moves are the same but not the opponents.

Furthermore, anyone who needs to clean the database from duplicates as you define them (for whatever purpose they may have) can do that trivially using their favorite tool.

[Graham Banks]

When two games are played by different pairs of players, we don't consider such games as "duplicates".

Best to you,
Kirill

Quite rightly so. I agree with this interpretation also.
We don't erase games from records of GM games just because two GMs played out exactly the same game as another two GMs did.

[jdart]

I have about 90,000 games, available here:

http://www.arasanchess.org/ccomp.zip

Mostly "standard" time control. Includes some from SSDF, etc. but mostly ICC + FICS + official events + some test tournaments.

--Jon

[Shaun]

Hi Norm,

When two games are played by different pairs of players, we don't consider such games as "duplicates".

Best to you,
Kirill

Hi Kiril,

That's the first time I heard that interpretation. It might be logical when it comes to computers, but I would not be satisfied with that interpretation when it comes to human v human games. Just suppose that by unintentional coincidence Kramnik-Topalov played a game that was exactly the same as a Alekhine-Capablanca game. I think almost everyone would consider that to be a "duplicate" game.

My feeling is that once a game has been played, it has sort of been "copyrighted", and any duplication of it thereafter, by computers or humans, is in my opinion, a "duplicate" game. Even if the duplication is non-intentional.

-Norm

Hi Norm,

thanks for highlighting these games - I had a quick look and most appear to be early draws - I assume this must also be common in GM play?

I agree with Kirill and Graham from our perspective and for statistical purposes I would want to keep these games - however I am not going to get drawn into the definition of 'duplicate'

Shaun

[Norm Pollock]

I'm not bringing up the issue of defining "duplicates" again. But I do have another issue concerning the 294 games mentioned above. Included in them are 56 games between Glaurung and Bright. It seems that the

28 games of Glaurung 2.0.1 64-bit 2CPU - Bright 0.2c have the "same moves" as the 28 games of Glaurung 2.0.1 64-bit 2CPU - Bright 0.2c 2CPU.

Here are the 2 files. Surprisingly the games are in the same order (at least those I checked).

http://www.mediafire.com/file/2a0322znzmm/g2b1.pgn
http://www.mediafire.com/file/ncwdymkhxt4/g2b2.pgn

I'm curious to know how this could have occurred.

[Graham Banks]

I'm not bringing up the issue of defining "duplicates" again. But I do have another issue concerning the 294 games mentioned above. Included in them are 56 games between Glaurung and Bright. It seems that the

28 games of Glaurung 2.0.1 64-bit 2CPU - Bright 0.2c have the "same moves" as the 28 games of Glaurung 2.0.1 64-bit 2CPU - Bright 0.2c 2CPU.

Here are the 2 files. Surprisingly the games are in the same order (at least those I checked).

http://www.mediafire.com/file/2a0322znzmm/g2b1.pgn
http://www.mediafire.com/file/ncwdymkhxt4/g2b2.pgn

I'm curious to know how this could have occurred.

That certainly looks like an error. Thanks for pointing it out.
I'd say that a pgn was submitted originally with the wrong info. When this was pointed out and a new corrected pgn submitted, the original was not removed as it should have been.
This can now be corrected.
If you find anything else Norm, please let us know. We appreciate it.

Cheers,
Graham.

[thoughtful]

I have about 90,000 games, available here:

http://www.arasanchess.org/ccomp.zip

Mostly "standard" time control. Includes some from SSDF, etc. but mostly ICC + FICS + official events + some test tournaments.

--Jon

thanks Jon