A hard endgame ... for who?

A lot has been written about Gijs van Breukelens "Elephant dance" (elephant = knight), presumably the hardest endgame that even Stockfish, Carlsen and others couldn't solve - but in my opinion largely and increasingly the result of mystification.

Today, SF is very well capable of solving the riddle of the dancing elephants, and with regard to the scores of grandmasters and worldchampions I would wish to know the precise conditions of their solving challenge. Many 'professional' problemists are capable of solving such endgames, simply by recognizing the themes (ideas) behind it. I am sure Carlsen could solve it as well given the information and tools available to the professional solvers and a fair amount of time.

Solving is a bit harder for SF, simply because, as a general purpose game analyzer, it will not be aware of an underlying design intent to wrongfoot game players. It will find the virtues of the first underpromotion but will treat it as incidental. A problemist will think: "ah, I see a pattern emerging, lets look for other knight promotions!" and that will give him a big search advantage.

But, the elephant study is from the seventies of the last century and things have changed a lot. Today, composers do their job from behind the computer screen very similar to our managers, our scientists, our office workers. Much of the arduous testing of complicated lines is taken care of by chess engines, leading to very different compositions. In the 21st century, the ideas of composers still dictate the main solution lines but you are likely to find more complicated tries and variations on the side than 30 years ago. Without the engines, todays endgame creators would be obliged to reject many of their new constructs for pure uncertainty about the status of some complex lines. On top of it, the engines with their quick evaluations (relative to human analysis) will help them much faster through the composition process, saving time for adding extra content to the endgames.

And so, many modern endgame studies are found to contain complex, often prozaic variations to support the thematic content of the composition. They won't show up in the presentations of these endgames because presenters focus on the themes, not on the sidelines. They are all in the smallprint.

Here comes the big point. The engines in the 21st century are not primarily used to analyze the quirky themes of composers, but to generate verdicts on many gamelike variations accompanying the crown jewel themes of the composition. And that is just regular chess and something they are very good at!

The typical example endgame below may appear very intricate to humans with its gamelike variations, but is in fact quite easy for an engine rating these quickly on a depth of 50 ply and more.

Concluding, a problem like the elephant dance is relatively easy for a human and relatively hard for an engine, while the opposite is the case for many modern studies designed with the help of an engine. Provided the human performs the same validation work on the sidelines and not trust it to the author of course. Stands to reason, right?

Arisktotle, is this your own composition? Has it appeared in print?

(1) Yep (2) Nope.

Since my pension I have become very careless with publication of my compositions. I placed a simpler version of this one on an obscure site rendering it ineligible for prize competitions. It's the first one I made with full assistance of an engine.

Note: still another version is with the bQ on c3 and Pe3 on e2. I preferred the play in that one but I liked the extra bQ sac on the first move in this one. Lots of saccing in this study!

I was thinking, if it has been published, it ought to go into HHdbVI.

Placing it on any website counts as published. Who mines all the websites for this material?

Arisktotle, if you want to write up the solution in the form you would submit for a competition and send it to me, I will forward it to Harold van der Heidjen and suggest he put it in HHdbVI, which is due out next year. He would also want your first initial and last name and the date it was first published in the chess.com forum.

Thank you for the offer, n9531l1! I had a slightly different plan. I have a profile on the ARVES composers site and I wanted to ask him (or whoever is in charge) to put my recent studies in my profile. Of course, he will then also include them in HHdbVI. I have delayed this action as I have some doubts on which studies and which versions to select. Like I am not sure whether I want the violent intro for the current study or a more refined version. I intend to send the backlog in one sweep and not in little packages, It will take some time to complete the gift wrap for all of them.

I would appreciate your help in testing a few positions I am not sure of, e.g. relating to the 7-piece endgame tablebase. There is one in a try at the 10th move in the current study as you can see in the comments. SF says it should be OK, but in endgames I'd rather trust the tablebase.

Lomonosov server is down now. I'll check again later.

Lomonosov reports that after 18. Kxg5, Black can win with Kf3 (DTM 42), Kd3 (DTM 46), or Kd4 (DTM 47).

Thanks! That puts my mind to rest. SF gave me -5 but the way it rates some endgame draws (remember the one with the 2 pawns and 6 bishops) you can't even be sure at -25.

The 7-piece Syzygy tablebases is a free alternative to Lomonosov and it's just as good for testing studies. Check out my blog More adventures with endgame tablebases for a short discussion of their differences.

That's right. When Lomonosov was down, I forgot about asking Syzygy. It gives the same three winning moves, all with DTZ 2.

Thanks Rocky64! Didn't know a free version was around though I've seen the name Syzygy a few times. Not one that sticks in your mind as a search term but I have it bookmarked now. Also read your interesting blog. I know Andrew Buchanan well and know the example you gave but I am surprised he was engaged in similar endgame activities as early as 2002.

This is very useful for analyzing an endgame I have on the shelf with several 2N vs 3P variations. Hope I won't run into the DTM/DTZ issues!

By the way, I had and will be having interesting discussions with Andrew. I generally disagree with his approach to self-classify problems and endgames - e.g. classify a diagram as retro because there appears to be a retro element. This leads to all kinds of problems in the retro field and it hurts peoples common sense of logic - you are entitled to know the rules before you start solving something. 

A wonderful study, deep and most surprizing and unexpected in many aspects. What's nice is that by the variations one can learn a lot even for "natural" chess.

I came to this through your solution of today's endgame pawn problem (many thanks for this as well).

How long did you work to figure this one (above) out? 

By the way, the strange "syzygies" appear in mathematics but seem to originate from astronomy, see

https://www.encyclopediaofmath.org/index.php/Syzygy

I made it a few years ago - and still look occasionally at perfecting it. I think I worked a few weeks on it at the time, which is really short by my standards!

Thanks for the compliments and the Syzygy information!

ok