How fair Chessbase is in including poorly annotated games in Megadatabase

[bmp1974]

How fair Chessbase is in including poorly annotated games in Mega database? Chessbase claims that they have included xyz number of high quality annotated games but take for example, the below high profile game played by Kasparov against Ivanchuk in Linares 1990.[pgn][Event "Linares 08th"]
[Site "Linares"]
[Date "1990.??.??"]
[Round "7"]
[White "Ivanchuk, Vassily"]
[Black "Kasparov, Garry"]
[Result "0-1"]
[ECO "B97"]
[WhiteElo "2665"]
[BlackElo "2800"]
[Annotator "Boensch,U"]
[PlyCount "86"]
[EventDate "1990.02.??"]
[EventType "tourn"]
[EventRounds "11"]
[EventCountry "ESP"]
[EventCategory "16"]
[SourceTitle "CBM 017"]
[Source "ChessBase"]
[SourceDate "1990.08.01"]
[SourceVersion "1"]
[SourceVersionDate "1990.08.01"]
[SourceQuality "1"]

1. e4 {Bönsch} c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4
Qb6 8. Qd2 Qxb2 9. Rb1 Qa3 10. f5 Nc6 11. fxe6 fxe6 12. Nxc6 bxc6 13. Be2 Be7
14. O-O O-O 15. Rb3 Qc5+ 16. Be3 Qe5 17. Bf4 Qc5+ 18. Kh1 Ng4 19. h3 e5 20. Na4
Qa7 21. Bc4+ Kh8 22. hxg4 exf4 23. Nb6 d5 {N} (23... Rb8 24. Rxf4 Bd7 25. Nxd7
Qxd7 26. Rf5 Qa7 27. Rxb8 Qxb8 28. Qf2 a5 29. g5 Qd8 30. g6 hxg6 31. Rf3 Rxf3
32. Qxf3 Bg5 $10 {Konovalov - Zablockij corr. 1989 -Inf.48/369}) 24. exd5 cxd5
25. Bxd5 Rb8 26. Nxc8 Rbxc8 $10 27. Rh3 Qb6 28. Re1 Bg5 29. Re6 Qd8 30. c4 Rb8
31. Qd3 Bh4 (31... h6 $2 32. Qg6 f3 (32... Rf6 $2 33. Qxg5) 33. Rxf3 Rxf3 34.
gxf3 Bf4 35. Kg1 $16) 32. Be4 Qg5 (32... Qxd3 $2 33. Bxd3 Rbd8 34. Rxh4 Rxd3
35. g5 $16) 33. Bxh7 $2 {>= 33.Dd5} Rfd8 34. Qc2 f3 $3 35. Rxf3 Rd2 36. Qe4
Rd1+ 37. Kh2 Re1 $19 38. Qf5 $19 Rxe6 39. Qxe6 Kxh7 40. Qe4+ g6 41. Rh3 Kg7 42.
Qd4+ Kg8 43. Qe4 Qf6 0-1

[/pgn]

When one looks into the game under the lens of Stockfish engine, there appears lot of mistakes in the annotations:

1. 31..h6 is not a bad move as noted
2. 32...Qxd3 is not a bad move as noted
3. 33. Bxh7 is not such a bad move as noted
4. 34...f3 is not a very good move that gives any advantage to Black
5. 37...Re1 is actually a blunder by Gary but shown as a winning move. correct was Qc1
6. 38. Qf5 is actually a move that leads to equality but shown as Black winning
7. 41. Rh3 is actually the final blunder by Ivanchuk that loses the game but not shown that way in the annotation (41.Rf7+/g3/Kh3/c5 all leads to equality)

[Cornfed]

How fair Chessbase is in including poorly annotated games in Mega database? Chessbase claims that they have included xyz number of high quality annotated games but take for example, the below high profile game played by Kasparov against Ivanchuk in Linares 1990.[pgn][Event "Linares 08th"]
[Site "Linares"]
[Date "1990.??.??"]
[Round "7"]
[White "Ivanchuk, Vassily"]
[Black "Kasparov, Garry"]
[Result "0-1"]
[ECO "B97"]
[WhiteElo "2665"]
[BlackElo "2800"]
[Annotator "Boensch,U"]
[PlyCount "86"]
[EventDate "1990.02.??"]
[EventType "tourn"]
[EventRounds "11"]
[EventCountry "ESP"]
[EventCategory "16"]
[SourceTitle "CBM 017"]
[Source "ChessBase"]
[SourceDate "1990.08.01"]
[SourceVersion "1"]
[SourceVersionDate "1990.08.01"]
[SourceQuality "1"]

