SYZYGY TABLEBASES LIST OF IMPORTANCE

[chri$]

Hello,can anybody can tell me the mostly used tablebases i would need instead of all of them.. i can just dopwnload a certain set that way narrow down my corrupt files for the syzygy tb's.... i tried researching online where i found a list at one time of like the top 10 or 12 tablebases u need and occur 96% of time .. and cant find them.. i think its like no more than 10 or less tablebases in 6men or 5 that u need and thats it.. rest very rarely occur...thanks.. for the help..

IM focusing on downloading just the mostly used and needed tb's for the syzygy tb's since im having difficulties with corrupt files and rather just get the ones i will use the most vs. all of them. thanks in advance guys..

[ChessDrone]

Code: Select all

+----------------------------------+--------------------------------+
| Details for Human games (3216177)| Details for Comp games (22949) |
+----+--------+----------+---------+----+--------+--------+---------+
| #  | pieces | #3216177 |     %   | #  | pieces | #22949 |     %   |
+----+--------+----------+---------+----+--------+--------+---------+
|  1 | krppkr | 27,902   | 14.029% |  1 | krpkrp | 1,148  | 11.566% |
|  2 | krpkrp | 25,779   | 12.962% |  2 | krppkr | 690    | 6.951%  |
|  3 | kppkpp | 19,15    | 9.629%  |  3 | kppkpp | 460    | 4.634%  |
|  4 | kpppkp | 8,474    | 4.261%  |  4 | kbpknp | 299    | 3.012%  |
|  5 | kbpkpp | 7,043    | 3.541%  |  5 | kbpkpp | 289    | 2.912%  |
|  6 | krpkpp | 6,966    | 3.503%  |  6 | krpkpp | 276    | 2.781%  |
|  7 | knpkpp | 6,487    | 3.262%  |  7 | kqppkp | 265    | 2.670%  |
|  8 | kbpknp | 5,389    | 2.710%  |  8 | kpppkp | 252    | 2.539%  |
|  9 | krpkbp | 4,648    | 2.337%  |  9 | kbpkbp | 238    | 2.398%  |
| 10 | kqpkpp | 4,63     | 2.328%  | 10 | krbkrp | 219    | 2.206%  |
| 11 | kbppkb | 3,68     | 1.850%  | 11 | knpkpp | 207    | 2.085%  |
| 12 | krppkp | 3,529    | 1.774%  | 12 | krppkp | 199    | 2.005%  |
| 13 | knppkn | 3,528    | 1.774%  | 13 | krpkbp | 199    | 2.005%  |
| 14 | kbppkp | 3,499    | 1.759%  | 14 | kbppkp | 189    | 1.904%  |
| 15 | knppkp | 3,098    | 1.558%  | 15 | kqpkqp | 186    | 1.874%  |
| 16 | krpknp | 3,095    | 1.556%  | 16 | kqpkpp | 186    | 1.874%  |
| 17 | knppkb | 3,083    | 1.550%  | 17 | krnkrp | 178    | 1.793%  |
| 18 | knpknp | 2,91     | 1.463%  | 18 | knpknp | 156    | 1.572%  |
| 19 | kqpkqp | 2,823    | 1.419%  | 19 | krbpkr | 132    | 1.330%  |
| 20 | kbpkbp | 2,525    | 1.270%  | 20 | knppkp | 131    | 1.320%  |
| 21 | kbppkn | 2,514    | 1.264%  | 21 | krpknp | 120    | 1.209%  |
| 22 | kqppkq | 2,302    | 1.157%  | 22 | kbppkb | 116    | 1.169%  |
| 23 | kqppkp | 2,24     | 1.126%  | 23 | kqpkrp | 108    | 1.088%  |
| 24 | krbpkr | 2,181    | 1.097%  | 24 | krnpkr | 106    | 1.068%  |
| 25 | kpppkr | 2,168    | 1.090%  | 25 | knppkb | 103    | 1.038%  |
| 26 | krbkrp | 2,017    | 1.014%  | 26 | kpppkr | 100    | 1.007%  |
| 27 | kqpkrp | 1,909    | 0.960%  | 27 | kqrpkr | 96     | 0.967%  |
| 28 | krnpkr | 1,875    | 0.943%  | 28 | kqppkq | 91     | 0.917%  |
| 29 | kbppkr | 1,644    | 0.827%  | 29 | kbppkr | 91     | 0.917%  |
| 30 | krnkrp | 1,555    | 0.782%  | 30 | knppkn | 84     | 0.846%  |
| 31 | kqrkrp | 1,521    | 0.765%  | 31 | kqrkrp | 84     | 0.846%  |
| 32 | kpppkn | 1,243    | 0.625%  | 32 | kbppkn | 83     | 0.836%  |
| 33 | knppkr | 1,208    | 0.607%  | 33 | kqbpkp | 80     | 0.806%  |
| 34 | kpppkb | 1,124    | 0.565%  | 34 | kqrpkp | 79     | 0.796%  |
| 35 | kpppkq | 1,027    | 0.516%  | 35 | krppkb | 77     | 0.776%  |
+----+--------+----------+---------+----+--------+--------+---------+

source: http://kirill-kryukov.com/chess/discussion-board/viewtopic.php?t=74

[chri$]

thank you chessdrone..

