Thursday, September 06, 2007

Another CTB boner [?] Or, another BDK boner

Warning: original post was incorrect. The position below is a draw! I've corrected the post, indicating revisions in brackets.

Two people (a kind reader via email) and Rise 'n' Shine have both provided error lists for CTB (Chess Tactics for Beginners). I am finding more and more errors, especially in the K/P endgames (I'll post the master list once I finish the Circles). Hey Convekta, how about you computer check your bloody computer software? Or at least let us modify problems to change the solution, or to mark problems so they aren't presented.

Below is [another counterintuitive problem] that [I thought was an error which had] slipped by our two censors. Black to move and "draw." Problem 1290:

The first time I did this problem, it was using the stare for 10 minutes method. This time I was using the 'think hard about the solution, question the solution, see if it makes sense, explain it to yourself' method, and red lights started a'flashing in my head. There's something that seems decidedly undrawish about a white queen on h8, with the black having K on e3 and a pawn on f2, with white to move!

[Note added: well, that 'red light' was wrong. While Fritz, for the first couple of minutes of 'infinite analysis' said the position I describe had a #36 (mate in 36 moves), some very trustworthy chess bloggers have said this is indeed a draw. (I thought the queen against seventh rank pawn won unless the pawn was a rook pawn, but apparently there is a clever stalemate trick with the bishop pawn, as described in the notes.) A little exploration on my own part has shown they are indeed correct. So it is I who have pulled a boner. I guess I really need to get Rybka: who knows what other lies Fritz is telling me? Fritz has been my go-to guy: when I don't understand or trust a solution in a book, I use him to tell me who is right.

(I just got back from work, and after Fritz analyzed the position for nine hours, it still had it at +15 rather than the correct 0.0 or so....how very odd).

Sorry CTB: while you have many errors, I was wrong about 1290. Also, good job censors.]

15 Comments:

Blogger Grandpatzer said...

I've found several examples where the program only recognizes one answer, when another is equally valid. For example, I find a mate in 2 but the computer only recognizes its "pet" mate in 2.

Haven't kept a record of the offenders, though. Sorry :(

9/06/2007 04:37:00 AM  
Anonymous ChessTyro said...

Unless you can force the queen to f1, I believe it's a draw.

9/06/2007 05:15:00 AM  
Anonymous Anonymous said...

Hello!

It is a draw indeed.
When a black pawn is on the 2nd rank
in K+Q vs K+P endgames white tries to move his King next to the pawn and then the queen takes it. Black tries to move her King off the promoting square so that the pawn can promote but black must also protect the pawn with her King.
So white's plan is to check the King by moving the Queen next to the pawn, and when black puts her king on the promoting square "under" the pawn white moves his King one step closer to the pawn.

But in the case of the f2 pawn (or c2) black should try to put her King on g1 threatening f1=Q. Then when white checks black plays Kh1.
If QxP then Stalemate!!. So black never plays Kf1 again and white's King hasn't got any time to reach the pawn.

Try it out

9/06/2007 05:33:00 AM  
Blogger Glenn Wilson said...

Appearances can be deceptive. It looks drawn to me. Maybe I am missing something?

Q v rook pawns and Q v Bishop pawn on the seventh (second) are known draws. If the other King is close enough there are mating chances. In the position you mention with the f-pawn Black can eventually play his K to h1 when Qxf2 will be stalemate.

See Pandolfini positions 131 and 132.

9/06/2007 05:49:00 AM  
Anonymous ookwelbekendalsemc said...

Pretty f'd up when you're trying to solve problems (putting in lots of time and energy) that aren't even correct.

9/06/2007 05:55:00 AM  
Anonymous svensp said...

I think, it's really a draw, even though it doesn't look like it. Whenever the queen might take the pawn on f2 blacks' king can be at h1, so there would be a stalemate. White doesn't have time to bring in his king because the black king doesn't need to block the pawn at f1 more than once) and as the queen can never conquer the pawn it's drawn.

kind regards,
svensp

9/06/2007 07:20:00 AM  
Blogger Loomis said...

"There's something that seems decidedly undrawish about a white queen on h8, with the black having K on e3 and a pawn on f2, with white to move!"

So far it looks like a draw to me. I'll give you my reasoning and you can tell me if there is a way to win I am missing.

White needs to be able to get his queen to f1 safely in order to win that position. Can he do it? If white allows the black king to e1 or e2 he'll never make it. So white with his next move needs to stop Ke2 and f1=Q. Since he can't go to the a6-f1 diagonal, the only choice is a check on the e-file. But then after Kd2, white must now prevent Ke2, Ke1 and f1=Q and he just can't do it.

The reason the white queen needs to get to f1, is that for white to win, he needs to bring his king down to help the queen. But he needs to prevent black from promoting. The only way to do this is to force the black king to f1 or put your own queen there. Here's what happens if you try to force the black king to f1:

Qe5+ Kd2 Qf4+ Ke2 Qe4+ Kd2 Qf3 Ke1 Qe3+ Kf1, mission accomplished, you can approach with the king, e.g. Kg5 Kg1 Qg3+ and now there is a very key move, Kh1! This time the king is not forced to f1 because Qxf2 is stalemate.

Anything I'm missing?

9/06/2007 09:45:00 AM  
Anonymous Anonymous said...

The position you describe is a draw. The key idea is that the Black king can run to h1 rather than f1 (the latter blocks the pawn and allows the White king to return). If the pawn is captured by the queen, it is stalemate. Thus White cannot make progress.

9/06/2007 10:49:00 AM  
Blogger Blue Devil Knight said...

Thanks very much to everyone who pointed out this is indeed a draw, and describing the trick. I've changed the original post to indicate this.

Luckily, about 0.001% of my games are not decided by endgame technique but non-subtle middlegame mistakes.

One thing CTB should do with this problem (it is for beginners after all) is extend the solution out a few more moves, to show the actual trick. This is a program for beginners, after all.

At any rate, this is motivating me to read up on Silman's book. If this is important for my level, it will be in there.

9/06/2007 02:25:00 PM  
Blogger Temposchlucker said...

If you have a fast connection, you should really think about downloading the Nalimove tablebases. There are some pretty fast servers out there. It took me only 2 hours to download 7 Gb! It is really amazing to get the announcement mate in 32 within less than a second. Rybka has its flaws too in the endgame.

9/06/2007 07:26:00 PM  
Blogger likesforests said...

Nice puzzle; it's a draw.

1... Kf4 2. Kg7 (2. Kxf7 Kf5 3. Kg7 Ke6 4. Kxh7 Kf7 =) 2... f5 3. Kxh7 Ke3 4. Kg8 f4 5. h7 f3 6. h8=Q f2 =.

9/06/2007 09:09:00 PM  
Blogger likesforests said...

Yes, in my estimation Fritz is rather weak at evaluating endgames. Just this morning it told me that I could mate with K+P+B vs K+P+P when I had the bishop and it was not the same color as my pawn's promotion square. Rybka and Pandolfini say otherwise. ;)

Even the *free* version of Rybka is very strong. The only weakness I know of in the free version is it doesn't spot solutions with underpromotions.

9/06/2007 09:14:00 PM  
Blogger katar said...

dont download tablebases, just go here
http://www.shredderchess.com/online-chess/online-databases/endgame-database.html

peace out.

9/07/2007 04:29:00 AM  
Blogger Blue Devil Knight said...

Katar: amazing.

9/07/2007 12:25:00 PM  
Anonymous Anonymous said...

Hi, Nice Blog

problem 763 has an error.

2/13/2010 03:23:00 PM  

Post a Comment

<< Home