1. e4 {Bönsch} c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4
Qb6 8. Qd2 Qxb2 9. Rb1 Qa3 10. f5 Nc6 11. fxe6 fxe6 12. Nxc6 bxc6 13. Be2 Be7
14. O-O O-O 15. Rb3 Qc5+ 16. Be3 Qe5 17. Bf4 Qc5+ 18. Kh1 Ng4 19. h3 e5 20. Na4
Qa7 21. Bc4+ Kh8 22. hxg4 exf4 23. Nb6 d5 {N} (23... Rb8 24. Rxf4 Bd7 25. Nxd7
Qxd7 26. Rf5 Qa7 27. Rxb8 Qxb8 28. Qf2 a5 29. g5 Qd8 30. g6 hxg6 31. Rf3 Rxf3
32. Qxf3 Bg5 $10 {Konovalov - Zablockij corr. 1989 -Inf.48/369}) 24. exd5 cxd5
25. Bxd5 Rb8 26. Nxc8 Rbxc8 $10 27. Rh3 Qb6 28. Re1 Bg5 29. Re6 Qd8 30. c4 Rb8
31. Qd3 Bh4 (31... h6 $2 32. Qg6 f3 (32... Rf6 $2 33. Qxg5) 33. Rxf3 Rxf3 34.
gxf3 Bf4 35. Kg1 $16) 32. Be4 Qg5 (32... Qxd3 $2 33. Bxd3 Rbd8 34. Rxh4 Rxd3
35. g5 $16) 33. Bxh7 $2 {>= 33.Dd5} Rfd8 34. Qc2 f3 $3 35. Rxf3 Rd2 36. Qe4
Rd1+ 37. Kh2 Re1 $19 38. Qf5 $19 Rxe6 39. Qxe6 Kxh7 40. Qe4+ g6 41. Rh3 Kg7 42.
Qd4+ Kg8 43. Qe4 Qf6 0-1

[/pgn]

When one looks into the game under the lens of Stockfish engine, there appears lot of mistakes in the annotations:

1. 31..h6 is not a bad move as noted
2. 32...Qxd3 is not a bad move as noted
3. 33. Bxh7 is not such a bad move as noted
4. 34...f3 is not a very good move that gives any advantage to Black
5. 37...Re1 is actually a blunder by Gary but shown as a winning move. correct was Qc1
6. 38. Qf5 is actually a move that leads to equality but shown as Black winning
7. 41. Rh3 is actually the final blunder by Ivanchuk that loses the game but not shown that way in the annotation (41.Rf7+/g3/Kh3/c5 all leads to equality)

Better question: How "fair" is it to judge 30+ year old non-verbal annotations by U Boensch by modern engine analysis?

I might point out that 31. Re1 is not called a 'winning move'...Boensch's appended symbol only indicates that the position is winning.

Ivanchuck vs Kasparov is pretty 'high quality'. Annotations...well, 30 years from now Stockfish's "annoations" may prove suspect as do some engine (and human!) analysis from 30 yrs back.

[KLc]

Oh that's not even bad. There's so much crap in the MegaDB: fragments, incorrectly dated, mix-up of blitz and classical rating, the "good games" filter throws out actually good (and well-annotated!) games, etc. They just keep on adding more blitz online games ("now 9.5 million high quality games" blah blah) instead of cleaning up the mess.

[Cornfed]

Oh that's not even bad. There's so much crap in the MegaDB: fragments, incorrectly dated, mix-up of blitz and classical rating, the "good games" filter throws out actually good (and well-annotated!) games, etc. They just keep on adding more blitz online games ("now 9.5 million high quality games" blah blah) instead of cleaning up the mess.

That said, I don't think one is going to find a database this large and well indexed with fewer issues. I can't think of one.

A while back I went thru some older Mega's to compare and found they do indeed add old games, tournaments (forgotten or missed), notes etc. It really is quite impressive when you think about it.

I do however create my own 'Reference base' (despite the fact that 'big data' generally points you in the right direction) with players just over (depends) 2200 or over 2400 where I augment with annotated data from books/magazines and older good players 'pre-elo'.

[bmp1974]

How fair Chessbase is in including poorly annotated games in Mega database? Chessbase claims that they have included xyz number of high quality annotated games but take for example, the below high profile game played by Kasparov against Ivanchuk in Linares 1990.[pgn][Event "Linares 08th"]
[Site "Linares"]
[Date "1990.??.??"]
[Round "7"]
[White "Ivanchuk, Vassily"]
[Black "Kasparov, Garry"]
[Result "0-1"]
[ECO "B97"]
[WhiteElo "2665"]
[BlackElo "2800"]
[Annotator "Boensch,U"]
[PlyCount "86"]
[EventDate "1990.02.??"]
[EventType "tourn"]
[EventRounds "11"]
[EventCountry "ESP"]
[EventCategory "16"]
[SourceTitle "CBM 017"]
[Source "ChessBase"]
[SourceDate "1990.08.01"]
[SourceVersion "1"]
[SourceVersionDate "1990.08.01"]
[SourceQuality "1"]

