Team Tournament for CCC Members

[mar]

ok, combined elo lower bound solver is here in case someone is interested: http://www.mediafire.com/download.php?6jaxy7v44q9g9b8
examples: turn test1.bat or test2.bat (takes ~10 minutes here to solve)

[mar]

best balanced using half-pairing, minimum combined elo limit:

bestvalue: 4461
best pairing:
KMTChess 1.2.1 32-bit + BikJump 2.01 (4461 combined)
Jazz 444 32-bit + Protej 0.5.7 (4457 combined)
ECE 11.01 + Jabba 1.0 (4447 combined)
Chesley r323 + Bubble 1.5 (4447 combined)
Prophet 2.0 + Clarabit 1.00 (4441 combined)
Parrot 07.07.22 + Carballo 0.5 32-bit (4438 combined)
FireFly 2.5.10 + PLP 1661571 (4435 combined)
Kurt 0.9.2 32-bit + Ice 0.1 (4435 combined)

[Sven]

Martin,

[this part would actually belong into Programming subforum ...]

4461 will work for automatic, "optimized" team selection. But what will happen in case of team selection by humans? Will the upper limit of 4461 guarantee that there will be no "unsolvable" situation, whatever choice of teams is made by humans? I don't think so, we saw a counter-example in this thread. I think you need a second rule, either a lower limit or the explicit rule that for a team size of 2, one player must be from the upper half of the ranking and one from the lower (could be generalized for team size N). Otherwise selecting a team where both players have a low rating would immediately break the system.

Any idea whether this can be solved in general? I tried it but was not successful so far ...

Sven

[mar]

Martin,

[this part would actually belong into Programming subforum ...]

4461 will work for automatic, "optimized" team selection. But what will happen in case of team selection by humans? Will the upper limit of 4461 guarantee that there will be no "unsolvable" situation, whatever choice of teams is made by humans? I don't think so, we saw a counter-example in this thread. I think you need a second rule, either a lower limit or the explicit rule that for a team size of 2, one player must be from the upper half of the ranking and one from the lower (could be generalized for team size N). Otherwise selecting a team where both players have a low rating would immediately break the system.

Any idea whether this can be solved in general? I tried it but was not successful so far ...

Sven

Hi Sven, it actually returns a lower bound and I'm 100% sure you can't go below 4461. I also added the condition that the first engine MUST come from the upper half (command line parameter -half) and it boosted the search. It now takes 20 seconds to find the lower bound for 16 input engines, which is not bad I think. I just uploaded a new version to mediafire, link remains the same.
It was just meant as a helper tool to find the lower bound so that the problem is still solvable (anything below the lower bound means 100% unsolvable). When the lower bound no longer works, it has to be reestablished. Ideally, the lower bound should be recomputed each time an engine pair is selected.
Considering the upper bound, assuming we have upper and lower half pairing, the upper bound should be simply the best engine in the upper half and the best engine in the lower half.
Btw. sorry to plague tournaments subforum but I think the problem is too specific

Martin

[mar]

sorry there is actually no upper limit of course it would be +inf. but anything above best engine in upper half + best engine in lower half has no effect and will always be solvable (I think).

Martin

[mar]

ok I just tried to look for the minimum solvable upperbound and verified by brute force search. My conclusion is that anything _below_ "combined elo of best two engines that form a valid pair" can _always_ end up being unsolvable.

Martin

[Graham Banks]

CCC TEAM TOURNAMENT

Intel Q6600 Quad Core
ChessGUI
128mb hash each
3-4-5 piece tablebases
Ponder off
HarrySchnapp.cgb book
40 moves in 32 minutes repeating (adapted for the CCRL)
2 cycles 30 rounds
All engines 32-bit 1CPU


Round 1

Jazz 444 v PLP 1661571 (1-0)
ECE 11.01 v Jabba 1.0 (1-0)
FireFly 2.5.10 v Protej 0.5.7 (draw)
KMTChess 1.2.1 32-bit v BikJump 2.01 (0-1)
Chesley r323 v Bubble 1.5 (1-0)
Parrot 07.07.22 v Carballo 0.5 32-bit (draw)
Prophet 2.0 v Clarabit 1.00 (1-0)
Kurt 0.9.2 32-bit v Ice 0.1 (draw)


