Pilot

Hello, my name is Martino, I'm italian but I live in Brazil. I want to try to talk about chess related stuff.

So, last week I was looking at the following wonderful study (Prokes 1946). White to play and win.

Here is the solution (select it with your mouse): 1.Kd2 f2 2.Rd1!! g2 3.Ke2+ and wins.

Another great one is the following (Troitzky 1895, Hannemann 1915).

White to move and win. Solution: 1.d7 Rg6+ 2.Ke5 Rg5+ 3.Ke4 Rg4+ 4.Ke3 (or 4.Kd3 Rg1 5.Kd2 Rg2+ 6.Kc3 Rg3+ transposing to the main line) Rg3+ 5.Kd2! Rg2+ 6.Kc3 Rg3+ 7.Kc4 Rg4+ 8.Kc5 Rg5+ 9.Kc6 Rg6+ 10.Kc7 and wins. Note that white cannot allow the black rook to get to the d file and has to take care of the fact that if the white king moves to the d file, for example Kd4, then black plays Rg1! with the idea of a skewer on the d file.

The previous study follows, I would say, from the extremely famous Saavedra-Barbier study (1895), which is the following (white to move and win).

Solution: 1.c7 Rd6+ 2.Kb5 Rd5+ 3.Kb4 Rd4+ 4.Kb3 Rd3+ 5.Kc2 Rd4! 6.c8=R!! (6.c8=Q? Rc4+! 7.Qxc4 stalemate) 6...Ra4 7.Kb3!! and wins.

This gave me the idea of trying to compose chess endgame studies, and let me tell you, it's not easy.

First of all I looked for chess study databases, and I found this.

The study above (the first displayed in this post) is included, and in fact it's only here that I found who composed it. Someone called Prokes in 1946.

So as I was saying I started trying to compose and the first idea that I had was to use the tablebase as a support:

The idea is very simple: set up any position with at most 6 pieces and the "magic" tablebase will solve it for you. Not as a chess engine solves it but as a mathematician would solve it. If it's draw, he tells you that it's draw and the move. If it's not draw it tells you how many moves the winning side needs to checkmate the opponent and he gives you the move. You can make that move, and in the resulting position he will tell you the same thing (who wins and the move). You don't need to be a mathematician (although maybe you do) to know that this process gives the exact and complete solution of every position with at most 6 pieces (when I say 6 pieces I'm including the kings .. kings are pieces aren't they?).

So what did I try to do? I tried to set up some positions to see what happened. At some point, I don't remember how, I was dealing with the following tricky position (white to move).

It's tricky because there is only one winning move, and it's 1.Kc4. The idea is 1...Rxd6 2.d5+! Ke5 3.Re7+ cutting off the king, and now it's a basic win using the Lucena maneuver. If instead 1.Ke4? then 1...Re5+! leads to a draw (although it's enough to play 1...Ra5). The idea is that the white king cannot go forward.

Ok, so you might say, what does this have to do with anything? My idea was to create a study that reached the above position in the main line.

This is an example (although quite trivial). The line goes 1.Rxh7 Rxc5 (1...Kxc5? 2.Rh5+! Kb6 3.Rxa5 Kxa5 4.c7 wins) and we are back to the previous position. But this is too trivial. The following also is too trivial.

Here the idea is 1.Nd3+ Kd6 2.Nxc5 Rxc5.

The following is also trivial.

Here the idea is to take twice on c5. The problem, or I should say one problem, is that the first move is not unique, white having the choice between bxc5 and dxc5. Notice that of course it is not just a matter of finding a winning combination. You need to create a position (like this one) where there is only one winning move/idea. So you need to check that cxd5, cxb5 and Rh6+ lead to nothing.

The reason why they are trivial is that white is challenged with the choice between b4 and d4 only when the key position occurs. I would like a situation where white has to commit (choose between those two squares) before reaching the key position. Let me tell you again, this is not easy to achieve.

I'll try to work more on this and maybe I'll come up with something. Cheers :-)