CCRL Discussion Board

Determining the W/D/L of a given position with a 99.9% probability to be correct is not 'simulating a tablebase'. Or with 99.999% probability. For one, an EGT will offer 100% correctness, and thus has an infinitely smaller probability to be wrong. But the most important distinction is of course that...

I think I got it working. I had a bit of trouble extracting the best move from the EGT in the proposed format. In particular the Duck part. Normally for an EGT that contains both wtm and btm positions you just do a 1-ply search from the current position, to select the target with the best DTM. But f...

I was distracted from this project for a long time. But it seems too interesting to abandon it entirely. In particular it would be fun to make a web applet where you could play the KRK end-game in Duck Chess. Since this is ony a 3-men EGT, efficiency is not really an issue. For propagating the DTx I...

And, does it run properly? You might be interest in the checkmating applets I wrote in JavaScript for 2-vs-1 and 3-vs-1 checkmating: https://www.chessvariants.com/page/MScheckmating-applet There also is a version that can handle bent sliders and hoppers: https://www.chessvariants.com/membergraphics/...

I am thinking about generating EGT for Duck Chess . The Duck can of course be treated like an extra piece. This would lead to an extremely large branching ratio, as both the Duck and another piece must be moved independently in each turn, and the Duck can be moved anywhere. It also drives up the mem...

In the abstraction in which an EGT is defined, all that exists is a set of game states that can be enumerated by an index, and transformations between those states (the move rules) for each of the two players. And two sub-sets of the game states where one or the other player has won. The latter can ...

Again sorry for the late response. This forum has so little trafic that I look at it perhaps only twice a year. For urgent questions I can be reached at talkchess.com, which I visit nearly every day. I never save the EGT I generate; I generate them only to get the WDL statistics. I might not even ha...

I added an easy trick to my EGT generator for getting a better impression of how many fortress draws a given end-game contains. Normally this would be masked in the statistics by the much larger number of draws due to grabbing an unprotected opponent piece (usually on the first move), changing the m...

I set up a web applet where one can practice checkmating with various pieces, including custom designed pieces, by playing against a (3-men) EGT. See http://hgm.nubati.net/rules/EGT.html . The applet is powered by JavaScript, and generates the required EGT on the client machine. I had expected this ...

I have started building End-Game Tables for Xiangqi (Chinese Chess), to enhance my engine for the upcoming computer Olympiad. It turns out that Xiangqi exngines tend to require a lot of technical skill. Quite unlike Chess, where end-games with a significant material advantage tend to 'win themselves...

In the Makruk end-game King + Silver General (S) + Ferz (M) vs King I encuntered the following manoeuvre to triangulate. It seems of general interest, as this is a case where really everything seems to be most unfavorable: neither Silver nor Ferz can triangulate, and the Silver breaks diagonal symme...

Interesting. It seems your intuition that larger boards are better is right, at least considering the Flamingo (F'). On 8x8 and 10x10 boards KBF'.K is a general draw, but on 12x12 it is a general win: $ ./4men allocate 432988480 bytes at 6fa040 mated mate King captures 66718336 mates 1800 ( 2.84 sec...

This chess on infinite edgeless boards is really something... It seems an active topic of math research, raising some very deep problems. In the mean time I found this thread about KBNK: http://chess.stackexchange.com/questions/1432/kbn-vs-k-checkmate-on-nonstandard-boards It contains an optimal mat...

I understood that due to abuse of the PGN Round tag the CCRL database is not compatible with SCID. If you have access to a Linux system or Cygwin you could use 'sed' to delete all Round tags.

I was wondering, is KBNK a won end-game on any (even) size board, and perhaps even on an infinite board? There doesn't seem to be a faster-than-linear increase in the length of the longest mate, as one usually sees in end-games that at some point cease to be won (e.g. K+K vs K). For KBNK I find 6x6:...

The funny thing is that 'any size board' is not the same as 'infinite board (with 1 corner)'. I am not sure what exactly the extra condition is that you would have to impose on the piece constellation on an infinite board to make it equivalent to 'any size'. I would expect something like "when ...

I created a novel Chess variant, which distinguishes itself by the peculiarity that no single piece occurring in it has mating potential. So Checkmates will also have to be performed by pairs of pieces. Hence the name 'Team-Mate Chess'. For people that like KBNK this would be the ultimate Chess vari...

My argument would be that in the end, to do the biggest thing that can be done, you would need that code complexity anyway. And once you have it, you might as well use it for the cases that did not really need it, rather than writing specialized code for that. I don't see why slicing would slow thin...

I would always slice by square. Why not use the extra reduction if you can?

Maybe the solution to this problem is to write down the results of each iteration on disk, and to keep on RAM only the information needed for next iterations. Indeed, this seems the way to go. If the working set consists of 2 bytes per board position (one byte for each side to move), the bytes coul...

OK, I'll join the discussion in that thread, then. But please note I have added something to my previous post, which probably crossed yours (as you don't quote it).

So do I understand correctly that you treat every (position, plycount) combination as a separate element of the EGT, multiplying its size by 100? What I had in mind was to let the EGT (or at least the working space for generating it) only contain one element per position, but keep both the DTZ and t...

I see no one has reacted to your posting for months. I can't help you with the Nalimovs; I never used those. You raise an interesting point about end-game tables: the information they contain is not very helpful to humans. There is little logic in the shortest lines of play they imply, and if there ...

What algorithm is used to generate DTM50? Is it sufficient to start from DTM50 successors, and start generating like you would for DTZ, except that you propagate DTM together with DTZ? And then, when a retro-move reaches a position that already was a DTZ win/loss (so you would normally ignore it), o...

I always make the effort to count symmetry-reduced positions twice when I take the statistics. This because I am interested in answering the question: "when you set up a random position, how large is the probability it is won (lost)?". When two different positions are symmetry equivalent, ...