83rd Amateur Series Division 6

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Graham Banks
Posts: 34830
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: 83rd Amateur Series Division 6

Post by Graham Banks » Sat Jan 09, 2021 9:22 pm

83RD AMATEUR SERIES (Division 6)

Intel i5 Quad
ChessGUI
256mb hash each
3-4-5 piece tablebases
Ponder off
ChessGUI.cgb book
40 moves in 25 minutes repeating (adapted for the CCRL)
4 cycles 44 rounds
All engines 64-bit where available
The top engine will gain automatic promotion and the bottom engine will be relegated. Any other changes of division will be at my discretion and may not necessarily involve playoff matches. This will be the final series that Arminius is eligible to play in, until a new version is released.
Standings after Round 36

22.5 - Velvet 1.1.0 64-bit
22.0 - ECE 20.1 64-bit
21.0 - Mayhem 2.5 64-bit
20.5 - OliThink 5.9.2 64-bit
19.0 - K2 0.95
18.0 - Arminius 2018-12-23 64-bit
17.0 - GreKo 2020.03 64-bit
16.5 - Drosophila 1.6 64-bit
15.5 - Shallow 2020-09-09 64-bit
14.5 - DanaSah 8.30
13.5 - Fridolin 3.10 64-bit
13.0 - Delocto 0.61n 64-bit


Games and standings will be available after every six rounds from the following link:
http://kirill-kryukov.com/chess/discuss ... =7&t=12609

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://kirill-kryukov.com/chess/discuss ... p?id=42959
Host - GrahamCCRL.dyndns.org Port - 16001

Linux users can use Livius:
https://github.com/kmar/livius

There is also a Livius windows version.
It has live pv boards as a nice addition.
http://www.crabaware.com/livius/
gbanksnz at gmail.com

User avatar
Graham Banks
Posts: 34830
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: 83rd Amateur Series Division 6

Post by Graham Banks » Tue Jan 12, 2021 12:42 am

83RD AMATEUR SERIES (Division 6)

Intel i5 Quad
ChessGUI
256mb hash each
3-4-5 piece tablebases
Ponder off
ChessGUI.cgb book
40 moves in 25 minutes repeating (adapted for the CCRL)
4 cycles 44 rounds
All engines 64-bit where available
The top engine will gain automatic promotion and the bottom engine will be relegated. Any other changes of division will be at my discretion and may not necessarily involve playoff matches. This will be the final series that Arminius is eligible to play in, until a new version is released.
Final Standings

27.5 - OliThink 5.9.2 64-bit
27.0 - Velvet 1.1.0 64-bit
26.0 - ECE 20.1 64-bit
25.5 - Mayhem 2.5 64-bit
23.0 - K2 0.95
22.0 - Arminius 2018-12-23 64-bit
21.5 - GreKo 2020.03 64-bit
20.0 - Fridolin 3.10 64-bit
19.5 - Shallow 2020-09-09 64-bit
19.5 - Drosophila 1.6 64-bit
17.0 - DanaSah 8.30
15.5 - Delocto 0.61n 64-bit


The complete tournament pgn (zipped) can be downloaded here:
http://kirill-kryukov.com/chess/discuss ... p?id=47666
gbanksnz at gmail.com

User avatar
Graham Banks
Posts: 34830
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: 83rd Amateur Series Division 6

Post by Graham Banks » Tue Jan 12, 2021 12:44 am

Congratulations to Oliver (OliThink author) for winning ths division and gaining promotion for the next series. :)

Thanks to all who took an interest in the tournament, including those who followed the broadcast.
gbanksnz at gmail.com

Guenther
Posts: 3662
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: 83rd Amateur Series Division 6 - OliThink 5.92

Post by Guenther » Wed Jan 13, 2021 1:11 pm

OliverBr wrote:
Fri Jan 08, 2021 9:39 am
Graham Banks wrote:
Thu Jan 07, 2021 10:30 pm
Small sample size, plus the unpredictability that tournaments can throw up.
OliThink 5.9.2 64-bit is performing around 2750 Elo in CCRL after a small number of games, but the gauntlet I'm starting today will sort out how accurate that rating is.
Thank you so much, Graham! You are such a huge support for the chess programmer community!

