44th Amateur Series Division 3

[Graham Banks]

44TH AMATEUR SERIES (Division 3)

Intel i5 Quad
ChessGUI
256mb hash each
3-4-5 piece tablebases
Ponder off
ChessOK2013-3.cgb book
40 moves in 25 minutes repeating (adapted for the CCRL)
4 cycles 44 rounds
All engines 64-bit 1CPU (where available)

The top engine will gain automatic promotion and the bottom engine will gain automatic relegation. Any other changes of division will be at my discretion and may not necessarily involve playoff matches.

Standings after Round 30

20.5 - DiscoCheck 5.1 64-bit
18.0 - Arasan 16.3 64-bit
17.5 - EXchess 7.18b 64-bit
17.0 - Octochess r5190 64-bit
16.5 - Cheng4 0.36a 64-bit
16.0 - Gaviota 0.86 64-bit
15.0 - GNU Chess 5.50 64-bit
14.0 - Bobcat 3.25 64-bit
13.5 - Nirvanachess 1.3 64-bit
13.5 - ProDeo 1.86
9.5 - iCE 1.0 64-bit
9.0 - Rodent 1.2 64-bit


This tournament can be followed more closely from the following link, where games will also be available for download after every two rounds:
http://kirill-kryukov.com/chess/discuss ... f=7&t=7239

Alternatively, if you install TLCV (Tom's Live Chess Viewer) on your computer, you can watch the games live move by move. You'll also be able to chat to others following the tournament in the chatroom there.
http://home.pacific.net.au/~tommyinoz/client.zip
Host - GrahamCCRL.dyndns.org Port - 16002

[Graham Banks]

44TH AMATEUR SERIES (Division 3)

Intel i5 Quad
ChessGUI
256mb hash each
3-4-5 piece tablebases
Ponder off
ChessOK2013-3.cgb book
40 moves in 25 minutes repeating (adapted for the CCRL)
4 cycles 44 rounds
All engines 64-bit 1CPU (where available)

The top engine will gain automatic promotion and the bottom engine will gain automatic relegation. Any other changes of division will be at my discretion and may not necessarily involve playoff matches.

Standings after Round 36

24.5 - DiscoCheck 5.1 64-bit
21.0 - EXchess 7.18b 64-bit
21.0 - Arasan 16.3 64-bit
20.5 - Gaviota 0.86 64-bit
19.5 - Cheng4 0.36a 64-bit
19.5 - Octochess r5190 64-bit
18.0 - GNU Chess 5.50 64-bit
16.5 - ProDeo 1.86
16.0 - Bobcat 3.25 64-bit
15.5 - Nirvanachess 1.3 64-bit
13.0 - Rodent 1.2 64-bit
11.0 - iCE 1.0 64-bit


This tournament can be followed more closely from the following link, where games will also be available for download after every two rounds:
http://kirill-kryukov.com/chess/discuss ... f=7&t=7239

Alternatively, if you install TLCV (Tom's Live Chess Viewer) on your computer, you can watch the games live move by move. You'll also be able to chat to others following the tournament in the chatroom there.
http://home.pacific.net.au/~tommyinoz/client.zip
Host - GrahamCCRL.dyndns.org Port - 16002

[Graham Banks]

44TH AMATEUR SERIES (Division 3)

Intel i5 Quad
ChessGUI
256mb hash each
3-4-5 piece tablebases
Ponder off
ChessOK2013-3.cgb book
40 moves in 25 minutes repeating (adapted for the CCRL)
4 cycles 44 rounds
All engines 64-bit 1CPU (where available)

The top engine will gain automatic promotion and the bottom engine will gain automatic relegation. Any other changes of division will be at my discretion and may not necessarily involve playoff matches.

Final Standings

30.0 - DiscoCheck 5.1 64-bit
26.0 - EXchess 7.18b 64-bit
25.5 - Arasan 16.3 64-bit
24.5 - Gaviota 0.86 64-bit
24.0 - GNU Chess 5.50 64-bit
23.0 - Octochess r5190 64-bit
21.0 - Cheng4 0.36a 64-bit
20.0 - Bobcat 3.25 64-bit
19.5 - ProDeo 1.86
19.0 - Nirvanachess 1.3 64-bit
17.0 - iCE 1.0 64-bit
14.5 - Rodent 1.2 64-bit


The complete tournament pgn (zipped) can be downloaded here:
http://kirill-kryukov.com/chess/discuss ... p?id=30917

[PK]

hard group for Rodent! I already have some improvements, but main problem is of general nature: Rodent's eval goes blind in multi-piece endgames, just after king safety is switched off. It is interesting that analyzing my own games I noticed exactly the same pattern

[lucasart]

hard group for Rodent! I already have some improvements, but main problem is of general nature: Rodent's eval goes blind in multi-piece endgames, just after king safety is switched off. It is interesting that analyzing my own games I noticed exactly the same pattern

I would not recommend "switching off" the king safety at any stage. Eval continuity is important. What I do is to compute everything (passed pawns, mobility, safety, etc.) and assign it an opening and an endgame score. Then interpolate linearly based on the material left. This is one of the many things I've learnt from Fruit.

