What are the most important 6 man TBs to have

[JAMerolle]

I have KQKPPP
KRKPPP

any other tips would be great , I dont want them all just for space sake but i do want the most critical ones.

Any sugestions would be great

[LiquidNitrogenOverclocker]

I would say, minimally, get these:

KPPKPP (great for king and pawn endings, of course)
KQPKPP (for when one side just barely outpromotes the other)

KQPKQP (equal material with 1 pawn each)
KRPKRP
KBPKBP
KNPKNP

KQPKRP (major vs. major with 1 pawn each)
KBPKNP (minor vs. minor with 1 pawn each)

Whether you want this one (sometimes the knights win) is a coin toss:

KNNKPP

Just my opinion, of course.

[JAMerolle]

Really I should ask which 6 man TBs are needed that engines would slip into draws without TB help

[LiquidNitrogenOverclocker]

Really I should ask which 6 man TBs are needed that engines would slip into draws without TB help

Take the longest win in Queen + Pawn vs. Queen + Pawn. Query the tablebase as you make a subtle change to the position, such as moving 1 piece to a different square close by. Find one position where the change resulted in the "win in 117" becoming a draw.

Now, feed both of them to your program, but WITHOUT the tablebases on.

It would be almost impossible for the program to distinguish between the two, even after a long search. In one case, the win just barely foils the attempt at perpetual check by the loser. In the other case, it must be a perpetual check, or some mutual annihilation that can't be avoided.

Based on that explanation, you tell me which tablebases are most valuable.

[JAMerolle]

thanks for the tip what would be your next 4 sugestions

[kasinp]

I have KQKPPP
KRKPPP

any other tips would be great , I dont want them all just for space sake but i do want the most critical ones.

Any sugestions would be great

Here is the top 50 6-man tablebases by probability of occurrence based on approx. 200,000 computer-computer games (SSDF, CCRL, etc.).

198,884 100.000%

1 krppkr 27,902 14.029%
2 krpkrp 25,779 12.962%
3 kppkpp 19,150 9.629%
4 kpppkp 8,474 4.261%
5 kbpkpp 7,043 3.541%
6 krpkpp 6,966 3.503%
7 knpkpp 6,487 3.262%
8 kbpknp 5,389 2.710%
9 krpkbp 4,648 2.337%
10 kqpkpp 4,630 2.328%
11 kbppkb 3,680 1.850%
12 krppkp 3,529 1.774%
13 knppkn 3,528 1.774%
14 kbppkp 3,499 1.759%
15 knppkp 3,098 1.558%
16 krpknp 3,095 1.556%
17 knppkb 3,083 1.550%
18 knpknp 2,910 1.463%
19 kqpkqp 2,823 1.419%
20 kbpkbp 2,525 1.270%
21 kbppkn 2,514 1.264%
22 kqppkq 2,302 1.157%
23 kqppkp 2,240 1.126%
24 krbpkr 2,181 1.097%
25 kpppkr 2,168 1.090%
26 krbkrp 2,017 1.014%
27 kqpkrp 1,909 0.960%
28 krnpkr 1,875 0.943%
29 kbppkr 1,644 0.827%
30 krnkrp 1,555 0.782%
31 kqrkrp 1,521 0.765%
32 kpppkn 1,243 0.625%
33 knppkr 1,208 0.607%
34 kpppkb 1,124 0.565%
35 kpppkq 1,027 0.516%
36 krppkb 805 0.405%
37 krppkq 789 0.397%
38 krpppk 719 0.362%
39 kqrpkr 670 0.337%
40 krppkn 651 0.327%
41 kqpknp 549 0.276%
42 kqpkbp 531 0.267%
43 kbnpkr 494 0.248%
44 kqpppk 479 0.241%
45 kppppk 456 0.229%
46 kbnpkb 424 0.213%
47 kbpppk 416 0.209%
48 krrkrp 400 0.201%
49 kqbpkp 396 0.199%
50 kqrpkp 381 0.192%

PK

[govert]

Interesting. How were these figures calculated?

[LiquidNitrogenOverclocker]

I wonder how many of these endgames were conversion due to extreme time pressure? Some of them seem to be unlikely to even be continued by human players.

I can see why RPP vs. R might be continued, in cases where there are pawns in the f- and h- files (since this can be drawn in practice) and the same for BPP vs. B.

[playjunior]

Roughly 50% of chess endgames are rook endgames.

Computers are notoriously bad at those, so getting them is a no-brainer.

[JAMerolle]

Roughly 50% of chess endgames are rook endgames.

Computers are notoriously bad at those, so getting them is a no-brainer.[/

quote]


Thanks for the imput but todays computers handle rook end games very well. This was true in the 1990s but no longer.