You are right about the small sample size, but there is a pattern that I observe since quite some time in other tourneys, too.

Actually I am running a Gauntlet, too, on my 32-core Linux server with 32 concurrent games. It's the same time control as yours, 25 minutes for 40 games and ponder off.
Differences: I am using 3moves book with thousands of opening, no tablebases at all, machines use their default hash size.

So there are not that many games, the error is still quite large. It's running for 12 hours and this is an intermediate result :

Code: Select all

   # PLAYER            :  RATING  ERROR  POINTS  PLAYED   (%)    W    D    L  D(%)  CFS(%)
   1 OliThink 5.9.2    :       0   ----   145.0     198  73.2  119   52   27  26.3     100
   2 Delocto 0.6       :    -121     75    22.5      66  34.1   12   21   33  31.8      93
   3 K2 v.0.96dev      :    -202     77    16.5      66  25.0    8   17   41  25.8      75
   4 Drosophila 1.6    :    -241     86    14.0      66  21.2    7   14   45  21.2     ---

White advantage = 80.18 +/- 24.87
Draw rate (equal opponents) = 33.23 % +/- 4.29
Of course we have to consider the fact that OliThink has been developed on that very Linux server, so it's optimizied for that.
Other than that, there is notable difference in terms of performance which I cannot explain.
With 43 games so far in CCRL (long time) the error bar is extremely large as already stated.
Here is a game when OliThink never realizes it will lose and long insists on being near draw.

I let it crunch here for a very very long time in WB (no popcount version) and I already thought it is somehow stuck
until it popped out depth 35! Look at the search explosion. Also it still had never considered the GreKo answer
to Qa1. It seems the move is sacrificed by some reduction?
(On top of this, this was done after some back and forward analysing for some moves before)

Code: Select all

dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 35	  0.00 	2.52G	26:32.88	g3g4 c3c2 a7b8 d5g2 h2g2 c2c1q b8f8 f6g6 f8g8 g6f6 g8f8 
 34	  0.00 	308.6M	2:57.45	a7a1 f6g6 a1a4 c3c2 g2c2 d4c2 a4c2 d5d4 c2a2 g6f6 h2h1 d4c3 h1h2 c3d4 
 33	  0.00 	179.8M	1:39.42	a7a1 f6g6 a1a4 c3c2 a4e8 g6f6 e8f8 f6g6 f8g8 g6f6 g8f8 
 32	  0.00 	121.3M	1:06.53	a7a1 f6g6 a1a4 c3c2 g2c2 d4f3 h2g2 f3e1 g2f1 e1c2 a4e8 g6f6 e8f8 f6g6 f8e8 
 31	  0.00 	63.8M  	0:34.91	a7a1 f6g6 a1a4 d4f3 h2h1 c3c2 a4c2 f3e1 c2a2 d5e4 h1g1 e1g2 a2g2 e4e1 g1h2 e1d1 g2f2 g6f6 f2e3 d1c2 h2g1 c2b1 g1f2 b1a2 f2g1 a2b1 
 30	  0.00 	54.2M  	0:29.55	a7a1 f6g6 a1a4 d4f3 h2h1 c3c2 a4c2 f3e1 c2a2 d5e4 h1g1 e1g2 a2g2 e4e1 g1h2 h6h5 g2g1 e1e2 h2h1 e2f3 h1h2 f3e2 
 29	  0.00 	39.3M  	0:21.32	a7a1 f6g6 a1a4 d4f3 h2h1 c3c2 a4c2 f3e1 c2a2 d5e4 h1g1 e1g2 a2g2 e4e1 g1h2 h6h5 g2g1 e1e2 h2h1 e2f3 h1h2 f3e2 
 28	  0.00 	30.1M  	0:16.27	a7a1 f6g6 a1a4 d4f3 h2h1 c3c2 a4c2 f3e1 c2a2 d5e4 h1g1 e1g2 a2g2 e4e1 g1h2 h6h5 g2g1 e1e2 h2h1 e2f3 h1h2 f3e2 
 27	  0.00 	22.9M  	0:12.32	a7a1 f6g6 a1a4 d4f3 h2h1 c3c2 a4c2 f3e1 c2a2 d5e4 h1g1 e1g2 a2g2 e4e1 g1h2 h6h5 g2g1 e1e2 g1g2 e2d1 g2b2 d1e1 b2d4 e1e2 h2g1 e2e1 g1g2 e1e2 g2h1 e2e1 h1h2 
 26	  0.00 	15.1M  	0:08.17	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 25	  0.00 	10.6M  	0:05.77	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 24	  0.00 	8.07M  	0:04.41	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 23	  0.00 	5.37M  	0:02.96	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 22	  0.00 	3.79M  	0:02.11	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 21	  0.00 	2.28M  	0:01.29	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 20	  0.00 	1.36M  	0:00.81	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 19	  0.00 	921028	0:00.56	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 18	  0.00 	559661	0:00.33	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 17	  0.00 	392830	0:00.22	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 16	  0.00 	269502	0:00.16	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 15	  0.00 	175579	0:00.11	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 14	  0.00 	121019	0:00.08	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 13	  0.00 	48440  	0:00.03	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 12	  0.00 	38798  	0:00.03	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 11	  0.00 	29579  	0:00.01	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
 10	  0.00 	21868  	0:00.01	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
  9	  0.00 	18916  	0:00.01	a7a1 f6g6 a1a4 d4f3 h2h1 f3e1 a4e8 g6g7 e8e7 g7g6 e7e8 
  8	  0.00 	7229    	0:00.01	g3g4 c3c2 a7c7 d5e4 c7d8 f6f7 d8c7 f7f6 
  7	  0.00 	6127    	0:00.00	g3g4 c3c2 a7c7 d5e4 c7d8 f6f7 d8d7 f7f6 d7d8 
  6	+0.50 	2875    	0:00.00	g3g4 c3c2 a7c7 e6e5 g2f2 d4f3 f2f3 
  5	+0.31 	1154    	0:00.00	g3g4 f5g4 g2g4 f6f5 g4g2 
  4	+0.38 	382      	0:00.00	g3g4 c3c2 a7c7 e6e5 
  3	+0.93 	142      	0:00.00	g3g4 c3c2 a7c7 
  2	+0.70 	22        	0:00.00	h2g1 c3c2 
  1	+0.70 	9          	0:00.00	h2g1 