Standings after Round 1

1.0 - Jazz 444
1.0 - ECE 11.01
1.0 - Chesley r323
1.0 - Prophet 2.0
1.0 - BikJump 2.01
0.5 - FireFly 2.5.10
0.5 - Parrot 07.07.22
0.5 - Kurt 0.9.2 32-bit
0.5 - Ice 0.1
0.5 - Carballo 0.5 32-bit
0.5 - Protej 0.5.7
0.0 - KMTChess 1.2.1 32-bit
0.0 - Clarabit 1.00
0.0 - Bubble 1.5
0.0 - Jabba 1.0
0.0 - PLP 1661571


Team Standings

1.5 - Sven
1.5 - Ray
1.0 - Martin
1.0 - Kirill
1.0 - Aser
1.0 - Graham
0.5 - Ben-Hur
0.5 - Sam


Round 1 pgn - http://kirill-kryukov.com/chess/discuss ... p?id=22611

[Sven]

I have created a new thread in the Programming subforum to avoid further "spamming" of this subforum with our non-tournament stuff

Sven

[Graham Banks]

CCC TEAM TOURNAMENT

Intel Q6600 Quad Core
ChessGUI
128mb hash each
3-4-5 piece tablebases
Ponder off
HarrySchnapp.cgb book
40 moves in 32 minutes repeating (adapted for the CCRL)
2 cycles 30 rounds
All engines 32-bit 1CPU


Round 2

ECE 11.01 v PLP 1661571 (1-0)
FireFly 2.5.10 v Jazz 444 (draw)
KMTChess 1.2.1 32-bit v Jabba 1.0 (1-0)
Chesley r323 v Protej 0.5.7 (draw)
Parrot 07.07.22 v BikJump 2.01 (draw)
Prophet 2.0 v Bubble 1.5 (1-0)
Kurt 0.9.2 32-bit v Carballo 0.5 32-bit (0-1)
Ice 0.1 v Clarabit 1.00 (draw)


Standings after Round 2

2.0 - Prophet 2.0
2.0 - ECE 11.01
1.5 - BikJump 2.01
1.5 - Carballo 0.5 32-bit
1.5 - Jazz 444
1.5 - Chesley r323
1.0 - Parrot 07.07.22
1.0 - FireFly 2.5.10
1.0 - Protej 0.5.7
1.0 - Ice 0.1
1.0 - KMTChess 1.2.1 32-bit
0.5 - Kurt 0.9.2 32-bit
0.5 - Clarabit 1.00
0.0 - Bubble 1.5
0.0 - Jabba 1.0
0.0 - PLP 1661571


Team Standings

2.5 - Sven
2.5 - Aser
2.5 - Ray
2.0 - Martin
2.0 - Graham
1.5 - Kirill
1.5 - Ben-Hur
1.5 - Sam


Round 2 pgn - http://kirill-kryukov.com/chess/discuss ... p?id=22613

[Graham Banks]

CCC TEAM TOURNAMENT

Intel Q6600 Quad Core
ChessGUI
128mb hash each
3-4-5 piece tablebases
Ponder off
HarrySchnapp.cgb book
40 moves in 32 minutes repeating (adapted for the CCRL)
2 cycles 30 rounds
All engines 32-bit 1CPU


Round 3

FireFly 2.5.10 v PLP 1661571 (1-0)
KMTChess 1.2.1 32-bit v ECE 11.01 (0-1)
Chesley r323 v Jazz 444 (0-1)
Parrot 07.07.22 v Jabba 1.0 (draw)
Prophet 2.0 v Protej 0.5.7 (0-1)
Kurt 0.9.2 32-bit v BikJump 2.01 (draw)
Ice 0.1 v Bubble 1.5 (draw)
Clarabit 1.00 v Carballo 0.5 32-bit (1-0)


Standings after Round 3

3.0 - ECE 11.01
2.5 - Jazz 444
2.0 - Protej 0.5.7
2.0 - FireFly 2.5.10
2.0 - BikJump 2.01
2.0 - Prophet 2.0
1.5 - Clarabit 1.00
1.5 - Parrot 07.07.22
1.5 - Carballo 0.5 32-bit
1.5 - Chesley r323
1.5 - Ice 0.1
1.0 - Kurt 0.9.2 32-bit
1.0 - KMTChess 1.2.1 32-bit
0.5 - Bubble 1.5
0.5 - Jabba 1.0
0.0 - PLP 1661571


Team Standings

4.0 - Graham
3.0 - Martin
3.0 - Sven
3.0 - Kirill
3.0 - Aser
3.0 - Sam
3.0 - Ray
2.0 - Ben-Hur


Round 3 pgn - http://kirill-kryukov.com/chess/discuss ... p?id=22620