CCRL Discussion Board

DTM50 work is making progress. Galen and me are currently checking our 3-5 men DTM50 EGTs against each other, and no bug has been found so far. So we are not far from getting a complete and fully verified 3-5 men DTM50 set, with 6-men expected to follow later (Galen's generator already has some 6-me...

I have thought about trying to hook Nalimov tables to my generator, and then using existing DTM tables which are known to be 50-move safe. But this might be a big undertaking as I know next to nothing about the format, and it might not help anyway, since underpromotions such as to kbbnkr have to be...

Ok, I think I got it. Our ideas are different because my generator back propagates DTM when galen's (and also the one h.g. muller has in mind) back propagates DTZ (a good sign being that my generator and galen's one have totally different designs and up to now have produced results that match exactl...

So do I understand correctly that you treat every (position, plycount) combination as a separate element of the EGT, multiplying its size by 100? You’re correct, but I don’t store as many as 100 values for every position, because a lot of values are identical. I represent DTM50 of a position as a s...

My DTM50 generator works as follows : 1) identify illegal positions in order to exclude them from all subsequent searches 2) do all decaptures and depromotions from all possible subendings, and get this way an upper bound (if white to move) or lower bound (if black to move) of DTM50, for every plyco...

Sent you a pm with my kpkp and kbbkb DTM50 TBs

I have this position as mated in 85, 87, 91, 95, 98, 101, 103, or 118, or draw, depending on the initial count. When it's zero, here is a mate in 85 line (not unique):1. ... Kd4 2. Nd1 Kd3 3. Ne3 Kd4 4. Kf4 Kd3 5. Nef5 Kc4[1] 6. Ke5 Kc5 7. Ng7 Kc4 8. Ne6 Kd3 9. Kf4 Kc3 10. Ke3 Kc4 11. Ke4 Kc3 12. K...

Of course, memory is the real limitation. I could pretty much right now build arbitrary 6-man pawnless non-5v1 tables with the existing codebase, except for the tiny wrinkle that it would require a few hundred GB of RAM to do it. Maybe the solution to this problem is to write down the results of ea...

. if we want to try comparing tables, maybe it'd be best to start with one or two-- perhaps you can upload to a hosting service-- and I'll see what I can figure out. Although, since you've checked your results against my text, it's looking pretty positive. Since our TB results have matched exactly s...

knnkp: To be super-precise, I find it begins to strike at depth "51.5"; e.g., there are positions where Black is mated in 51 DTM but more in DTM50 (e.g., Ka4 Nc3 Nh3 / Ka8 Nc4, Black mated in 55). Other than that, same results. Yes some positions lost in 51 (DTM) for Black are also 50-mov...

(continued) checking galen's results : 2 bishops vs knight : same results – only a typo in 2nd position (PC=80 or 81 instead of PC=79 or 81) rook+bishop vs rook : results for the 2nd position (Kb4 Rh3 Bg8 / Kb1 Rf6) are off 1 move : My TBs give mate in 17 for PC up to 67, in 19 for PC=68-69, in 20 f...

Hi, Sorry to have been so long to reply. I'm extremely busy nowadays... First of all, I shall congratulate galen for his DTM50 work, and above all for his blog that perfectly explains what DTM50 is, what is the influence of the move counter and how two positions identical on the board with different...

I've corrected the pgn above, there was an error at move 81 (81... Qc3+ shortens the mate by 3 moves by allowing 82. Qc5, the correct move is 81... Qa4+). Thanks to ernest who identified the error. The result is validated by two independantly written DTM50 generators and can now be considered as a c...

Hi ernest, glad to see you again. I've corrected the pgn above. For now I use a self-made probing code that works poorly and must be ameliorated. I produce the pgns manually, so there can be a few errors. 49. Kc5 is 49. Nc5 instead. 18. Ne6 is also not a check. The rest is correct. I'm glad to learn...

Ernest, If you're still with us, I have (finally) the solution of your problem. I hope you won't mind getting the answer 6 years after the problem was posted :lol: , but now the knnkp DTM50 tablebase exists and if you or anybody else has the same demand about any other knnkp position, the response s...

Yes, the 50-move rule hasn't always a sterilizing effect on very deep positions. It will certainly have one on pawnless endings, rubbing out all the records that have been found there (mate in 262 in krnknn, mate in 545 in kqnkrbn, winning nature of krbknn etc...) and limiting pawnless ending depths...

I’ve started generating 5-men DTM50 tablebases. I’ve almost no time to do it and progress is very slow, but I expect to generate all 3 vs 2 / 2 vs 3 with 0-1 pawns before the end of 2013. The remaining ones (2-3 pawns, 4 vs 1 / 1 vs 4) will need new code and I expect to have them generated next year...

Hello muzzoid ! I'm currently working on 5-men DTM50 tablebases. These tablebases are the only ones who can give distance to mate playing with 50-move rule, and currently only exist for 3-4 men (see my previous posts past year). They are more complicated and more difficult to compute than standard D...

A post discussing first ideas about DTM50 compression strategy – I’ve very little experience in this area and about what to do exactly in order to achieve good compression rates and access time for future probing engines. Help welcome :) ! DTM50 is a movecount (MC) dependant metric. This means that ...

3Q4/3K4/8/8/8/3k4/3p4/8 w - - 48 1 MC=48 plies for the initial position. This position is also maximal for DTM, but DTM=28 whereas DTM50=34 if MC=48 precisely (if MC<48 DTM50=DTM=28 and the winning lines are the same, and if MC>48 the position is drawn). The 23 first moves are the same for DTM and ...

All 3-men and 4-men DTM50 tables are created (also kpkp with en-passant rule). Tables are in raw format with no compression and there are still many redundancies to be eventually discarded, but they already shrink to less than 300 MB using winzip. I expect that with specific compression and no redun...

I've not been sitting still, though . Just finished generating the 6-piece WDL50+ and DTZ50+ tables for regular chess Nice. My own vision is that the obvious path for developing new EGTs for regular chess – more men on board – is going to require fast growing computing resources that already exceed...

Syz, good to see you again. This forum is so quiet nowadays… Except that best play will not lead to MC >= 82 in such positions Of course. Clearly these first DTM50 curiosities are academic and will never happen in real games, but this won’t hold true when endings such as knnkp or kqpkq will be probe...

Look at this krkp position : Ka8, Rf5 / Kg8, Ph7 (white to move) The DTM50 EGT shows the following results : If MC<82, DTM50=19 If MC = 82 or 83, DTM50 = 20 If MC = 84 or 85, DTM50 = 28 If MC>85, draw Why should the mate be delayed by as much as 9 moves for some particular values of MC ? Let’s follo...

I’m currently looking at DTM50 (distance to mate metric, playing with 50-move rule). I’ve built a generator capable of building raw (uncompressed) DTM50 tables for all 3-men, pawnless 4-men , pawnful 4-men (soon), and 5-men later (I need to upgrade my hardware to handle them, I’m waiting for Haswell...