1. e4 {Bönsch} c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6 7. f4
Qb6 8. Qd2 Qxb2 9. Rb1 Qa3 10. f5 Nc6 11. fxe6 fxe6 12. Nxc6 bxc6 13. Be2 Be7
14. O-O O-O 15. Rb3 Qc5+ 16. Be3 Qe5 17. Bf4 Qc5+ 18. Kh1 Ng4 19. h3 e5 20. Na4
Qa7 21. Bc4+ Kh8 22. hxg4 exf4 23. Nb6 d5 {N} (23... Rb8 24. Rxf4 Bd7 25. Nxd7
Qxd7 26. Rf5 Qa7 27. Rxb8 Qxb8 28. Qf2 a5 29. g5 Qd8 30. g6 hxg6 31. Rf3 Rxf3
32. Qxf3 Bg5 $10 {Konovalov - Zablockij corr. 1989 -Inf.48/369}) 24. exd5 cxd5
25. Bxd5 Rb8 26. Nxc8 Rbxc8 $10 27. Rh3 Qb6 28. Re1 Bg5 29. Re6 Qd8 30. c4 Rb8
31. Qd3 Bh4 (31... h6 $2 32. Qg6 f3 (32... Rf6 $2 33. Qxg5) 33. Rxf3 Rxf3 34.
gxf3 Bf4 35. Kg1 $16) 32. Be4 Qg5 (32... Qxd3 $2 33. Bxd3 Rbd8 34. Rxh4 Rxd3
35. g5 $16) 33. Bxh7 $2 {>= 33.Dd5} Rfd8 34. Qc2 f3 $3 35. Rxf3 Rd2 36. Qe4
Rd1+ 37. Kh2 Re1 $19 38. Qf5 $19 Rxe6 39. Qxe6 Kxh7 40. Qe4+ g6 41. Rh3 Kg7 42.
Qd4+ Kg8 43. Qe4 Qf6 0-1

[/pgn]

When one looks into the game under the lens of Stockfish engine, there appears lot of mistakes in the annotations:

1. 31..h6 is not a bad move as noted
2. 32...Qxd3 is not a bad move as noted
3. 33. Bxh7 is not such a bad move as noted
4. 34...f3 is not a very good move that gives any advantage to Black
5. 37...Re1 is actually a blunder by Gary but shown as a winning move. correct was Qc1
6. 38. Qf5 is actually a move that leads to equality but shown as Black winning
7. 41. Rh3 is actually the final blunder by Ivanchuk that loses the game but not shown that way in the annotation (41.Rf7+/g3/Kh3/c5 all leads to equality)

Better question: How "fair" is it to judge 30+ year old non-verbal annotations by U Boensch by modern engine analysis?

I might point out that 31. Re1 is not called a 'winning move'...Boensch's appended symbol only indicates that the position is winning.

Ivanchuck vs Kasparov is pretty 'high quality'. Annotations...well, 30 years from now Stockfish's "annoations" may prove suspect as do some engine (and human!) analysis from 30 yrs back.

My point is if the quality of annotations are not accurate and misleading, why can't Chessbase remove them?
They can also recheck all important games with the best engine and correct as necessary.

[Uri Blass]

I am against removing annotations because it may be interesting to see also the mistakes that people did in the past in the analysis.

I think that it is better to add new annotations without removing old annotations so people can see 2 versions of the game(with old wrong annotations and with new annotation that is verified by stockfish).

[KLc]

UI do however create my own 'Reference base' (despite the fact that 'big data' generally points you in the right direction) with players just over (depends) 2200 or over 2400 where I augment with annotated data from books/magazines and older good players 'pre-elo'.

Could you give some details how to do this? I’m having a hard time figuring out how to extract the really “good/quality” games. After 1970 one could take Elo. But I’m more interested in the “classical era”. There are many games in Mega from the 19th and 20th Century by club player etc which I would somehow like to filter out.

[Cornfed]

My point is if the quality of annotations are not accurate and misleading, why can't Chessbase remove them?
They can also recheck all important games with the best engine and correct as necessary.

It's about history.

You don't take the commentary on games from tournament bulletins or old newspaper articles or really anything produced by an individual and just 'change it' because some would sacrifice historical accuracy for ones desires. Heck, there may even be legal questions in some cases.

Quality is a product of the time it was made in. If you found a Model T or a 'state of the art' car from 1958 and have someone throw away the insides just to make it 'work better? You could...but...I can't think of anyone who appreciates such things ever doing so.

[Look]

It seems to me there is a ChessBase product called "Big Database". Without any annotated game.

[jdart]

Megabase has issues, but as noted upthread, it is better than most of what is out there.

I have seen vast quantities of crap games available for download or even purchase. These typically have mangled, inconsistent or missing headers (abuse of the "Event" and "Site" fields is rampant), improper PGN, transcription errors, duplicates, etc. chessgames.com is one example, although they've been slowly cleaning it up.

As for annotation, yes, pre-computer era annotation is full of errors, especially in the endgame. Pre-tablebases and before engines that can see 40-ply lines, GMs working on their own often made wrong assessments.