World Draughts Forum

Syzygy table-bases are not the holy grail, they contain WLD and DTZ (distance to the 50 move rule) but lack DTM (distance to mate). Nalimov table-bases contain DTM (without considering the 50 move rule), but they are larger 1.12 TB vs 150 GB (and somewhat slower to probe). I usually combine them, Sy...

I have gotten distracted with many other interesting machine learning projects (non-draughts) that kind of had to happen right now before anything else. But I will soon start working again on my Keras/Tensorflow pipeline for Scan-like pattern evals. One improvement that is possible for any eval tra...

In practice 1.4B positions are difficult to handle (one needs fast amounts of memory for this) so I used only 240M. I have gotten distracted with many other interesting machine learning projects (non-draughts) that kind of had to happen right now before anything else. But I will soon start working ...

Joost, very interesting, thanks for posting this update. Your results with floating point and non-incremental eval are already very good. It wasn't clear to me, are your training positions labeled as WLD, or with an eval score from a quick search? Also, what time controls did you use for the matche...

Inspired by Berts enthusiasm about NN and ML in general, also by the good results I've got with a NNUE style NN in my chess engine, I started 2 weeks ago to implement a fully connected NN in the draughts engine I'm sometimes working on. Contrary to Berts implementation with int16 SIMD code I decided...

Is it possible to connect two computers with each other so that the draughts program can use two processors and RAM memory. Let's imagine a situation where I play a game of draughts with a human and I would like the program to play with maximum force. You need a draughts program with a distributed ...

Thank you Joost. I have a huge request to you Joost. You wouldn't have made the Nalimov (6P) database available online. Krzysztof, I don't think my provider likes it when I upload 1.2 terabyte. I'm willing to put Nalimov 6P on a 1.5 TB portable hard-drive, but than you have to sent me one, or I can...

Thank you for your answer Joost. It doesn't make sense to buy such a disk. And how does it relate to chess Joost. For chess it's the same, it only helps with EGDBs. I have a few of these 2TB drives with Nalimov 6P (1.2 TB), syzygy 6P and Kingsrows EGDBs on them. In practice I only use Nalimov becau...

I have a few of these NVME drives in use (although 2TB), the tranfer-rate (random and sequential) is about 5 to 6 times as high as what you get from regular IDE SSDs. They are very useful for endgame databases, for other applications I don't see much difference. You are probably better off with the...

I would like to buy a hard disk Samsung 970 Pro 1TB M.2. I have a question. Is it worth buying it and what will be the disk performance for draughts such as Kingsrow, Damage, Dragon, Scan, Maximus , Sjende Blyn, Moby Dam and other draughts programs. I'm also interested in what the performance will ...

If someone sends me a tool (binary or source) then I think permission is implied. So if that tool can parse a file of positions into a table of features and do gradient descent on it to get a bunch of feature weights, I would certainly use it to see what comes out of it. Using the source to reverse...

Perft is addictive. With the same processor the Intel Core i7-8700K, and frequency 4.5 GHZ, another test. This time the perft for the initial position has really improved, others almost the same. I forgot that I have a switch in my program to disable updating hash etc. when running perft, so I rera...

It was not my intention to post here anymore, but your perft numbers are amazing and you clearly beat the old perft numbers I've got back in 2017 with my move generator based on magics. I just ran perft() on my old Broadwell i7-6950X at 4.4 GHz. At the second position I'm still a little bit better, ...

Rumors indicate that also a 16-core Ryzen is in the works :) Bert Indeed! There are also rumors that AMD will add 4 way interleaved memory to Ryzen to make it competitive with Intel HEDP. However one of the lead developers of Ryzen says that there are no intentions to add 4 way memory, so I don't k...

Bert, Right now both my machines (8 and 10 core) are busy producing a huge number of games of considerable depth, if you wish I can run your experiment at a larger depth when this job is finished. As an alternative I am willing to give you a remote login to my 8 core machine (which is on a separate ...