Shortest-proof-game challenge

Yes, it's lovely! Rocky's 2x2 is good as well but a bit technical. Labelle's is more pleasing to the eye.

Euclide also considers it sound with 4 distinct solutions. I didn't test it because I saw the "C+" next to the diagram on the database. Looking at that page again, it actually mentions the software used was Euclide!

I still haven't tried Stelvio, even though when I googled it I learned that it's supposed to be much faster than Euclide. Maybe I'll check it out and cover it as a blog topic.

When you do, here's a problem you may want to check out. I saw a report of Euclide solving it in about 5 hours. Stelvio solves it in 8 seconds.

rn4n1/p7/bpppKpp1/k1r1p1qp/7b/5P1P/PPPPP1P1/RNBQ1BNR
36

P.S. I just found it at P1004103, where it says Natch solved it in 40 seconds, so the report on Euclide might be wrong.

That's a nice one. Labelle is a problemist/programmer who discovered the only known unique proof game that ends with just the two kings left on the board. An amazing task! Guess how much computer time was required to find this game? See my blog for the answer!

That's amazing indeed! I recall Andrew Buchanan mentioning that no such proof game existed for the complete massacre task yet but I suppose this is a rather recent one [edit: 2012 is not that recent but Andrew and I go a long way back]. It even conquered the final obstacle of a dead position before the last move(s) by leaving a pawn (even 2) till the end. I wonder if that was one of Labelle's conditions or that was sheer luck!

The next challenge is the Holy Grail960. With any luck, in a few years time, you can place 2 kings on an empty board and tell the contestants that the position was reached after 18.0 moves. Then you ask them to determine what the Chess960 setup position was. Solving time: 20 minutes. Start: NOW

what about this?

Here's a position which technically has two unique solutions that diverge from the first move. It's not impressive at all since both solutions are still related to each other, but maybe it will give somebody here inspiration.

(5951, Leither123)

@5952

Not that bad! The knight twin is perfect. Similarities in the solutions are often a plus but identities (like the last two black moves) not so much. But there are some 'filler' moves in almost every proof game!

One of several proof games in 7.5.

Yes! I tried to find 6.5 but that looks impossible. Btw, it is standard practice to specify the number of moves in composed SPG's. People with proof game solvers don't care because the solving programs find the shortest one anyway bur for humans it's hard without a number. Especially when it's unknown if the SPG is sound!

Actually, proof game programs require that the move count be specified. Short of a lucky guess when submitting a position, multiple attempts are needed to find the SPG, and further submissions are required to verify it really is the SPG.

This was also my initial solution. Here's one of the alternates:

I've noticed how most people on this forum tend to not include the length of the SPG, although most of the time it is because the position was randomly generated, so nobody has an SPG yet. However, even when I have an intended solution, I usually omit the intended move length in order to make the SPG more difficult to find, which I suppose is a little uncharitable.

Your assumption is correct. The recent posts gradually shifted from random diagrams to designed unique proof games. In the random diagrams the shortest number of moves is yet unknown so "finding an even shorter proof game" is always part of the challenge. Even when they are concluded nobody dare claim that the last version is the shortest possible one. For instance I always found shorter versions for those random diagrams whenever I tried!

Designed proof games are very different. Not only does the composer know the solution, he is also responsible for making sure it is correct. Which includes making sure there are no other solutions beyond the designed thematic alternatives. That becomes more important when also the number of solutions is unknown such that the solver cannot know what he is to find. E.g. I quickly saw there are too many unthematic solutions but couldn't tell if the composer intended shorter solutions and how many. So in creations like this one it is vital to specify the number of moves and the number of solutions unless the complete challenge was C+ tested on all intended solutions.

In the composition community the argument of "difficulty" to hide the number of moves for almost any solving challenge is rejected. Endgame studies are an exception because solvers can objectively identify that they found the correct solution on the basis of their chess knowledge (and by tablebase today). For most challenges they need the feedback of something telling them "mate in 4 is not good enough you gotta do it in 3". So indeed not giving the number of moves (and number of solutions) is harder but it is also considered an unfair challenge. Another well -known example are the helpmates which quite often have 2 to 4 solutions. It is immensely irritating to a solver not to know how many more solutions he must find after the first one! Imagine you believe there are 2 and must conclude there is just one after wasting many hours to find more. And finally, everyone knows that composers are not flawless. Specifying the (theoretically redundant) details of the objective enables the solver to find flaws the composer was unaware of before engaging in an infinite goose chase. So SPG challenges follow the format: SPG in x moves, y solutions.

Btw, chess.com puzzles are all outside the composition standard. Not because it is harder that way, or it is better, or they believe in it - but only because chess.com is 100% geared towards game play where nobody ever tells you it's mate in 3.. They really have no clue what a puzzle is, let alone a composition!

Btw, to illustrate my point, I did waste many hours on your SPG(s) until I read in subsequent posts that the goose I was chasing after had never seen the light of day

Well, worry no more, as I will now present a position with an intended length.

I intend for this to have PGs in 10.0, and I have my own solution ready (though I feel there are more solutions). To be honest, I didn't bother checking thoroughly to see if it's possible to get below ten, so feel free to try getting there

(Interestingly, due to the way I created the positions, the move length is hidden in the FEN! Has anyone noticed this before?)