$20,000 ICUC - a report on wins

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
Ozymandias
Posts: 978
Joined: Sun Oct 25, 2009 12:30 am

$20,000 ICUC - a report on wins

Post by Ozymandias » Wed May 24, 2017 5:08 pm

After the 2014 ICFB, I was asked to make a report about the tournament, as any other prize winner; this time, no one's asked me, but I felt it was even more relevant than three years ago. The proportion of blunders per game is very similar, but this time the draw rate is higher, so they become key. How high? That depends on the DB you look at. The official PGN has it at 87.1% (just 101 wins out of 781 games), the one used for my performance calculations, rises to 90.3% (74 out of 768) and just by taking out the six players at the bottom (*) we reach 93.4% (45/677).

While in 2014 I discarded every game without a clearcut result, this time around I’ll be changing results to draw, for almost every blunder/disconnection. I will only discard games, where losing opening lines were played, as they tell nothing about the opponent’s performance.

The data collected below, was originally compiled by Thomas Zipproth, this is the result after we compared notes:

Code: Select all

Players: 80
Theoretical number of Games 80*21/2:     840
Number of scheduled matches:             790
Number of games in pgn:                  781


1,Thomas_A_Anderson,,2648,deepkrack11,,2475,90'+30'',04/22/2017,1-0 (Forced white win)    in pgn, dc
6,Gamelover,,2462,Fahad,,2450,90'+30'',04/29/2017,1-0 (Forced white win)                  in pgn, dc
9,Thomas_A_Anderson,,2648,fati64,,2510,90'+30'',05/03/2017,1-0 (Forced white win)         in pgn, dc in lost position, regular win, result kept
10,FireFather,,2357,Wolle,,2508,90'+30'',05/05/2017,1-0 (Forced white win)                in pgn, regular win, result kept
11,Regina_H.Milch,,2658,Katzenmaier,,2483,90'+30'',05/06/2017,1-0 (Forced white win)      in pgn, regular win, result kept
14,Coco69,,2375,MrScience,,2400,90'+30'',05/10/2017,1-0 (Forced white win)                ----
14,maximus,,2676,ArKheiN,,2400,90'+30'',05/10/2017,1-0 (Forced white win)                 in pgn, dc
15,Phyrros,,2450,Dinkelberger,,2400,90'+30'',05/12/2017,1-0 (Forced white win)            ----
17,KingSlayer,,2524,FireFather,,2357,90'+30'',05/14/2017,1-0 (Forced white win)           in pgn, regular win, result kept
17,Longin,,2461,KhalidOmar,,2450,90'+30'',05/14/2017,1-0 (Forced white win)               in pgn, dc, unclear ending postion, discarded for performance calcualtions
18,woodrun,,2500,Katzenmaier,,2483,90'+30'',05/16/2017,1-0 (Forced white win)             in pgn, regular win, result kept
20,mateinfour,,2563,Sirocco,,2498,90'+30'',05/19/2017,1-0 (Forced white win)              in pgn, regular win, result kept
8,Claexit,,2400,Fahad,,2450,90'+30'',05/02/2017,0-0 (Absent)                              ----

3,Ultra-d,,2689,maximus,,2676,90'+30'',04/25/2017,0-1 (Forced black win)                  ----
3,hafaba,,2444,suzy,,2450,90'+30'',04/25/2017,0-1 (Forced black win)                      ----
7,ugurpc,,2400,Biertrinker,,2620,90'+30'',04/30/2017,0-1 (Forced black win)               ----
11,SOLIDOR,,2410,hafaba,,2444,90'+30'',05/06/2017,0-1 (Forced black win)                  ----
13,maximus,,2676,TheForce,,2450,90'+30'',05/09/2017,0-1 (Forced black win)                ----
13,Fahad,,2450,Longin,,2461,90'+30'',05/09/2017,0-1 (Forced black win)                    in pgn, dc
13,Sunnytown,,2365,retep1,,2450,90'+30'',05/09/2017,0-1 (Forced black win)                ----
17,SkyPilot,,2477,miguello,,2488,90'+30'',05/14/2017,0-1 (Forced black win)               in pgn, dc
20,Phyrros,,2450,A_h_a_n_i_b_a_l,,2450,90'+30'',05/19/2017,0-1 (Forced black win)         in pgn, dc


