Nordlandia wrote:
How long will it take to generate 5-men syzygy in DTM format on high-end hardware (in my case i7-5960X 8-core) and estimated volume with good compression technique?
The generation time is ultimately based the size of the TB's. Using a 12 core Mac Pro with slightly older CPUs (x5690) was able to generate syzygy all 6 Man EGTB's in less than 72 hours - let's call it 60 hours . This is about 150 GB.
Simply amazing.
He already said, he ain't interested in generating DTM-Syzygy
Because there is no such thing, and he gets tired of people asking the same stupid questions all the time. For one, DTM EGTs already exist (Nalimov, Gaviota). So what is the point in generating new ones? Just use those that already exist.
Because there is no such thing, and he gets tired of people asking the same stupid questions all the time. For one, DTM EGTs already exist (Nalimov, Gaviota). So what is the point in generating new ones? Just use those that already exist.
Peter explained some time ago how Texel combines information from both Syzygy and Gaviota in a game theoretically correct way
hgm wrote:Because there is no such thing, and he gets tired of people asking the same stupid questions all the time. For one, DTM EGTs already exist (Nalimov, Gaviota). So what is the point in generating new ones? Just use those that already exist.
To be fair, I think the question is for DTM50, which is ~natural play. These apparently do exist (because I've seen some discussion of TB results), but for some reason are not publicly available (and so no one uses them).
Of course DTZ50 and DTM50 are not the same thing, and I imagine converting a generator from one to the other is not easy. Then again, the szyzygy generator is open source, so anyone interested in this can work on it if they want to...
Doesn't magically avoid tossing away material to enter a won TB position, of course, but I suppose that isn't why people want DTM50.
To be fair, I think the question is for DTM50, which is ~natural play. These apparently do exist (because I've seen some discussion of TB results), but for some reason are not publicly available (and so no one uses them).
But what would the specs be for a DTM50 TB? For different values of the 50move counter at root you will have a different values of DTM50.
For game play this is _perhaps_ not relevant as one enters TB territory through a move which resets the 50move counter. On the other hand I have some difficulty understanding how (game theoretically correct) followup play would work as now 50move_counter < 100 but the TB would still assume 50move_counter=100 leading perhaps to incorrect moves being played.
Ideas=science. Simplification=engineering.
Without ideas there is nothing to simplify.
To be fair, I think the question is for DTM50, which is ~natural play. These apparently do exist (because I've seen some discussion of TB results), but for some reason are not publicly available (and so no one uses them).
But what would the specs be for a DTM50 TB? For different values of the 50move counter at root you will have a different values of DTM50.
For game play this is _perhaps_ not relevant as one enters TB territory through a move which resets the 50move counter. On the other hand I have some difficulty understanding how (game theoretically correct) followup play would work as now 50move_counter < 100 but the TB would still assume 50move_counter=100 leading perhaps to incorrect moves being played.
From what I've read, in DTM50 you treat the ply counter as an extra variable element of the position, inflating the size of the tablebase by a factor 100. This is what makes DTM50 impractical. I haven't given much thought to how this could be improved, but presumably people who have worked more on tablebases have. I'd be interested in learning how that can be improved...
Michel wrote:For game play this is _perhaps_ not relevant as one enters TB territory through a move which resets the 50move counter.
The problem is that you don't have to enter the won sector of the EGT through a reset of the counter. When you enter the EGT in a drawn position, and try to swindle the opponent, he could make the losing mistake only after 45 reversible moves. And then you are really in a hurry to convert.
So indeed, DTM50 does not exist in he same sense as DTZ50 or DTM. You would need a different table for every value of the ply counter, hugely driving up the size compared to DTM. Of course you could make compromises, by only keeping the DTM50 table needed to play the end-game when you enter it in a won position, and play by DTZ50 in the rare cases where you could swindle the opponent, just to be sure. But then you might as well play by existing DTM plus DTZ50. No need to generate anything new.
The reason for requesting DTM-Syzygy is that Stockfish/Komodo is much faster giving definite evaluation when probing tablebases compared
to Houdini.
Stockfish probing tablebases is faster in reaching winning EGTB position than Houdini. If Stockfish/Komodo is able to probe DTM, then they're able to reach winning EGTB position much faster than Houdini.
The frequency of EGTB probing on Houdini is very slow compared to Stockfish and Komodo.
Well, so DTM is bulky and slow, while DTZ50 is small and fast. Now what? You want even bulkier and fast? Well, I want a job that pays a million a month for which I have to work only 10 min a day...
IMO DTM is an obsolete concept today, WDL is enough in real games and in studies the engine will easily analyse an endgame (so with few pieces) up to very high depths in short time, and will work out the mate by itself, including 50 moves rule and everything else.