King safety naturally fades off, as the interpolation factor goes and weighs more the endgame (where there is no safety). Also, the fewer pieces the less king attacks on average. So it fades away naturally rather than artificially and discontinuously.

I've tried to switch off the safety in late endgames as a speed gain, but never gained any elo in doing it. Generally my testing results tend to confirm that accuracy is more important than speed in NPS.

Plus, I've found that DiscoCheck is particularly vicious at unexpected king attacks in the endgame.Possible explanations: 1/ DC does a single reply extension (even when not in check) unlike most engines 2/ DC continues to account for safety.

[mar]

Thanks Graham.
I'm happy with the results even though I admit I expected a tiny bit more
I have one comment though: I have noticed that in one game cheng was +1.8 out of opening against iCE (both engines agreed on similar score).
It was very unfair (I would even say lame, no offense , maybe you should filter your books because lines like that just add noise.
Plus it will save you time as you won't have to replay such bad openings (if you do).

[Graham Banks]

Thanks Graham.
I'm happy with the results even though I admit I expected a tiny bit more
I have one comment though: I have noticed that in one game cheng was +1.8 out of opening against iCE (both engines agreed on similar score).
It was very unfair (I would even say lame, no offense , maybe you should filter your books because lines like that just add noise.
Plus it will save you time as you won't have to replay such bad openings (if you do).

Hi Martin,

I don't know how to filter the ChessGUI opening books unless I check each line individually for computer evaluations. These books are too big to do that, which is why I have to rely on ChessGUI's auto-replaying games with unfair lines or by manually ensuring that a game gets replayed when missed.

However, for the very first time, I'm having a go at constructing a book from handpicked and computer checked lines.
I have 245 lines in it now and I'm looking forward to running a tournament with it to see how good a job I've done.

Cheers,
Graham.

[lucasart]

Thanks Graham.
I'm happy with the results even though I admit I expected a tiny bit more
I have one comment though: I have noticed that in one game cheng was +1.8 out of opening against iCE (both engines agreed on similar score).
It was very unfair (I would even say lame, no offense , maybe you should filter your books because lines like that just add noise.
Plus it will save you time as you won't have to replay such bad openings (if you do).

Hi Martin,

I don't know how to filter the ChessGUI opening books unless I check each line individually for computer evaluations. These books are too big to do that, which is why I have to rely on ChessGUI's auto-replaying games with unfair lines or by manually ensuring that a game gets replayed when missed.

However, for the very first time, I'm having a go at constructing a book from handpicked and computer checked lines.
I have 245 lines in it now and I'm looking forward to running a tournament with it to see how good a job I've done.

Cheers,
Graham.

You could use this book. At least it has been filtered automatically:
http://open-chess.org/viewtopic.php?f=7&t=2522
Filter is quite tolerant though (DiscoCheck 12 ply search score within +/-0.7 for 2 moves).

[Graham Banks]

You could use this book. At least it has been filtered automatically:
http://open-chess.org/viewtopic.php?f=7&t=2522
Filter is quite tolerant though (DiscoCheck 12 ply search score within +/-0.7 for 2 moves).

Hi Lucas,

I have many ChessGUI books available and most are based on games between GMs.
However, every book I've come across (including ctg and abk books) has a few howlers. It's hard to find the perfect book!

I'm currently working with Dann Corbit on producing what I hope will be a very balanced and reliable ChessGUI book.
I've provided him with 242 lines that I've selected and he is pruning out those that give an evaluation of over +0.30 or worse than -0.20 from the exit points.
It will therefore be a small book, but hopefully a good one for engine v engine testing.

Congratulations on DiscoCheck's win in a tough field. I'm looking to run a gauntlet for this new version soon, so that it gets esrablished in our 40/40 list.

Regards,
Graham.

[lucasart]

Congratulations on DiscoCheck's win in a tough field. I'm looking to run a gauntlet for this new version soon, so that it gets esrablished in our 40/40 list.

Thanks.