On-line engine blitz tourney April

[Joost Buijs]

The monthly on-line engine blitz tourney for April will take place on:

Saturday April 16, 2:00 PM EST (Boston time), 20:00 CET (Amsterdam time)

To connect:

winboard -zp -ics -icshost nightmare-chess.nl -icshelper timeseal -fcp ENGINE.exe -fd ENGINEFOLDER -autoKibitz -keepAlive 30

(for UCI engines, add -fUCI)

People without account can submit a request for an account to admin@nightmare-chess.nl

Our tournament manager will be: Eendje (Lisebeth at ICC)

The format will be: 9 rounds Swiss 5/1

Lisebeth likes to start the tournament on top of the hour, please try to be on-line 15 minutes prior to tournament start, late joins are not supported.

[dangi12012]

Engine and Hardware Infor for all people who provided it.

Code: Select all

BelFish           [3001]  4.5  [3074] [   0] +13w +04w =03b =05w =06b +08w 
BelFish - a StockFish derivative, modified, maintained and operated by belka
 2: experimenting with evaluation function, with small techniques for speeding up code execution, with compiling options ...
 3: running commonly unattended
 4: ...
 5: ...
 6: ...
 7: SET NOESCAPE 1 !!!
 8: average ping statistics on nightmare-chess.nl: 21 ms
 9: Engine approval: "I've added computer account BelFish" (Oct. 10th 2019)
10: Thanks to Joost Buijs for maintaining this chess server!

pontifex          [2954]  4.0  [2941] [  13] +16w =09w =06b =10w =05b +13w
 1: RTX 3080
 2: Lc0 v0.29.0-rc 
 
SeerRB            [2873]  4.0  [2930] [ 118] +17w =06b =01w =09w +13b =05w
  1: Seer 2.4.0 by Connor McMonigle
 2: Run by Ray Banks. May be unattended.
 3: Book: Chosen on the day
 4: 20 threads on a Ryxen 5900X
 5: Tablebases: Syzygy on NVME SSD
 6: Hash: 4096MB 
 
Goldbar           [2870]  4.0  [2851] [  51] +12w -01b +16w =07b =09w +15b 
 Timeseal: yes

 1: Goldbar 4.50a
 2: Playing on 16x3.8GHz and 5 or 6 tablebases
 3: By Bart Goldhoorn, Zeist
 4: Thx to my friend Joost Buijs
 
ChessSystemTalX   [2863]  4.0  [2972] [ 222] +19w =07b +08w =01b =02w =03b 
 NONE
 
LczTinker         [2981]  3.5  [2875] [  10] =14b =03w =02w +15b =01w =09b 
  1: lc0-v0.29.0-rc0-windows-gpu-nvidia-cuda
 2: Net 791429
 3: Nvidia RTX 2080ti GPU
 4: 6 and some 7 piece syzygy
 5: No book
 6: Just for fun
rofChade          [2973]  3.5  [2809] [   0] +15w =05w =09b =04w =10b =11b 
 NONE
 
NightmareX        [2911]  3.5  [2812] [   0] +11w +10w -05b =13b +16w -01b
 1: Nightmare-NN v7.93 by Joost Buijs
 2: Running on a TR3-3970X @ 3.7 GHz. using 32 cores
 3: Nalimov 6 men EGTB on a NVME SSD

ArasanX           [2856]  3.5  [2880] [ 184] +20w =02b =07w =03b =04b =06w 
 Arasan, Threadripper 3970x (32 cores), 32G hash, Ubuntu x64
fics% 
WaspX             [2831]  3.5  [2750] [ 132] +18w -08b +11w =02b =07w =14b 
Timeseal: yes
 1: Wasp development version, by John Stanback
 2: Ryzen 3950x at 3.9 Ghz using 26 threads and 4G Hash
 3: Small NN for eval with 1826 inputs, one layer with 768 nodes, and 7 outputs
 4: Inputs are 2x2x7x64 PSQ and 32 material difference
 
mhouppin          [2459]  3.5  [2664] [ 257] -08b +18w -10b +19w +12b =07w 
 NONE
 
RookieMonster     [2333]  3.0  [2536] [ 260] -04b +17b -13w +14b -11w +18w 
 NONE
 
Marvin            [2748]  2.5  [2710] [  81] -01b +14w +12b =08w -03w -02b
 NONE 
 
atomNightmare     [2592]  2.5  [2507] [ 313] =06w -13b =19b -12w +20b =10w 
 NONE
 
WaDuuttie         [2576]  2.5  [2531] [   0] -07b +19w =20b -06w +18b -04w 
1: private single core 32-bit engine on Intel i7-9700 @ 3.0 Ghz


atomGoldbar       [2571]  2.0  [2556] [   0] -02b +20w -04b +18w -08b *17w
 1: Goldbar v4.50a by Bart Goldhoorn on a J3455 @ 1.5 GHz. using 1 core


