The ever alert <CG> member <Oliveira>, well actually <jjones5050>, spotted an inconsistency in the game score (i.e. movelist) for the game:
<Spielmann — Eliskases (R7) Austrian ch m, Semmering (1936)>
Here is the contemporaneous coverage of the game:
But, as <Oliveira> found, <ChessBase> has the move 24.Kh2, as does essentially all the databases (<MillBase>, <Chessbase>, <ChessTempo>, <365chess>, <redhotpawn>, etc). Playing 24.Kh2 then makes 24…Nxg3 sensible, and comports with Eliskases typically good play.
Of course, even GM’s make mistakes, sometimes for several moves. Or play decidedly suboptimal moves, for various reasons. But if all the other databases are in agreement, I strongly feel that proactive affirmation evidence should be on hand to change the score. And it should be documented within the game.
It’s the “Principle of Congruence” (or “Conformance” if you prefer).
Biographers such as myself need to trace the origins of changes in order to understand and verify. And these changes are best contained within the PGN, because that’s the currency used in our business. At the very least, if the full documentation isn’t embedded, an indication of the change should be made in a comment, plus a pointer to the detailed change log (rationale).
An example of where <CG> changed a move, with no indication whatsoever in the actual game data, is here:
<Lein, A — Shulman, Y (R2) Foxwood op, Ledyards CT (2006)>
The game can be played, within 0.5-pawn differential eval, with both 2…b6 and 2…b5.
Myself, I would prefer to leave the score unchanged, i.e. to keep the original 2…b5, without affirmative evidence to the contrary, in order to maintain congruence with all the other databases, include the definitive <ChessBase> database. Of course, there is an admittedly high likelihood that the move 2…b6 was played, and that information would make a most welcome comment in the score.
Unfortunately, I was overruled, because several individual players on <CG> felt that they knew how Shulman would play. Indeed, having the pawn on b5 causes the eval to “sawtooth”, but never more than to change += to -=, for instance. Not nearly as decisive as the eval differences discuss in our 1936 game. There, the matter changes the very outcome of the game. In the 2006 game, the destiny of the game is completely unchanged whether 2…b5 was played, whether or not either side makes a pawn capture involving it.
The real reason for changing the move is because it spoiled the esthetic appreciation of certain individuals.
What disturbed me more, is that Shulman was apparently known to several of these same individuals. And so contacting him to verify the score seems an simple and obvious way to remedy the situation, i.e. to obtain affirmative evidence that the change is valid. But it was deemed too trivial an issue to contact him, or too bothersome. Of course, it wasn’t too bothersome to send an email to Shulman, if a bet or a point of pride were on the line. Somehow, I fail to understand why a player would be bother with a question about their actual play, especially if the game is stored worldwide on various databases.
Consider the 1936 game. It is clear that <CG>‘s score is wrong on several fronts. Congruence with other databases, the eval of the game play (using a non-arbitrary threshold of an Informant level, from -+ to =), and finally, and most definitively, the historical data provided above.
Why bother to dig out the actual move-list shown above? Why not just change the move because “we just know”, on the authority of what should be obvious to any strong player?
Is this some sort of metaphysical question?
I can ask other, related questions:
Why bother to correct illegal PGN in the <CG> database?
Why bother to ensure correction crosstables can be generated from tournament data?
The answer is simple. Because it’s the right thing to do. It respects the value of the game data by respecting the process of data-handling. And if the high standard of the process is adhered to by all, there is a collective benefit to the data integrity, the user’s, and the researchers. Plus, there should be a consistency, that applies to all users, respecting their capabilities, while at the same time ensuring fairness. And this has the valuable benefit of encouraging more individuals to contribute to the effort.
For instance, consider another game where the moves “don’t make sense” :
<Polgar, J — Spassky, B (R10) Budapest (1993)>
Here, apparently neither GM noticed that a transposition was made in the typical order move order at moves 11-12, one which was not exploited – so it was missed by both players. This was not a matter of half a pawn, but what a decisive +- which was missed. There would be a natural tendency to “correct” these moves. Indeed, according to the discussion on <CG>, it seems that <ChessBase> at one time did change the move order (or, if they didn’t make the change themselves, incorporated a “corrected” version). One should note that it’s been corrected in the online version.
OK, then, let’s return to the 1936 game. Knowing what the correct data is, one can naturally ask:
Where does the error in the <Spielmann — Eliskases (R7)> game come from?
One can learn from mistakes as well. And here we can thank <Oliveira> for suggesting the most likely cause:
This game is contained in “500 Games of Master Chess” by Tartakower and Du Mont, and the moves given there are the same as on this site. However, ChessBase gives 24.♔h2, after which the subsequent moves make perfect sense.
So, we know who to blame (well, where to place the blame, as it could be a typesetting error, etc.).