Complete 7-piece tablebases are out!

[syzygy]

I don't understand your post. You are saying you would not have any TB entries for KPPPKPP?

It is possible to create tablebases with "incomplete" knowledge, basically by setting positions that need to probe unavailable subtables as "don't know". Hgm claims that the positions in KPPPKPP that will have "don't know" as value will be relatively few and uninteresting.

See here for a long story.

[jshriver]

It is possible to create tablebases with "incomplete" knowledge, basically by setting positions that need to probe unavailable subtables as "don't know". Hgm claims that the positions in KPPPKPP that will have "don't know" as value will be relatively few and uninteresting.

See here for a long story.

Was a great read thanks for posting this, and from someone I greatly respect.

[hgm]

@ hgm. I was referring to reducing the size of TB and not calculating them.

Let's say I download the entire 7 piece TB somehow. I remove all the entries that are considered draw. Then I compress just the win/loss on to disc(s)...so an engine looks up a position that has 7 pieces...if it can't find the position, then it implies that its a draw.

So there should be no reason to store the draw positions...

Perhaps this simplified example will make it clear:

Suppose you had a WDL bitbase for a game with only 16 positions, (numbered 0-15), and it was

WWLD.WLDD.LWDW.WWLD

(where the periods were inserted just to facilitate counting). When you would remove the draws you would be left with:

WWLW.LLWW.WWL

which you could further compress in a reversible way. When you now want to probe for the result of position number 8, how would you know where to probe?

[geots]

Mate in 549

So it is an (actual, legal) theoretical draw.

The pawn can only make 2 moves till promotion. And there can be only 4 captures (including the capture of one of the 2 white Queens). So a maximum of 50*6 = 300 moves.

So before reaching the 549 move, there should be at least a 51 move sequence without moving a pawn nor capturing... that implies draw according the 50 move rule.

That is true except if this tablebases are going to be taking into account to avoid the 50 move rule, as there exists a "know" forced mate sequence in such position.

Me, I am going to wait on the 14-piece bases to come out.


Best,

[geots]

Now that is not written in stone. There is always the possibility, based on the theory of "multiples of 4", that I might just sit back and wait on the 28 piece tablebases. (I would really like to see the bases that kick in as soon as you come out of the opening.)


gts

[M ANSARI]

Wow that is big news can't believe I missed it ... that happened much quicker than I thought would be possible. Of course now there should be an effort to reduce size to something more palatable or there should be an emphasis on making the tables more efficient. Many of the positions are easily solved with today's average hardware and good engine ... so these could be truncated out. Also making the tables into bitbases would shrink things dramatically. Would be interesting if data could be compiled to see which positions could be solved in say 1 minute of engine time on quadcore with one of the top engines ... I bet a vast majority of the positions would disappear. I still think a powerful daughter PCI card with some sort of Monte Carlo algo would be a shortcut to 7 or maybe even 8 piece TB's. For sure it would be an enhancement to present day engines.

[casaschi]

Here is the website of the author/creator

https://plus.google.com/100454521496393505718/posts

A couple of months ago they posted this:

For now, there is a web-service for accessing the tables. Currently it is used by Aquarium only (and we are still testing it), but later it will be added to other products and hopefully published.

I hoped this meant some sort of public availability of the 7-men tablebase (i.e. non restricted to aquarium), but after two months nothing happened.
Does anyone know more about these plans for public APIs and public access?

[Albert Silver]

Here is the website of the author/creator

https://plus.google.com/100454521496393505718/posts

A couple of months ago they posted this:

For now, there is a web-service for accessing the tables. Currently it is used by Aquarium only (and we are still testing it), but later it will be added to other products and hopefully published.

I hoped this meant some sort of public availability of the 7-men tablebase (i.e. non restricted to aquarium), but after two months nothing happened.
Does anyone know more about these plans for public APIs and public access?

It won't happen before 2014 at least IMHO. They are still posting all the 7-piece combinations.