It's easy for me to locate 6-piece tablebases online for free (e.g. lichess), and I know 7-piece tablebases exist (cf. Lomonosov tablebase), but I don't know of any easy way to access them.

However, I'm not really interested in all the weird positions like K4B vs KQ that a ≥7-piece tablebase would "honestly" have to include to be called as such. Haven't there been any restricted tablebases generated that only bother to include useful positions, such as e.g. KR vs. KR with more than a total of two pawns? Or is such a thing still technically infeasible?

(I realize that in theory, such a tablebase may not be 100% perfect, since calculations for common positions may be inextricably linked to those for uncommon ones; I would hope this issue could be circumvented with a minimum of resources spent brute forcing the uncommon positions and a minimum loss of accuracy for the common positions, but I could be wrong)

  • 1
    Interesting question, to which I don't know the answer. BTW, the 7 piece Lomonosov tablebase you can access for free through the android app. The Lomonosov tablebase does not include positions with 6 pieces against a lone king (because they are not interesting), so in some sense it is an example of what you are looking for. Commented Jan 26, 2017 at 19:13

4 Answers 4


I understand where you are getting from. Unfortunately, it doesn't quite work like that. To generate a 7-man KRPPvKRP you would also need 7-man endgames for the promotion. You would need:

  • .......

Can you see, it's really not that easy to generate an endgame with a pawn, because that can transpose to another endgame with the same number of pieces.

We do have technology to generate Syzygy 7-piece, but you'd need a super-computer or a high-performance-computing to do it. Google can definitely do it if they want to. So far, nobody wants to cover the costs.


As of August 2018 a complete set of 7-piece Syzygy endgame tablebases is available


I'm not expert in this field but my opinion is that even if you managed to create selected tablebases of 10% size of full package, what would be the point? I understand if you made 0.0001% size that you could be thinking about adding some pieces count but I highly doubt yo can eliminate that much and still call it tablebase. I would expect size of database for every next piece added to go extremely bigger so I think there is no point in doing this. How more precise would heavily incomplete tablebase be compared to engine? I think engine itself is way too precise these days.


Creating those tablebases requires a lot of resources (CPU time, memory, storage). The Lomonossov tables were built in the course of months and require >100TByte of storage.

For 'interesting' tables with pawns, you need to compute all tables which can result from pawn promotions first. So it requires quite some effort to create even a single table.

OTOH, you can be pretty sure that engines equipped with 6-man TBs are sufficent to analyze 7-man positions well enough for all practical needs.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.