CCRL Discussion Board

I'm not as good at programming as I once thought, either! However, I believe Ronald de Man is as good a programmer as he thinks he is (which is very high). Based on his comments over the years he only works on projects if and when he wants to, and is not likely to take on someone else's plan. So, yo...

I have a couple extra questions that I have not been able to find answers for, and which I think wouldn't hurt to put on public record. i) When running BayesElo, what values are used for anchoring the rating lists? ii) Was anchoring to FIDE rating lists performed, and if so, what historical notes do...

Clock speeds these days vary depending on the load on the CPU. For example on my i9 10900 if I run the benchmark on a fresh boot on its own, the clock will boost as high as 5.2Ghz but usually 5.0GHz. If I'm running 10 threads, it will be running at more like 4.0 GHz ish That surprises me. Is there ...

I have a question in regards to relating the benchmark calibration to running conditions. The answer is perhaps obvious, but better to ask since I have not seen it addressed in this forum thread. i) How many simultaneous games do CCRL testers typically run? ii) If it is more than one, is benchmark c...

The part of your algorithm that stands out are the square map tables and indexing function. But I must be missing something because I'm puzzled. What problem is this supposed to solve? That is to say, how does it bring 8-man and beyond tablebase construction in the the realm of current computer capa...

<r>I think distributed computing of large tablebases (large=8-9 men, at the time of writing) is technically possible. But this going on in real life I only see it happening if egtb generation were somehow how a by-product of time-space-proving crypto-currency mining. I am thinking in particular of C...

Found:
http://www.arves.org/arves/index.php/en ... plorations
https://drive.google.com/drive/folders/ ... PuGCkwXDTI

It is Christmas break for me and one of my annual habits during the yearly downtime is to spend a little time searching for interesting endgame news. I may have found one. In the Y-Combinator Hacker News discussion board, while discussing Centaur chess and Leela, one commentator makes a brief aside ...

The tablebases are in the mail. (When was the last time you heard that phrase?) By plan, jt will send along the full set to their recipient by the end of the week.

I don't know what "tfy" means. "Thanks for your" is my guess. In which case, you're welcome.

jk

I do. Also, some months ago I promised to John Tamlin's that I would top off his collection. I've filled but haven't mailed the hard drive yet. Procrastinating me. So by rights he should have a complete set, but is short a few. Based on earlier discussion Norm Pruitt (aka Pakman) may be closing in. ...

WDL50 should an order of 100 times less than DTZ50. WDL50 will be significantly smaller than DTZ50, but I wonder where you get the factor 100 from. Not sure, probably from Vincent. It would be great if you could get a list of actual sizes. I'm with syzygy on this one: I expect a factor of less than...

Please read what I write carefully . I'll repeat my definition once more: k does not stand for an integer; (k) stands for a (possibly infinite) sequence 0 < k_1 < k_2 < k_3 < ... < k_n. The DTZ(k)-metric for a position p is given by the pair: (min{i : p is won under k_i-move rule}, distance to zero...

Can a correct en-passant-aware tablebase be generated without considering en-passant rights in root? (Or without storing the positions with the en-passant right). I don't think so. Please correct me if I am wrong. Ah yes. You're correct on both. I missed second argument, positions with en-passant r...

Oh it does. The "+" implies there is no bound on the distance and one must accommodate single and double-byte entries. So in that regard the simplicity gained in going from DTM to DTZ50 is lost. True, but that holds for DTZ as well. Indeed it does - a factor to weigh into the consideratio...

I do. Difficulty of writing correct code, complexity of the file format, ease of use, and uncertainty of validation. In the simple-versus-complicated debate I side solidly with Kirill. In my opinion only "difficulty of writing correct code" is a relevant argument. The other aspects you me...

Competition format: prize (several complete solutions may compete in performance and features) vs bounty (the first successful solution wins). While this thread has generally veered off into discussions of what kind of tablebases to construct, I am reminded that there are many open issues. For exam...

Vincent brought some good points here about the WDL. What I did not realize at first is that WDL and DTZ50 are totally interchangeable during the generation. If you have the comlpete set of WDL tables, you can generate ANY DTZ50 table you wish. It makes perfect sense to use WDL tables necessary for...

I can understand that there is much interest in theoretical aspects of EGTB, especially among the theorists gathering here. :-) But please consider that we need something very practical and simple for wide adoption. And we need wide adoption to compute the important 7-men tables sooner. If it turns...

WDL tables give three possible position status. This is to be stored on 2 bits. I'm not sure of this, but a fourth position status can be used for free as it doesn't change the number of bits and therefore won't change final size of compressed data. As an identified advocate for cursed positions :w...

Congratulations Kirill! You've started a thread so enticing it brought Vincent out of the woodwork. The most improtant question is: Do you want all 7 men WDL quickly, say within 1 or 2 years, or do you want to get all DTZ in 16 years from now? Vincent Exactly. What are the ambitions of "the com...

Almost two years ago to the the day, in a teaser that might be considered this forum's earliest bounty hunting, kp1139 ventured willingness to pay for 7-men generator development. No telling how serious he was at the time since nothing came of it. Is the interest still extant? jk From the thread &qu...

The idea is to encourage creation of a useful 7-men EGTB generator and probing code, and construction of 7-men tablebases, by creating a bounty fund that will be paid to the first successful developer. An interesting idea! And you've introduced a novelty. Traditionally the prize originator is a mul...

KPPKPP does not promote directly to KQPKQP, only to KxPKPP. Given KPPKPP, and pawns on the 7th rank respectively, its possible for KPPKPP to convert to KQPKQP if the position does not check on queening. guyhaw's statement stands. (Nor does he require instruction that 4 pieces need 3 pieces for capt...

I have use of several mainframes and I have tested chess engines and retrograde analysis and machines with faster memory access such as Cray's machines are much faster then some other machines such as a Sun Enterprise. Nice. Conduct some carefully designed experiments and show us scaling numbers. jk

Endgame tablebase generation can be highly parallelized, but not uniformly, and not necessarily efficiently. When I say "not uniformly" I mean two things. Different force combinations are not equally easy to run in parallel, nor offer equal parallelism. Secondly, the threading code must be...