Setting SF to SyzygyProbeDepth=0

[RJN]

Hi, in an attempt to get SF to probe Syzygy more aggressively, I edited the ucioptions to allow me to set it to 0, then compiled. Komodo can go to 0, which made me wonder what would happen if I did this to SF. Normally, the minimum is 1 for SF, and 0 for Komodo.

I did not really see a difference when comparing depth 1 vs 0 in tests, but nothing bad seemed to happen.

Is there another variable I must change in the code to accomplish this? Or is 1 and 0 equivalent in the way SF Syzygy works?

[syzygy]

Equivalent. That's why the minimum is set to 1...

[RJN]

OK, thanks for confirming. But is there another way to get Komodo-like TB hit rates from SF?

[Norm Pollock]

Just based on my observations and testing, I do not believe that Syzygy 3-4-5 TBs make one bit of difference for Komodo or Stockfish. These engines are so advanced that they don't need any help. I have been using Syzygy 3-4-5 in a ram disk, which is as fast or faster than an SSD, and I have tried different settings, and no matter, they just don't improve the results for either K or S.

As for Syzygy 6, it is 160x larger, and that alone cancels out any help it will give.

[Laskos]

Just based on my observations and testing, I do not believe that Syzygy 3-4-5 TBs make one bit of difference for Komodo or Stockfish. These engines are so advanced that they don't need any help. I have been using Syzygy 3-4-5 in a ram disk, which is as fast or faster than an SSD, and I have tried different settings, and no matter, they just don't improve the results for either K or S.

As for Syzygy 6, it is 160x larger, and that alone cancels out any help it will give.

Wrong. Search and browse the forum.

[Eelco de Groot]

Ronald himself keeps saying that RAM is better dedicated to hash alone under all circumstances and let the operating system figure out if it wants to store any probes in RAM (I hope this is a sufficiently correct version of what the author stated several times I can't vouch for it). I think Norm should have read that. I do not understand the statement

Syzygy it is 160 x larger and that alone cancels out any help it will give.

In the case of Stockfish evaluation for any 5 men endgame should be sufficiently good except maybe for some Rook endgames where Syzygy may help. I do not know this for a fact but that would be my guess. That leaves only 6 men endgames which are the only ones where I would want to have them in tablebases. And that only in correspondence game circumstances or analyzing a game. I do not have them, but for correspondence or analysis I understand why you would want to have the Syzygys installed. So what if it is larger, as long as you have a disk available or SSD, that you can fit them on. I do not understand at all what can be the benefit of probing those more often. If you would be playing a game you would want to hit them as deep in the tree as possible, to have any effect? You can have endgame adjudication at the root, by the GUI in the case of Shredder for instance and that is 6 men endgames, if you let Shredder access Stefan's website. This will not bring any elo because probing at the root does not give you any advantage. So why probe more often than the default depth? It does not make any sense. Not that it is very important, I'm sure Rob has a reason.

Eelco

[Eelco de Groot]

It does not make any sense.

Should add: unless increasing the probing frequency will increase probes near the endleaves as well. Then I understand why Rob wants to do this.

[Norm Pollock]

I can see the use of tb's for analysis for correspondence chess for example, but as for game playing, the only improvement I see that tb's gives (for Stockfish and Komodo) is that it spots tb draws and tb wins quicker. For example K+R+B vs k+r, in a harmless situation with the Kings away from the rim, Stockfish without tb's will evaluate +0.40, and with tb's +0.00. The end result will still be a draw.

The 160x reference was to the total file size of the 6-man Syzygy tb's (160G) compared to the 3-4-5 Syzygy tb's (1G).

[Jesse Gersenson]

The 160x reference was to the total file size of the 6-man Syzygy tb's (160G) compared to the 3-4-5 Syzygy tb's (1G).

Is there a projected size for 7-man Syzygy tb's?

[whereagles]

You may know of this already, but the 7-men hosted at Lomosonov are like 120 TB, which is around ~1000x more than 6-men. Not sure if it's compressed, though.

8-men would be like a whole wall full with 3 TB hard drives.. kinda heavy (in kg)