I was putting together a Sample Database of about 25 games for Tactics Time so that people could download it, and get a better idea of the product, and what it is like.
I wanted to include a game or two that was a miniature.
I was looking at this game:
[Event “uChess rated”]
[Annotator “Fritz 12 (10s)”]
1.e4 e5 2.Nf3 f6 3.d4 d6 4.Bc4 Ne7 5.dxe5 fxe5 6.Nxe5 g6 7.Bf7# 1-0
which is one of thousands I have collected over the last couple of months. I think that I found this on the RedHotPawn.com forum, in a thread related to minatures, where people were posting their games related to quick kills in less than 20 moves. Here is a link to the thread, which has been going on actively since 2005: http://www.redhotpawn.com/board/showthread.php?subject=CHECKMATE_in_under_20_moves&threadid=26440.
This position came up as one of the variations:
When I was running Fritz in “Full Analysis” mode, it recommended: Qh5+ and Bb5+ in this position, but failed to mention the read killer: Bf7+! which does a Remove the Defender tactic – forcing Kxf7 followed by Qxd8, winning the queen.
This tactic is very common when the d file is open, and the black king can be checked on f7 with the light square bishop. This tactic often comes up in the “Smith Morra Gambit”, and is great to know, especially for blitz.
I also tried this game on Fritz 10, and it didn’t do any better (and was actually worse).
I also experimented with giving the engine more time, but this did not improve the output at all.
Here you can see the Full Analysis output, with the questionable recommendations for the spite check Qh5+ or Bb5+
So the lesson of the story is – chess software can make mistakes – so be sure to think for yourself