[hyatt]

One note. If you have a KXPKXX you need to at LEAST have KXQKXX, otherwise the program will refuse to promote the pawn since that leads to a position with no table base, and no mate score... I'd feel safer with all 4 promotion files since the mate-in-N (or just 'mate' for the current discussion) might depend on an under-promotion to avoid a stalemate or to provide a check as when promoting to knight.

[lucasart]

One note. If you have a KXPKXX you need to at LEAST have KXQKXX, otherwise the program will refuse to promote the pawn since that leads to a position with no table base, and no mate score... I'd feel safer with all 4 promotion files since the mate-in-N (or just 'mate' for the current discussion) might depend on an under-promotion to avoid a stalemate or to provide a check as when promoting to knight.

Indeed. More generally, whenever you have a TB you need to make sure you have all the other TB that it can transform into, or else you are likely to see some buggy transitions. For example, if you have KRNKN, make sure you also have KRKN, KNKN, KRNK, KRK, KNK.

It's probably better to have the whole lot. If you want to save space, you can download only the WDL ones. Assuming probing code is done correctly there should be no ELO value in the DTZ tables, other than making certain endings faster and cleaner (should never change the outcome of a game).

[hyatt]

I don't see how you are NOT going to change the outcome of some games if you only have w/d/l tables. Yes, if you reach a won position at the root you can exclude all non-won moves and search those. There's no guarantee you will find the win. And more specifically, you greatly increase the chances that the 50-move rule will end the game before you do. If you continue to use w/d/l tables once the game has reached 5-6 pieces, you are just about guaranteed to turn many wins into draws by oscillating back and forth between the many winning choices, without shortening the distance to mate...

knnkp is one dangerous example.

[lucasart]

I don't see how you are NOT going to change the outcome of some games if you only have w/d/l tables. Yes, if you reach a won position at the root you can exclude all non-won moves and search those. There's no guarantee you will find the win. And more specifically, you greatly increase the chances that the 50-move rule will end the game before you do. If you continue to use w/d/l tables once the game has reached 5-6 pieces, you are just about guaranteed to turn many wins into draws by oscillating back and forth between the many winning choices, without shortening the distance to mate...

knnkp is one dangerous example.

Yes, but syzygy WDL tables account for the 50 move rule. So that should solve the problem, no? In some (extremely rare) complicated wins, the path to win chosen by the engine may be very slow and laborious, but it should eventually win.

[hyatt]

I don't see how you are NOT going to change the outcome of some games if you only have w/d/l tables. Yes, if you reach a won position at the root you can exclude all non-won moves and search those. There's no guarantee you will find the win. And more specifically, you greatly increase the chances that the 50-move rule will end the game before you do. If you continue to use w/d/l tables once the game has reached 5-6 pieces, you are just about guaranteed to turn many wins into draws by oscillating back and forth between the many winning choices, without shortening the distance to mate...

knnkp is one dangerous example.

Yes, but syzygy WDL tables account for the 50 move rule. So that should solve the problem, no? In some (extremely rare) complicated wins, the path to win chosen by the engine may be very slow and laborious, but it should eventually win.

I am not convinced, no.

You have no idea which moves mates soonest so you bounce around between the choices that say "winning". But as you bounce back and forth, you have to avoid repetitions by excluding the ones that lead to a direct repetition that you can see, which forces you to go toward other positions. Only marked as "w". Not sure how the 50-move rule will work inside those. In normal DTM/DTZ (which I thought he did deal with the 50 move rule) it is pretty obvious how it would work. With DTM/DTZ you start at a known position when you make a capture to reach that position. And if you don't screw up and choose a move that leads to a longer mate, they will work. But with WDL how do you know which to play?

[marcelk]

Yes, but syzygy WDL tables account for the 50 move rule. So that should solve the problem, no? In some (extremely rare) complicated wins, the path to win chosen by the engine may be very slow and laborious, but it should eventually win.

The WDL tables assume the halfmove clock to be zero. In other words: the halfmove clock is not indexed. You need the DTZ tables to make guaranteed progress in all positions, otherwise some wins can become draws.

[ChessDrone]

I found other sources

• spreadsheet statistics (6-men): http://kirill-kryukov.com/chess/discussion-board/viewtopic.php?f=6&t=6941
• spreadsheet statistics (6-men): http://www.openingmaster.com/index.php?option=com_content&view=article&id=168&Itemid=178
• statistics (3-4-5-men) http://en.wikipedia.org/wiki/Chess_endgame#Frequency_table
• blunder statistics (3-4-5-men): http://chess-db.com/public/research/endgame_statistics.html?lan=0