Games won without fight (including dc):  15
Number of game results changed to draw, for performance calcualtions: 6
Number of games discarded for performance calcualtions: 1


4,A_h_a_n_i_b_a_l,,2450,Fahad,,2450,90'+30'',04/26/2017,1-0 (Forced white win)            BOOK LOSS


1,Zor,,2563,suzy,,2450,90'+30'',04/22/2017,1-0                       BOOK LOSS in line played by engine, but already known as book loss in 30 games
1,fati64,,2510,hafaba,,2444,90'+30'',04/22/2017,1-0
1,Medusa,,2500,OlivierEvan,,2405,90'+30'',04/22/2017,1-0             MOUSE SLIP (and bad opening)
2,MGR,,2468,jernjoffen,,2464,90'+30'',04/23/2017,1-0
2,Harry_Potter,,2449,ChessLion,,2440,90'+30'',04/23/2017,1-0
2,Coco69,,2375,ASSA,,2369,90'+30'',04/23/2017,1-0
3,THE_MACHINE,,2447,ChessLion,,2440,90'+30'',04/25/2017,1-0
3,ASSA,,2369,Sunnytown,,2365,90'+30'',04/25/2017,1-0
4,CoffeeOne,,2606,Biertrinker,,2620,90'+30'',04/26/2017,1-0
4,LordSirKnight,,2450,Longin,,2461,90'+30'',04/26/2017,1-0
4,ChessLion,,2440,OlivierEvan,,2405,90'+30'',04/26/2017,1-0          MOUSE SLIP
4,Sunnytown,,2365,IronMan,,2500,90'+30'',04/26/2017,1-0
5,Zor,,2563,Medusa,,2500,90'+30'',04/28/2017,1-0                     BOOK LOSS, lost line played out off book, known as loosing in 30 games
5,Soloman,,2461,KRAKEN,,2475,90'+30'',04/28/2017,1-0                 BOOK LOSS in line played by engine, but already known as book loss in 60 games
6,KRAKEN,,2475,jernjoffen,,2464,90'+30'',04/29/2017,1-0
7,Time_bandit,,2500,Wolle,,2508,90'+30'',04/30/2017,1-0              BOOK LOSS, some bad book moves caused the loss
7,jernjoffen,,2464,Claexit,,2400,90'+30'',04/30/2017,1-0
8,Zor,,2563,CoffeeOne,,2606,90'+30'',05/02/2017,1-0                  
8,Katzenmaier,,2483,Murx,,2510,90'+30'',05/02/2017,1-0
8,kobudera,,2450,SOLIDOR,,2410,90'+30'',05/02/2017,1-0
8,ChessLion,,2440,jernjoffen,,2464,90'+30'',05/02/2017,1-0
8,KULLANI,,2458,hafaba,,2444,90'+30'',05/02/2017,1-0                 BOOK LOSS, some bad book moves caused the loss, unclear why a book should play that line
9,Akhtar,,2479,LordSirKnight,,2450,90'+30'',05/03/2017,1-0
9,KhalidOmar,,2450,ChessLion,,2440,90'+30'',05/03/2017,1-0           BLUNDER
9,Dovahkiin,,2622,Marsell,,2519,90'+30'',05/03/2017,1-0
9,XXAAXX,,2493,Sunnytown,,2365,90'+30'',05/03/2017,1-0               BLUNDER
9,ugurpc,,2400,Claexit,,2400,90'+30'',05/03/2017,1-0
9,jernjoffen,,2464,OlivierEvan,,2405,90'+30'',05/03/2017,1-0
11,EtaoinShrdlu,,2599,woodrun,,2500,90'+30'',05/06/2017,1-0
11,Soloman,,2461,kolelo,,2474,90'+30'',05/06/2017,1-0                BOOK LOSS, loosing move 16 ..Qf6 played in book
11,LordSirKnight,,2450,retep1,,2450,90'+30'',05/06/2017,1-0
12,Jpn,,2520,fati64,,2510,90'+30'',05/07/2017,1-0
12,Marsell,,2519,Sunnytown,,2365,90'+30'',05/07/2017,1-0
12,Wolle,,2508,OlivierEvan,,2405,90'+30'',05/07/2017,1-0
13,woodrun,,2500,xeonrocks,,2504,90'+30'',05/09/2017,1-0
13,jkec,,2420,MiG29,,2643,90'+30'',05/09/2017,1-0
13,Brtroy,,2485,Claexit,,2400,90'+30'',05/09/2017,1-0
13,ArKheiN,,2400,OlivierEvan,,2405,90'+30'',05/09/2017,1-0
15,Soloman,,2461,CoffeeOne,,2606,90'+30'',05/12/2017,1-0             BOOK LOSS, looks like regular win, but repeated lost line from 7 rounds earlier
15,miguello,,2488,OlivierEvan,,2405,90'+30'',05/12/2017,1-0
15,MrScience,,2400,Claexit,,2400,90'+30'',05/12/2017,1-0
16,Coco69,,2375,KhalidOmar,,2450,90'+30'',05/13/2017,1-0
16,Knightmare,,2505,PublicEnemyN1,,2554,90'+30'',05/13/2017,1-0
16,xeonrocks,,2504,OlivierEvan,,2405,90'+30'',05/13/2017,1-0
17,Medusa,,2500,Time_bandit,,2500,90'+30'',05/14/2017,1-0
17,newton296,,2566,Claexit,,2400,90'+30'',05/14/2017,1-0
18,Time_bandit,,2500,Claexit,,2400,90'+30'',05/16/2017,1-0           BOOK LOSS, bad moves from Houdini, seemed to be playing without book, wasting time from move 1
19,maximus,,2676,LordSirKnight,,2450,90'+30'',05/17/2017,1-0
19,Prepared,,2568,Sunnytown,,2365,90'+30'',05/17/2017,1-0
20,Gamelover,,2462,MrScience,,2400,90'+30'',05/19/2017,1-0
21,jkec,,2420,Longin,,2461,90'+30'',05/20/2017,1-0
21,KULLANI,,2458,Coco69,,2375,90'+30'',05/20/2017,1-0
21,Team_Komodo,,2500,kolelo,,2474,90'+30'',05/20/2017,1-0
21,MrScience,,2400,OlivierEvan,,2405,90'+30'',05/20/2017,1-0