Myrddin           [2334]  2.0  [2314] [   0] -03b -12w -18b +20w +19b *16b 
 1: Myrddin chess engine version 0.88a (testing version), written by John Merlino
 2: Running on an AMD Ryzen 5 @ 3.6GHz, using eight processors
 3: Myrddin uses the complete set of Gaviota tablebases
 4: Myrddin will not accept draw offers and cannot resign. No aborts, adjourns or takebacks. Autoflag is on.
fics% 
Barbarossa        [2216]  1.0  [2250] [ 118] -10b -11b +17w -16b -15w -12b 

NONE

Nemeton           [2320]  0.5  [2244] [ 136] -05b -15b =14w -11b -17w *20w 
 1: i7-4770 3.5 GHz 4 cores

atomFloyd         [2229]  0.5  [2265] [ 173] -09b -16b =15w -17b -14w *19b 
 1: Floyd v1.0 by late Marcel van Kervinck on a J3455 @ 1.5 GHz. using 1 core

If you participated it might be a good idea to check if its the current information - and if there is a possibility to script that somehow?

I think is imperative to put hardware and engine performance into context - as well as seperate authors of original engines and people just testing hardware etc. Anyway its a lot of fun.

Greetings - Daniel

[Joost Buijs]

Belfish (a Stockfish derivative) won the tournament, this was more or less expected.
Seer did pretty good like before, which is amazing since I believe he is using a floating point network that must be rather slow.
ChessSystemTalX improved a lot since last month, I wonder what Chris did.
Four engines ended with 5.5 points, LczTinker, rofChade, Goldbar and Wasp, all very close to each other.
Nightmare and Arasan both ended with 5 points and were not very lucky this time.
mhouppin (also with 5 points) improved his performance by 200 Elo points, which is pretty amazing.
pontifex (Daniel Infuehr, as it seems with LC0) was not able to win the tournament and left 3 rounds early. Please don't enter a tournament if it is not your intention to finish it.

Code: Select all

:Tourney Players: Round 9 of 9 
:
:     Name              Rating Score Perfrm Upset  Results 
:     ----------------- ------ ----- ------ ------ ------- 
:  1 +BelFish           [3001]  7.0  [3070] [   0] +11w +06w =02b =03w =04b +08w =05b +15w +10b 
:  2 +SeerRB            [2873]  6.0  [2970] [ 187] +18w =04b =01w =09w +11b =03w =08b +06w =05b 
:  3 +ChessSystemTalX   [2863]  6.0  [2960] [ 281] +20w =05b +08w =01b =15w =02b =04w =09b +14b 
:  4 +LczTinker         [2981]  5.5  [2862] [  10] =12b =02w =15w +13b =01w =09b =03b =08w +16w 
:  5 +rofChade          [2973]  5.5  [2875] [  14] +13w =03w =09b =06w =07b =10b =01w +11b =02w 
:  6 +Goldbar           [2870]  5.5  [2863] [ 135] +16w -01b +14w =05b =09w +13b +15w -02b =07w 
:  7 +WaspX             [2831]  5.5  [2800] [ 163] +19w -08b +10w =15b =05w =12b =09w +13w =06b 
:  8 +NightmareX        [2911]  5.0  [2842] [  35] +10w +07w -03b =11b +14w -01b =02w =04b =09w
:  9 +ArasanX           [2856]  5.0  [2876] [ 214] +17w =15b =05w =02b =06b =04w =07b =03w =08b 
: 10 +mhouppin          [2459]  5.0  [2654] [ 313] -08b +19w -07b +20w +16b =05w =14w +18b -01w 
: 11 -Marvin            [2748]  4.5  [2687] [  81] -01b +12w +16b =08w -02w -15b +17b -05w +20b 
: 12 +atomNightmare     [2592]  4.5  [2507] [ 313] =04w -11b =20b -16w +17b =07w +18b -14w +19b 
: 13 +WaDuuttie         [2576]  4.5  [2565] [   0] -05b +20w =17b -04w +19b -06w +16w -07b +18w 
: 14 +atomGoldbar       [2571]  4.5  [2603] [  21] -15b +17w -06b +19w -08b +18w =10b +12b -03w 
: 15 -pontifex          [2954]  4.0  [2727] [  13] +14w =09w =04b =07w =03b +11w -06b -01b -17w 
: 16 +RookieMonster     [2333]  4.0  [2521] [ 260] -06b +18b -11w +12b -10w +19w -13b +20w -04b
: 17 +atomFloyd         [2229]  3.5  [2485] [ 989] -09b -14b =13w -18b -12w +20b -11w +19w +15b 
: 18 +Myrddin           [2334]  2.0  [2241] [   0] -02b -16w -19b +17w +20b -14b -12w -10w -13b 
: 19 +Barbarossa        [2216]  1.5  [2205] [ 170] -07b -10b +18w -14b -13w -16b =20w -17b -12w 
: 20 +Nemeton           [2320]  1.0  [2172] [ 136] -03b -13b =12w -10b -18w -17w =19b -16b -11w 
:
:     Average Rating    2674.6

[Modern Times]

Yes Seer did well again ! And on what was relatively modest hardware (5900X using 20 threads) compared to some of the other participants.