Image

Guenther
Posts: 3662
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: 83rd Amateur Series Division 6

Post by Guenther » Thu Jan 14, 2021 8:17 pm

OliverBr wrote:
Thu Jan 07, 2021 9:13 pm
Graham Banks wrote:
Thu Jan 07, 2021 12:39 am
Standings after Round 30

19.0 - Velvet 1.1.0 64-bit
17.5 - ECE 20.1 64-bit
17.0 - OliThink 5.9.2 64-bit
16.5 - K2 0.95
16.0 - Mayhem 2.5 64-bit
15.0 - Arminius 2018-12-23 64-bit
15.0 - Drosophila 1.6 64-bit
13.0 - GreKo 2020.03 64-bit
13.0 - DanaSah 8.30
13.0 - Shallow 2020-09-09 64-bit
13.0 - Fridolin 3.10 64-bit
12.0 - Delocto 0.61n 64-bit
Hmmm, something is still strange.
K2 (0.95/0.96) was a main sparring partner of OliThink.
They played more millions games and OliThink 5.9.1 beats K2 quite easily with an ELO diff of about 130.

Code: Select all

delocto         uci     -5      0.6     https://github.com/moterink/Delocto
drosophila      uci     -260    1.6     https://github.com/gustafullberg/drosophila
k2              xb+uci  -130    0.96    https://github.com/serg-meus/k2
olithink5       xboard  0       5.9.1   https://github.com/olithink/OliThink
Of course I never do 25 minutes for 40 moves, because I need quickly many games.
But changing time controls never affected the results against K2.

This is a general observation that other engines do much better in those tournaments compared to my results.
What could this be?

PS: With Drosophia it's an ELO diff of about 260 and yet Drosophila has 15 points and OliThink 17...
PPS: Delocto???
Note that Delocto 0.61n was compiled solely as non-popcount version (n) already in 2020-01-01 by me
(this information is clearly stated in my download section) after I noticed a compilation of Deloctro is long overdue,
but we don't know what your compilation is exactly?

Post Reply