7-men Syzygy attempt

[Sesse]

I'm not seeking donations at this point; in particular, it would seem inappropriate as long as others are donating time on really expensive hardware for free (and there is already the Lichess mirror).

Coincidentally, I see tablebase.lichess.ovh now points to a nice 7-man position

[main line]

We need the most important syzygy in first time. After we can put others important syzygy. But we don't need useless 7 men syzygy like QQQQKvPK or so.
SSD for the most important 7 men syzygy is about 2 Tb

[syzygy]

We need the most important syzygy in first time. After we can put others important syzygy. But we don't need useless 7 men syzygy like QQQQKvPK or so.
SSD for the most important 7 men syzygy is about 2 Tb

Please come back when you have something to contribute (and not earlier).

[Ozymandias]

I don't know if anyone ever tried to quantify how many Elo each level of tablebases help?

3 and 4-men probably add close to nothing, among the top 3 engines. 5-men has been cheap to enable form the get go, so the only thing I ever tested was 6-men against 5. This was years ago, before investing in fast NVMe SSDs I wanted to know if it was worth it. It fluctuates a lot depending on the engine, but I recall H4 tactical version, at the very low depth of 10, to give a very sizeable increase in strength. I don't expect results to hold for H6.

Under fixed depth testing conditions, it's my believe that more time equals smaller gaps; I never really tested it, but seeing who asks I'll give it a shot.

Engine tested: asmFish 9 (April 10th version with correct default contempt value)
Parameters: everything at default except for SyzygyProbeLimit, which was given the values 0, 5 and 6
Openings: Hert500
Book: 46 plies deep, very wide book (more than 10 million positions at ply 36)
6-men Syzygy adjudication
Resign at cp=300
Draw if score of 0 found for 5 moves, starting at the 25th
Ordo calculations done with version 1.2.4, using the -W and -s 1000 switches

2 round robins played at depth 12 and 15, plus a connecting match between the strongest depth 12 and the weakest depth 15 engine.

Code: Select all

   # PLAYER    :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)     W     D     L  D(%)  OppAvg  OppErr  OppN  OppDiv
   1 6a9d15    :  2454.9   14.4  1045.0    2000    52      99   469  1152   379    58  2438.9    13.0     2     2.0
   2 5a9d15    :  2440.9   14.6   986.0    2000    49      73   407  1158   435    58  2445.9    12.9     2     2.0
   3 0a9d15    :  2436.8   11.4  1789.5    3000    60     100  1070  1439   491    48  2354.3    13.5     3     3.0
   4 6a9d12    :  2167.3   11.6  1228.0    3000    41      97   620  1216  1164    41  2245.7    13.5     3     3.0
   5 5a9d12    :  2155.6   14.4   999.0    2000    50      95   529   940   531    47  2155.9    13.1     2     2.0
   6 0a9d12    :  2144.5   14.6   952.5    2000    48     ---   486   933   581    47  2161.4    13.0     2     2.0

White advantage = 41.83 +/- 3.06
Draw rate (equal opponents) = 50.00 % +/- 0.00

[syzygy]

Testing TBs at fixed depth will not give a complete picture. On the one hand, they take time to probe. On the other hand, they allow the search to go deeper. By fixing the depth you eliminate these two important components from the equation.

[Ozymandias]

Testing TBs at fixed depth will not give a complete picture. On the one hand, they take time to probe. On the other hand, they allow the search to go deeper. By fixing the depth you eliminate these two important components from the equation.

What do you mean by "time to probe"? CPU throttle? There isn't.

As for allowing the search to go deeper... well, yes, under regular TCs engines will search (much) deeper, but that only masks the effect TBs really have. As the test shows, more depth equals less of a gap, if you increase said depth you'll hardly register any differences at all. Just because it's the more chosen time control, it doesn't mean you should jump onboard blindly.

[syzygy]

Testing TBs at fixed depth will not give a complete picture. On the one hand, they take time to probe. On the other hand, they allow the search to go deeper. By fixing the depth you eliminate these two important components from the equation.

What do you mean by "time to probe"? CPU throttle? There isn't.

Time to probe is simply time to probe. It takes many cpu cycles to probe a position, and data may have to be read from SSD.

As for allowing the search to go deeper... well, yes, under regular TCs engines will search (much) deeper, but that only masks the effect TBs really have.

It masks their effect? On the contrary, the ability to search deeper is one of the main benefits of using TBs.

As the test shows, more depth equals less of a gap, if you increase said depth you'll hardly register any differences at all.

No, your test artificially limits the depth reached by the engine playing with TBs. At the same time, it artificially removes the speed penalty of probing.

Just because it's the more chosen time control, it doesn't mean you should jump onboard blindly.

Playing with fixed depth is simply not a time control at all.

[syzygy]

As for allowing the search to go deeper... well, yes, under regular TCs engines will search (much) deeper, but that only masks the effect TBs really have.

It masks their effect? On the contrary, the ability to search deeper is one of the main benefits of using TBs.

OK, what you meant is that playing with a time control as opposed to fixed depth lets both engines search deeper in endgame positions. That is indeed true and fixing depth indeed artificially lowers the level of endgame play of both engines.

But what I way saying is that searching with TBs allows an engine to search late middlegame and endgame positions deeper than without TBs. This is because each probe cuts a branch of the tree, so the tree becomes smaller. Fixing depth removes this benefit.

[main line]

Please come back when you have something to contribute (and not earlier).

Oh boy. Please look at this http://tablebase.lichess.ovh/tables/standard/7/
What if some child come here and sees what you are doing, what will you answer him???
What is this KQQQQQvK.rtbw and this KQQQQQvK.rtbz???
I don't know how to set 5 white queens on the board and black king still isn't checkmated.
What is this KRNNNNvK.rtbz or this KRBBBBvK.rtbz? Where did you find chessboard with 4 bishops and 4 knights ???
Please delete all those tablebase and stay only these:

KBBNNvKQ KQBBvKQB KQBBvKQN KQBBvKQQ KQBBvKQR KQBNvKQB KQBNvKQN KQBNvKQQ KQBNvKQR KQNNvKQB KQNNvKQN KQNNvKQQ KQNNvKQR KQQBvKQQ KQQNvKQQ KQRBvKQQ KQRBvKQR KQRNvKQQ KQRNvKQR KQRRvKQQ KRBBNvKQ KRBNNvKQ KRBNvKQN KRRBvKQB KRRBvKQN KRRBvKQR

These tablebases are 2-3% of all others useless bases. It means we need maximum 2 Tb for all useful 3-4-5-6-7 men.

[syzygy]

Please come back when you have something to contribute (and not earlier).

Oh boy. Please look at this http://tablebase.lichess.ovh/tables/standard/7/
What if some child come here and sees what you are doing, what will you answer him???

A child already came here.

What is this KQQQQQvK.rtbw and this KQQQQQvK.rtbz???
I don't know how to set 5 white queens on the board and black king still isn't checkmated.

Why do you feel the need to expose your ignorance?

Feel free to write your own generator.