UHO 2024 released

[Krzysztof Grzelak]

I have uploaded it to git lab; let me know when you download it, so I will delete it.
https://gitlab.com/beejaganita/chess-data
But I suggest do learn to use pgn-extract, it is very useful and easy to use.

Thank you, can you share the zipped file without any login. Thanks in advance.

[chesskobra]

Sorry, I just noticed that it cannot be accessed without login. I have now made it public. Please try again.

[chesskobra]

A question for Stefan. If the end position of a line is analysed, and, for example, has an eval of 040, it may not necessarily be a balanced opening line if both players have made errors and have arrived at a balanced position. How is this problem resolved? Ideally, if the end position is to have an eval of 040, then all positions on the line should have an eval of at most 040.

[pohl4711]

A question for Stefan. If the end position of a line is analysed, and, for example, has an eval of 040, it may not necessarily be a balanced opening line if both players have made errors and have arrived at a balanced position. How is this problem resolved? Ideally, if the end position is to have an eval of 040, then all positions on the line should have an eval of at most 040.

This problem is not solved. And cannot be solved with a reasonable effort of evaluating, sorry. Only the endpositions of the UHO-opening lines were evaluated, because from there, the engines start playing. It is possible, that white and black both played a bad move, before reaching the endposition, so that the eval of the endposition is "normal", but there are bad moves in the opening line, leading to this endposition.
But this should be a really rare case, because all opening-lines were played by players with 2200 Elo or more and the UHO opening lines are very short (6 moves and 8 moves). And mention, that this has no impact on the results, because the engines start playing at the endposition of the opening-lines.
And, because all UHO-endpositions give white a measureable advantage (that is the concept!), in all opening-lines, the human player with the black pieces has played a (at least) weak move, otherwise white would not have an advantage after 6/8 moves... So, without weak human moves in the opening-lines, building UHO openings would be impossible.

[chesskobra]

I agree that for engine testing after opening phase it does not matter. But here is some investigation I did.

From Stefan's UHO database of 8 move games, I filtered for eval [039,059], and I got 334357 distinct games (uho-8moves-[-039,059].pgn).

Then I looked at Norman Pollock's database, games after year 1980, and truncated the games to 16 ply, did sort, uniq on the lines, and I got 261441 distinct 8 move lines. If we further filtered for lines that have appeared many times in the games of Elo 2400+ players, we would get even fewer lines (e.g., my polyglot book from this database constructed with 'min-game 16' has only about 36000 lines).

So either there is enormous unexplored territory in human opening preparation even within the first 8 moves, or the database of uho-8moves-[-039,059].pgn has a very large number of lines that have balanced end position only because of mutual errors. I conjecture the latter, but it is also possible that both are true.

We can check the uho-8moves-[-039,059].pgn database for mutual errors by random sampling to estimate the fraction of lines with mutual errors, so there is no need to analyse all. (I am not asking anyone to do this, of course; I may try it.)

Could Stefan or other experts comment on these observations. I am asking questions out of curiosity, and I am no expert.

[Krzysztof Grzelak]

Sorry, I just noticed that it cannot be accessed without login. I have now made it public. Please try again.

Thank you very much, now you can download. I'm interested in uho-8moves. I have a question whether this book includes debuts with the code A00.

[Krzysztof Grzelak]

I agree that for engine testing after opening phase it does not matter. But here is some investigation I did.

From Stefan's UHO database of 8 move games, I filtered for eval [039,059], and I got 334357 distinct games (uho-8moves-[-039,059].pgn).

Then I looked at Norman Pollock's database, games after year 1980, and truncated the games to 16 ply, did sort, uniq on the lines, and I got 261441 distinct 8 move lines. If we further filtered for lines that have appeared many times in the games of Elo 2400+ players, we would get even fewer lines (e.g., my polyglot book from this database constructed with 'min-game 16' has only about 36000 lines).

So either there is enormous unexplored territory in human opening preparation even within the first 8 moves, or the database of uho-8moves-[-039,059].pgn has a very large number of lines that have balanced end position only because of mutual errors. I conjecture the latter, but it is also possible that both are true.

We can check the uho-8moves-[-039,059].pgn database for mutual errors by random sampling to estimate the fraction of lines with mutual errors, so there is no need to analyse all. (I am not asking anyone to do this, of course; I may try it.)

Could Stefan or other experts comment on these observations. I am asking questions out of curiosity, and I am no expert.

Then please do it. Thank you in advance.

[chesskobra]

Sorry, I just noticed that it cannot be accessed without login. I have now made it public. Please try again.

Thank you very much, now you can download. I'm interested in uho-8moves. I have a question whether this book includes debuts with the code A00.

You have to look in the file. It is a text file that opens in any editor (on windows you may have notepad or something like that). All games have ECO tags. I see there are many games with A00 code.

[chesskobra]

I
We can check the uho-8moves-[-039,059].pgn database for mutual errors by random sampling to estimate the fraction of lines with mutual errors, so there is no need to analyse all. (I am not asking anyone to do this, of course; I may try it.)

Could Stefan or other experts comment on these observations. I am asking questions out of curiosity, and I am no expert.

Then please do it. Thank you in advance.

This, unfortunately, I won't be able to do anytime soon,