Sunday, December 09, 2007


Four Problems Part II

In a recent post comparing CT-Art and PCT the subject of the relative difficulty of their exercises came up. I had made the comment that at the highest level of difficulty they were about the same difficulty but some other posters commented that CT-Art had more difficult problems at the highest levels.

So, I decided to investigate by selecting a couple problems from each to examine closer. I am not suggesting that my selections are representative or even fair and this is certainly not a scientific sampling. I decided to take "black to move" problems from the middle of the highest level of difficulty (or as reasonably close as I could easily navigate to, in the case of PCT). Any conclusions apply only to these four problems and not necessarily to all of the problems for the highest levels of either piece of software.

I've created a link where you can play through the problems and variations with added analysis and evaluation by Fritz. The link is: fourproblems. For each problem at that link, ignore the initial white move as it is just setting up the problem and it is not part of the exercise as given by PCT or CT-Art. In the earlier post I gave the initial positions and main line of the solution as provided by the software. I repeat that below with some additional commentary. My comments refer to the additional analysis at fourproblems so you might want to review that or follow along with that open in another tab or browser if you can.

The analysis and evaluations are based on a minimum ten-ply look ahead at each move using the Deep Fritz engine (Nov 2000 edition) from ChessBase 8.0.

Problem A
From PCT Tactics Module 6 Unit 26 Exercise 4

1...e4+ 2.fxe4 Nb2+ 3.Kd2 Rxc3 4.Bd4 Rfc8 5.Bxc3 Bh6+ 6.Ke2 Rxc3
Fritz agrees with this solution. This problem is quite difficult with significant competing lines to consider at three branches.

Problem B
From CT-Art Problem 1177

1...Bxh3 2.gxh3 Rxd4 3.Bxd4 Nf3+ 4.Kg2 Nf5 5.Rg1 N5xd4 6.Bd3 Nxg1 7.Qxd4 Nxh3
Fritz prefers 5. Qe3 (a variation given by CT-Art) but gives an improvement in that line. In this main line Fritz gives 6. ...Qe5 as a significant improvement. I find this position to be more challenging that "Problem A".

Problem C
From PCT Tactics Module 6 Unit 26 Exercise 5

1...Ba3+ 2.Kxa3 b4+ 3.Ka4 Nb6+ 4.Kxb4 Rb5+ 5.Ka3 Qc1+ 6.Bb2 Nc4+
Fritz gives the better defense of 3. Kb2. PCT does not attempt to show multiple variations in a single exercise and I do not know if they elsewhere show this with this alternative defense. To see / know 1...Ba3+ works one would have to consider the 3. Kb2 line. As Liquid Egg Product pointed out in the comment, since white has mate in one, black's key move is easy to find (it has to stop the mate and/or be a check). This is the easiest of the four problems.

Problem D
From CT-Art Problem 1179

1...Nb4 2.fxg4 Qxa2+ 3.Kc1 a5 4.Rhf1 Ne4 5.Qe1 Rc4 6.Rf4 a4 7.Rxe4 Rxe4 8.Qg3 h6
Fritz prefers 2. Nc1 (a variation given by CT-Art) but otherwise agrees with CT-Art. I found this problem to be similar in difficulty to "Problem A".

My conclusions: The most challenging problem of the four ("B") came from CT-Art and the easiest ("C") from PCT. I thought that the other two "A" (PCT) and "D" (CT-Art) to be similar in difficulty. If these are typical then CT-Art does indeed have more difficult problems at the highest levels than PCT.

My intent was to consider difficulty but I must add a word about quality. CT-Art made multiple errors in their analysis of what appears to be the most difficult problem. The only way I can get real value from CT-Art on such problems is to independently determine and verify the solutions. While I learn something by doing that, I can do that completely without CT-Art in the first place.

As always, your thoughts, comments, and criticisms are welcome.

Labels: , ,

The problemset of CT-art is from the pre-computer-check era. Such problemsets I found to have an error rate of about 20-25%. I tested early George Renko CD's and Polgar's middlegame brick.
I have to agree with Tempo here, the CT-Art problems are riddled with errors at the top levels. Arguing the merits of CT-Art based on the quality of problems at the top levels is, in my opinion, a poor argument.

By the way, a more detailed analysis of the errors of problem 1177 can be found here.
When I generate javascript for multiple games in chessbase 8.0, I get separate playing boards. How do you generate one board with mutple liks below, like your problem boards.

Convekta really needs to computer check their problems. I have a list of about 50 errors in CTB, and the frequency of errors increases with problem difficulty (which doesn't bode well for CT-Art, whose level 10 is about CTB Phase 3).
loomis: Thanks for the link to your analysis of 1177!

anon: I'll try to post some screen shots of what I do (probably this weekend or later)...
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]