Manual: How to use Syzygy (or any other) 6-men without SSD

[Milos]

So if you don't have expensive SSD and want to get accurate endgame analysis. This is rather trivial but I guess ppl don't know it so it would still help some.
Basically you need to move only 2-3 most common wdl dtz 6-men tables (you can always determine which these are according to the position you are analysing) into separate directory and point that one (together with 5-men directory) in SF path for Syzygy. Don't point other 6-men because it will kill the search speed totally.
Even though many of 6-men bases are more than 200MB each, it will still be enough even for machine with little RAM to load couple together with 5-men into RAM and have no slow down.
Slowdown mainly comes from the engine accesing all possible 6-men from the given position (which could be quite a few) even though most of the accessed 6-men are just useless positions deep in search that appear in only couple of % of the search nodes.

[bupalo]

good suggestion, thanks, but do you know if in the sygyzy do you have to put all the wdl and the dtz in the same root directory? I use stockfish but even with 5 pieces it doesn't give istantaneous the right result even if it accesses 1 tablebase.

[syzygy]

good suggestion, thanks, but do you know if in the sygyzy do you have to put all the wdl and the dtz in the same root directory?

You can distribute the files over multiple directories however you like. Stockfish will look for each file in the directories listed in SyzygyPath.

I use stockfish but even with 5 pieces it doesn't give istantaneous the right result even if it accesses 1 tablebase.

See here, section "What to expect", second paragraph.

[bupalo]

thanks all clearer

[Albert Silver]

So if you don't have expensive SSD and want to get accurate endgame analysis. This is rather trivial but I guess ppl don't know it so it would still help some.
Basically you need to move only 2-3 most common wdl dtz 6-men tables (you can always determine which these are according to the position you are analysing) into separate directory and point that one (together with 5-men directory) in SF path for Syzygy. Don't point other 6-men because it will kill the search speed totally.
Even though many of 6-men bases are more than 200MB each, it will still be enough even for machine with little RAM to load couple together with 5-men into RAM and have no slow down.
Slowdown mainly comes from the engine accesing all possible 6-men from the given position (which could be quite a few) even though most of the accessed 6-men are just useless positions deep in search that appear in only couple of % of the search nodes.

And what are the ones you recommend keeping for search?

[mvk]

So if you don't have expensive SSD and want to get accurate endgame analysis.

You can also opt for a cheap SSD.

[Albert Silver]

So if you don't have expensive SSD and want to get accurate endgame analysis.

You can also opt for a cheap SSD.

Here in Brazil there is no such thing.

[Milos]

And what are the ones you recommend keeping for search?

Depending on the position you are analysing. If position is currently 8-men, there are only comb(6, 2) posible 6-men outcomes (after 2 captures, and you can trade only 6 pieces out of 8 since you can't capture king).
Out of those 15 possible 6-men most are useless since they never happen (like one side loosing 2 pieces except pawns). Useful ones are after each side captures a piece (Q for Q or R, R for R or B or N, B for B or N, N for B or N, N or B or R for P, etc.)
No need to include bases for after promotion since usually the side that promotes first has an easy win so it doesn't need TB to see it.

[Milos]

So if you don't have expensive SSD and want to get accurate endgame analysis.

You can also opt for a cheap SSD.

Slow SATA SSD doesn't help much, you gain from 80MB/s common HDD to 160MB/s slow SSD. You gain quite a bit in latency, but once TBs are fatched loading them doesn't get much faster.
Expensive PCIe SSD is a different thing, but as I said they are expensive.

[M ANSARI]

I wonder if an efficient algo could be written in an engine to quickly decide which bases to look at and which bases to totally avoid on the fly. It would probably reduce any speed hit for the engine.