CCRL Discussion Board

My question is:

how is it possible to obtain a Nalimov-type EGTB (say, for 5-men) which takes into account the 50-move rule?

I am not sure at all that the combination of the DTM and DTZ50 tables are sufficient to reach that goal.

A friend of mine is trying to program such an EGTB, starting with knnkq and knnkp, but he has difficulties with a sizable proportion of positions which seem to bring about endless loops...

Do you have comments, or links or references to reflections on the subject?

Isn't that simply a matter of seeding the tablebase from the successor TBs with the true DTM initially, and then retrogradely building exactly 50 cycles?

unfortunately for the makers of chess engines, this entire thread is a moot point, since it has been the fide ruling for a few years now that it is up to the arbiter to decide whether an appealed position is a draw, thus meaning arbiter's discretion on the entire 50 move rule, meaning it is sometimes 40 moves, and sometimes played out until one player runs out of time or the game is mated or agreed drawn by players.

academically, dtz50 bases do exist, though they never became popular due to the facts that a: subperfect play can turn mate in 60 into mate in 50 and b: assuming that the position is drawn in some cases where mate is in 3 or 4 moves from the 50 move cutoff is not appropriate due to the displayable mate ruling by fide. if memory serves, the dtz50 bases that were created were all thompson bases though.

gambit3 wrote:unfortunately for ...

Thanks for your interest, gambit3, but I asked a clear question (I think...) so your ramblings don't help much!

ernest,
I would recommend that you were more inclusive about encouraging input from gambit3, some of whose points are well made.

DTZ50 is a relevant metric if the 50-move rule applies, but ... FIDE may change the 50-move rule (and have in the past), and there are events (e.g. ICGA CCCs) where the arbiter might suspend the 50-move rule.
However, I don't know of any recent FIDE ruling giving arbiters the power to suspend the 50-move rule in competitions for carbon chess players.
Also, the DTZ50 EGTs that have been produced emanate, not from Ken Thompson, but from John Tamplin using a code written by Marc Bourzutschky.

DTZ50 EGTs must _not_ be created by building on the DTM EGTs of sub-endgames. KBBKNN is one of the clearest demonstrators of this. Take away those KBBKN wins which are 'beyond' a 50-move draw-claim, and you majorly impact the wins in KBBKNN - even reducing maxDTM and maxDTC/Z. Data on the impact of the 50-move rule has been published by the (Bourzutschky, Tamplin, Haworth) troika.

However, the DTZ50 EGTs are built by creating a DTZ EGT, based on the DTZ50 EGTs of sub-endgames, and stopping when all wins of DTZ = 100 plies have been identified. This can be done without back-tracking, i.e., without resetting depths already given to positions - so your friend should not be 'getting into loops' as you describe.

g

I forgot to say that gambit3 is quite right in saying that DTZ50 EGTs have limited use in that they don't tell you 'how far away' you are from winning under the 50-move rule.
I define 'how far away' by how much the defender has to concede before you can limbo under the 50-move rule. If you could win under a 52-move rule, by definition, you are '2' from winning under the 50-move rule.
That is why I defined the DTR metric: DTR is the minimaxed maximum-length of any phase of play, i.e. the maximum move-count value when the sides are trying to minimax this maximum move-count value. If DTR is 52, you can win under the 52-move rule but not under the 51- or 50-move rule.
So, more kudos to gambit3.

g

In the chess programs I use they enforce rules depending on options, they ignore engines as some do not enforce the rules.

I apologize ; for I know you all are in an important discussion on how to get a program to take the 50-move rule into account in real game play / competitive environments -- but my "uneducated question" is : how to get my program to (even temporarily) IGNORE the 50-move rule when for example testing it with special "FEN's"/position like Wilhelm -- I want to at least see the end result for instructive purposes.

To add ; with me not yet having read Fritz12's Instruction booklet; can't I change the parameters of "null move" or "draw value" to get the desired result ?

Hennie Oosthuizen alias cracae

If WILHELM or another chess-engine is using EGTs, they are probably DTM EGTs which take no account of the 50-move rule.

g