Games here if anyone is interested:
http://ccrl.chessdom.com/public/Seer_Ga ... pr2022.zip

[RubiChess]

pontifex (Daniel Infuehr, as it seems with LC0) was not able to win the tournament and left 3 rounds early. Please don't enter a tournament if it is not your intention to finish it.

I am not surprised about this remembering his (dangi/pontifex) behaviour on his very first appearance at online blitz.
I will come back when this guy constantly stays away from the tourney.

Regards, Andreas

[dangi12012]

pontifex (Daniel Infuehr, as it seems with LC0) was not able to win the tournament and left 3 rounds early. Please don't enter a tournament if it is not your intention to finish it.

I am not surprised about this remembering his (dangi/pontifex) behaviour on his very first appearance at online blitz.
I will come back when this guy constantly stays away from the tourney.

Regards, Andreas

It was not my Intention to leave this one - I would not have joined otherwise. Also I am always suprised of how good all the engines perform. But some have 32 cores or 3090ti with 6man egtb etc

What I would still like to know: What is the confidence interval and standard deviation for the Elo outcome in a 9 Round Swiss tournament with 20 Partisipants? IE what percentage is luck.
But maybe its best to ask that in a math forum.

[Joost Buijs]

pontifex (Daniel Infuehr, as it seems with LC0) was not able to win the tournament and left 3 rounds early. Please don't enter a tournament if it is not your intention to finish it.

I am not surprised about this remembering his (dangi/pontifex) behaviour on his very first appearance at online blitz.
I will come back when this guy constantly stays away from the tourney.

Regards, Andreas

It was not my Intention to leave this one - I would not have joined otherwise. Also I am always suprised of how good all the engines perform. But some have 32 cores or 3090ti with 6man egtb etc

What I would still like to know: What is the confidence interval and standard deviation for the Elo outcome in a 9 Round Swiss tournament with 20 Partisipants? IE what percentage is luck.
But maybe its best to ask that in a math forum.

There were at least 2 entries with 64 cores and 3 entries with 32 cores, and a lot with 16. I assume most of them use 6 or partially 7 man EGTB.

You were probably thinking that it would be a walkover with LC0 on your RTX-3080, and to your surprise it wasn't. My guess is that this is the reason that you suddenly wanted to leave the tournament because you seemed a bit disappointed by the result you got.

You seem to be a good programmer, why don't you try to write your own engine and join the tournament with that instead of trying to beat other programmers with an unmodified Stockfish or LC0?

And yes, a nine round Swiss tournament with 20 players is like a tombola, you have to be lucky to win, but you can always look at the performance rating this gives a better indication of playing strength.

[JVMerlino]

Myrddin is already a mediocre engine at best, and it runs on these tournaments with "only" 8 cores and "only" 5-man EGTBs. With all of the super strong engines running on much better hardware, coupled with the (relatively few) people running clones and/or engines that they didn't write themselves, it is no surprise that Myrddin finished 18th out of 20 this time around. Less than a year ago, when we regularly had 30+ engines, Myrddin could occasionally finish in the middle third of the group. No longer....

Very discouraging and making me rethink whether I want to participate in these tournaments.

[Modern Times]

Myrddin is already a mediocre engine at best, and it runs on these tournaments with "only" 8 cores and "only" 5-man EGTBs.

Yes there is some impressive hardware in these tournaments. It must be quite discouraging as an author if you have a hardware disadvantage. One thing that was enjoyable though as a spectator for the old ICGA in-person world championships was in fact the huge hardware - big clusters in some cases with hundreds of cores. Must have been scary to compete against, but very interesting to follow.

[dangi12012]

Myrddin is already a mediocre engine at best, and it runs on these tournaments with "only" 8 cores and "only" 5-man EGTBs.

Yes there is some impressive hardware in these tournaments. It must be quite discouraging as an author if you have a hardware disadvantage. One thing that was enjoyable though as a spectator for the old ICGA in-person world championships was in fact the huge hardware - big clusters in some cases with hundreds of cores. Must have been scary to compete against, but very interesting to follow.

The administrator of the tournament should make it required to add all information to your winboard "finger <name>" information.
So that the rankings can be split into authors of original work vs the people who just test hardware and open source software in a tournament setting.
But yes - having 7man vs 5man egtb and CPU cores is a difference in the endgame espc. when time is low. (IE pay to win)

IMO this 5 lines should be made mandatory by Lisebeth at ICC to participate
Engine Name + Version
CPU + GPU + RAM
Hash Size / Threads
EGTB no/5/6/7
Special Settings

And afterwards the ranking is author engines and open source engines seperately.
This will helpful because first of all the ranking is fair - and second of all everyone can see how much elo difference the settings make when running open source engines. And third finally you can see exactly what you play against and its not just a blackbox player with a name.