Looking for EPD's of the first 10 moves of games

[Rebel]

For the creation of a new ultra short Polyglot book (only the first 10 moves ) I extended the existing EPD database of 14 million unique opening position with 2.5 million, now 16.5 million. Before analyzing those with SF12 I would like to have 20 million.

Anyone having large EPD opening sets, or large PGN's with the more unusual openings?

[Dann Corbit]

For the creation of a new ultra short Polyglot book (only the first 10 moves ) I extended the existing EPD database of 14 million unique opening position with 2.5 million, now 16.5 million. Before analyzing those with SF12 I would like to have 20 million.

Anyone having large EPD opening sets, or large PGN's with the more unusual openings?

I don't have a ply count with my EPD positions.
But I have a lot of games.
I guess if you used CCRL and CEGT and TCEC and OMCorr and TWIC that would give you plenty.
CCRL and CEGT is about 2.5 million games.
I have about a million correspondence games.
Twic is smaller, but there are a lot of interesting openings in there.

If you just want volume there is Lichess and Playchess
Lichess has 1,620,424,788 standard rated games now.

[Rebel]

For the creation of a new ultra short Polyglot book (only the first 10 moves ) I extended the existing EPD database of 14 million unique opening position with 2.5 million, now 16.5 million. Before analyzing those with SF12 I would like to have 20 million.

Anyone having large EPD opening sets, or large PGN's with the more unusual openings?

I don't have a ply count with my EPD positions.
But I have a lot of games.
I guess if you used CCRL and CEGT and TCEC and OMCorr and TWIC that would give you plenty.
CCRL and CEGT is about 2.5 million games.
I have about a million correspondence games.
Twic is smaller, but there are a lot of interesting openings in there.

If you just want volume there is Lichess and Playchess
Lichess has 1,620,424,788 standard rated games now.

I use all of thes but haven't turned to Lichess yet, will be fun. But I am very much interested in your one million correspondence games because correspondence players often try unusable openings.

[pohl4711]

For the creation of a new ultra short Polyglot book (only the first 10 moves ) I extended the existing EPD database of 14 million unique opening position with 2.5 million, now 16.5 million. Before analyzing those with SF12 I would like to have 20 million.

Anyone having large EPD opening sets, or large PGN's with the more unusual openings?

I don't have a ply count with my EPD positions.
But I have a lot of games.
I guess if you used CCRL and CEGT and TCEC and OMCorr and TWIC that would give you plenty.
CCRL and CEGT is about 2.5 million games.
I have about a million correspondence games.
Twic is smaller, but there are a lot of interesting openings in there.

If you just want volume there is Lichess and Playchess
Lichess has 1,620,424,788 standard rated games now.

I use all of thes but haven't turned to Lichess yet, will be fun. But I am very much interested in your one million correspondence games because correspondence players often try unusable openings.

I saw, that you linked my website on your project-site (a huge honor for me - you are one of my computerchess heroes!), but "only" the Download & Links section. There I missed (until now) to offer a download to the AB-testrun games played from the beginning of 2020 until now. That download was only on my main-site, so I am afraid you missed it? 485000 games were played in 2020 until today... A huge number of data.
Here the direct download link:
https://www.sp-cc.de/files/archive_hert_2020.zip

And do you know Thomas Zipproth's awesome LittleBlitzerGUI tools? Even though you dont use that GUI, in that downlad is a very cool tool: pgn2epd
The tools awaits a pgn-file and it builds a epd-file with the FEN codes of all endpositions of all games stored in the pgn games-file. So, if you cut all games in a pgn-file after move 10 and use that tool, it will build the epd of all endpositions after move 10 very fast and automatically.
https://www.sp-cc.de/files/zipproth_lbg_tools.zip

[Rebel]

Added 12 million positions from a Lichess database of 2200+ elo, so I have now 28 million unique positions from the first 10 moves, twice as much as the first USB database of 2 years ago.

The 14 million USB book analyzed with Stockfish 10 at 1000ms has 250+ downloads and 22 likes, not much, but enough for an update.

Making a new 28 million USB book at 1000ms with Stockfish 12 (estimated elo 3200) would certainly a good improvement and would take 8 days to complete.

But I wonder if there is interest to make it a commumity project splitting the job in parts of (say) 1 million positions at a higher time control or (even better) fixed depth.

[Rebel]

For the creation of a new ultra short Polyglot book (only the first 10 moves ) I extended the existing EPD database of 14 million unique opening position with 2.5 million, now 16.5 million. Before analyzing those with SF12 I would like to have 20 million.

Anyone having large EPD opening sets, or large PGN's with the more unusual openings?

I don't have a ply count with my EPD positions.
But I have a lot of games.
I guess if you used CCRL and CEGT and TCEC and OMCorr and TWIC that would give you plenty.
CCRL and CEGT is about 2.5 million games.
I have about a million correspondence games.
Twic is smaller, but there are a lot of interesting openings in there.

If you just want volume there is Lichess and Playchess
Lichess has 1,620,424,788 standard rated games now.

I use all of thes but haven't turned to Lichess yet, will be fun. But I am very much interested in your one million correspondence games because correspondence players often try unusable openings.

I saw, that you linked my website on your project-site (a huge honor for me - you are one of my computerchess heroes!), but "only" the Download & Links section. There I missed (until now) to offer a download to the AB-testrun games played from the beginning of 2020 until now. That download was only on my main-site, so I am afraid you missed it? 485000 games were played in 2020 until today... A huge number of data.
Here the direct download link:
https://www.sp-cc.de/files/archive_hert_2020.zip

And do you know Thomas Zipproth's awesome LittleBlitzerGUI tools? Even though you dont use that GUI, in that downlad is a very cool tool: pgn2epd
The tools awaits a pgn-file and it builds a epd-file with the FEN codes of all endpositions of all games stored in the pgn games-file. So, if you cut all games in a pgn-file after move 10 and use that tool, it will build the epd of all endpositions after move 10 very fast and automatically.
https://www.sp-cc.de/files/zipproth_lbg_tools.zip

Thanks Stefan, I will look into those.

[Rebel]

The HERT pgn added 154.983 unique positions to the database, good to have these type of openings.

[Dann Corbit]

But I am very much interested in your one million correspondence games because correspondence players often try unusable openings.

See:
http://talkchess.com/forum3/viewtopic.php?f=2&t=75739