7 Man Syzygy and SSD

[MikeB]

Went though many of the decisive games - so far only found that had an impact on play, but it did not impact final result.

[d]7q/2R4P/6k1/p7/Pp6/1K6/8/8 b - - 0 1

With 7 man syzygy, it plays Qxh7 immediately, with 6 man syzygy, it will play Qxh7 at depth 37 in about 11 seconds. With no egtb, it will be quite a while before it plays Qxh7.

It will eventually announce a mate-in-25

Code: Select all

 #-25 Qxh7 Rxh7 Kxh7 Kc2 Kg6 Kd3 Kf5 Kd4 Kf4 Kd3 Kf3 Kd2 Ke4 Kc2 Kd4 Kb3 Kd3 Kb2 Kc4 Ka2 Kc3 Kb1 Kb3 Kc1 Ka2 Kd2 b3 Ke3 b2 Kd4 Kb3 Kc5 b1=Q Kd5 Qf5+ Kd6 Qe4 Kc7 Qd5 Kb6 Kb4 Ka6 Kc5 Ka7 Qd7+ Ka8 Kb6 Kb8 Qb7# 

In the game 6 man syzygy McCain played Rc6+, which shorten the mate-in-25 to mate-in-15. This is the more typical type of artifacts one will see by going up to a higher man egtb - better play , but minimal Elo impact.

[Paloma]

Hi MikeB,

Any news about McCain release ?

[mbabigian]

I don't seem to be able to find a list of MD5 hashes for 7 piece tables. I'd like to verify my downloads are not corrupt. Could someone point me to a list of hashes?

Thanks
Mike

[duncan]

Went though many of the decisive games - so far only found that had an impact on play, but it did not impact final result.

[d]7q/2R4P/6k1/p7/Pp6/1K6/8/8 b - - 0 1

With 7 man syzygy, it plays Qxh7 immediately, with 6 man syzygy, it will play Qxh7 at depth 37 in about 11 seconds. With no egtb, it will be quite a while before it plays Qxh7.

It will eventually announce a mate-in-25

Code: Select all

 #-25 Qxh7 Rxh7 Kxh7 Kc2 Kg6 Kd3 Kf5 Kd4 Kf4 Kd3 Kf3 Kd2 Ke4 Kc2 Kd4 Kb3 Kd3 Kb2 Kc4 Ka2 Kc3 Kb1 Kb3 Kc1 Ka2 Kd2 b3 Ke3 b2 Kd4 Kb3 Kc5 b1=Q Kd5 Qf5+ Kd6 Qe4 Kc7 Qd5 Kb6 Kb4 Ka6 Kc5 Ka7 Qd7+ Ka8 Kb6 Kb8 Qb7# 

In the game 6 man syzygy McCain played Rc6+, which shorten the mate-in-25 to mate-in-15. This is the more typical type of artifacts one will see by going up to a higher man egtb - better play , but minimal Elo impact.

Thanks for that.

[duncan]

Going back to the 4 man EGTB, it probably gain 5 or 6 Elo compared to none, going from 4 to 5 Man EGTB , perhaps gain 2 or 3 Elo, going from 5 man to 6 man EGTB maybe gain another 1-2 Elo and like wise with going from 6 to 7 man EGTB, another 1-2 Elo. And alway less at longer TC.

As today's software is smarter, would you expect to get the same elo difference today.
e.g you say it probably gained 5 or 6 Elo compared to none, going from 4 to 5 Man EGTB . Do you know if it would be less than that for for latest stockfish ?

[MikeB]

Going back to the 4 man EGTB, it probably gain 5 or 6 Elo compared to none, going from 4 to 5 Man EGTB , perhaps gain 2 or 3 Elo, going from 5 man to 6 man EGTB maybe gain another 1-2 Elo and like wise with going from 6 to 7 man EGTB, another 1-2 Elo. And alway less at longer TC.

As today's software is smarter, would you expect to get the same elo difference today.
e.g you say it probably gained 5 or 6 Elo compared to none, going from 4 to 5 Man EGTB . Do you know if it would be less than that for for latest stockfish ?

Generally speaking, the weaker engines gain the most and the strongest engines gain the least from these type of enhancements. Function of diminishing returns. 5 years ago for Stockfish, we would see many patches worth 5 Elo or more, today, a 5 Elo patch is a rare bird indeed.

[duncan]

Going back to the 4 man EGTB, it probably gain 5 or 6 Elo compared to none, going from 4 to 5 Man EGTB , perhaps gain 2 or 3 Elo, going from 5 man to 6 man EGTB maybe gain another 1-2 Elo and like wise with going from 6 to 7 man EGTB, another 1-2 Elo. And alway less at longer TC.

As today's software is smarter, would you expect to get the same elo difference today.
e.g you say it probably gained 5 or 6 Elo compared to none, going from 4 to 5 Man EGTB . Do you know if it would be less than that for for latest stockfish ?

Generally speaking, the weaker engines gain the most and the strongest engines gain the least from these type of enhancements. Function of diminishing returns. 5 years ago for Stockfish, we would see many patches worth 5 Elo or more, today, a 5 Elo patch is a rare bird indeed.

Do you have an opinion if all the pieces were in Ram, would there be more of an elo gain than currently ?

[Sesse]

I don't seem to be able to find a list of MD5 hashes for 7 piece tables. I'd like to verify my downloads are not corrupt. Could someone point me to a list of hashes?

Thanks
Mike

No MD5, but something better:

• http://tablebase.sesse.net/syzygy/7-DTZ/SHA256SUMS
• http://tablebase.sesse.net/syzygy/7-DTZ/checksums.txt
• http://tablebase.sesse.net/syzygy/7-WDL/SHA256SUMS
• http://tablebase.sesse.net/syzygy/7-WDL/checksums.txt

Use the tbcheck checksums if you can (really, it doesn't even need the checksums file). Use the SHA-256 checksums if you can't get tbcheck, or don't trust it for whatever strange reason.

[MikeB]

I don't seem to be able to find a list of MD5 hashes for 7 piece tables. I'd like to verify my downloads are not corrupt. Could someone point me to a list of hashes?

Thanks
Mike

SHA256

[Jouni]

MikeB your example position is not good. ALL engines I try play -Qxh7 immediately with only 5 piece syzygy and Fritz even with 4 piece Nalimov! BTW I have found, that testing syzygy benefit (self test) needs 5000 games minimum.