Chess.com's Weekly Study: August 28th 2016

Hello Chess.com!

Starting with 2016, we will be posting a Weekly Study, courtesy of Yochanan Afek, Grandmaster for composing endgame studies. These challenging positions are designed to stimulate (and improve upon!) your creativity, depth of calculation, pattern recognition and pure imagination.

A. Gilbert, 1986

Enjoy! 

In case anyone is wondering whether the composer of this study is named Gilbert or Gillberg, it's Gillberg.

The most brilliant part of the solution content for me is 3. ... Kg3 4. Rf3+! (not .Rg1+? Kh2! .Rh1+ Kxh1 .Bc6+ Rxc6 draw!) Kg2 5. Bc6! These are the jewels that turn a good looking study into a great looking study!

After 3...Kg3 4.Rg1+, how easy is it to demonstrate the draw in order to confirm the analysis?

I checked it on the tablebase; that is totally satisfactory in study land today. Nobody feels the urge to provide proofs to convince club players or the ELO-1000 community on chess.com. Nothing personal of course, it is about evaluation standards. As I often write: "the only puzzle standard is perfection".

I wasn't sure whether an appeal to the tablebase is permitted in presenting the solution. If it is, without any conditions, there is a simpler solution: 1.e7 winning, per the tablebase.

That is an interesting point! You might argue that the tablebases have pushed thousands of endgames studies into oblivion but I don't think it is interpreted that way. I am not sufficiently active in the endgame world - I am a member of the "problem" society but not of the "study" society in the Netherlands - to know how they discuss these things and whether or not there is an official viewpoint. I do have an intuition though which I will describe. By the way, there is another involvement of engines that has become standard. The evaluations of diffcult solution variations (not in the tablebase) may be left to engines. When Komodo reports a +3.0 score for a position eval and no concrete analysis refutes it, the author may assume that the variation is a win. I am not sure of the cut-off value but I recall it is somewhere between 2.0 and 3.0.

In general, it is not assumed though that positions in a tablebase or knowledge system cannot be valid chess compositions - and that applies to old as well as new entries. This may change when there are good computer programs which can trace and identify interesting endgame studies in a tablebase. I assume work is in progress on this chore.

Which moves are part of the composition and which of the tablebase is IMO a choice of the endgame composer. Endgame studies feature ideas, mostly in a patterned fashion and in combination with other ideas. The composition stops where the presentation of those ideas stops. What remains are technical finishes which are considered (tablebase) theory. I don't think it is possible to separate the two in an objective, unambiguous fashion but such is not required. The author decides and it is up to him to convince the public and judges that his endgame is a unified and whole product.

The current study is a great example of this approach. The idea is obviously about the cooperation of B+R to harrass the black king and finally enable the promotion move. A technical ending of Q vs R+B clearly does not fit that pattern. Nevertheless, the author might include one or two moves of the tablebase theory provided they are unique.

In the final analysis the author decides and it is not in his interest to refer to a tablebase at move 1 of a solution since that would leave him with a 'no content' study. But he can quite legimately do so at any point a position is no longer part of the desirable content.

If I understood, you said it's OK to use a tablebase to dispose of variations unrelated to the theme of the study. I assumed the rule, if there is one, would be something like that. I have many favorite studies with fewer than eight men that I expect to continue amazing my chess club friends with for years to come, unfettered by any tablebase considerations.

That is true. But remember that old studies were also occasionally accompanied by long analytical variations which nobody cared to study but were necessary to prove the study correct. Today, some of the analysis is replaced by referral to the tablebase or chess engine.

Solvers and presenters of studies always concentrate on the ideas of the study and not so much on boring, difficult variations. They will not be bothered a lot when some of those are present besides the attractive main line + tries. They will trust the rest of it and rightfully so!

Studies are not primarily made to solve but to appreciate. Some of the modern studies are less suitable for solving (though still a small percentage) and should not be used in solving contests. I can tell you by experience that in the problemist world great care is taken when selecting problems and endgames for such an event.

I agree though that the interference of the machine takes some of the romance out of the artistic chess puzzles. if I am well informed, that's what they call 'progress'

Note: If you look at the pre-tablebase study by Gijs van Breukelen "la danza del elefante" often referred to as 'the hardest problem in the world', you will notice black can escape from the main line by playing ... Kg4 at the approriate time leading to a very difficult ending. in the old days everyone lazily said: "white wins by his material superiority" but modern engines are unable to produce really convincing wins. Still the study is amazingly popular and presented in chess venues all over the world. Not to mention its monthly repeat performance on chess.com. It just never goes away! 

Personally, I think it is more questionable to say "we think this position wins but it is too difficult to prove" than "we know this position wins (as per tablebase) but it is too complicated to prove to you". There's the view from the other side.

Interesting parallel between this conversation and what passed through the mathematical community a [relatively] short while ago regarding computer assisted proof. Most famous example is the Four Colour Theorem. The proof is solid, but carried out by exhaustive checking of mind bogglingly large amounts of unique cases and subcases by computer.  No human could ever hope to check it without finding a more analogue method of proof. Which ironically did come about, using theorem generating software.

As a former student of mathematics, the four colour theorem has had my life long interest. Always felt there had to be a four line proof for it somewhere :-) Does it exist now? Or at least something that can be explained on the level of Gödel, Escher, Bach?

Another parallel between the four-color theorem and some of the problems here is the large number of faulty solutions put forward by amateurs.

Only matched by those for Fermats last theorem  

By current math standards Fermat was a highschool amateur. If he ever proved it, an amateur might have hoped to walk in his steps. All over now, sigh ...