1,KULLANI,,2458,EtaoinShrdlu,,2599,90'+30'',04/22/2017,0-1
1,THE_MACHINE,,2447,Murx,,2510,90'+30'',04/22/2017,0-1
1,Claexit,,2400,woodrun,,2500,90'+30'',04/22/2017,0-1
3,XXAAXX,,2493,Gilgamesch,,2494,90'+30'',04/25/2017,0-1
3,Gamelover,,2462,kolelo,,2474,90'+30'',04/25/2017,0-1
4,MiG29,,2643,Regina_H.Milch,,2658,90'+30'',04/26/2017,0-1          BOOK LOSS
4,Brtroy,,2485,Akhtar,,2479,90'+30'',04/26/2017,0-1                 BLUNDER, two bad moves which cannot be explained, engines wouldn't play them
5,OlivierEvan,,2405,IronMan,,2500,90'+30'',04/28/2017,0-1
7,maximus,,2676,CoffeeOne,,2606,90'+30'',04/30/2017,0-1
7,Dovahkiin,,2622,Ozymandias,,2670,90'+30'',04/30/2017,0-1
7,miguello,,2488,Katzenmaier,,2483,90'+30'',04/30/2017,0-1          MOUSE SLIP
8,Marsell,,2519,MCDermont,,2545,90'+30'',05/02/2017,0-1
9,newton296,,2566,Paul,,2516,90'+30'',05/03/2017,0-1
10,RJN,,2477,KULLANI,,2458,90'+30'',05/05/2017,0-1                  
10,Sunnytown,,2365,jernjoffen,,2464,90'+30'',05/05/2017,0-1         
11,Coco69,,2375,Victorious,,2610,90'+30'',05/06/2017,0-1
11,Prepared,,2568,deepthroat,,2647,90'+30'',05/06/2017,0-1
11,ArKheiN,,2400,TheForce,,2450,90'+30'',05/06/2017,0-1             
12,jernjoffen,,2464,izanagi,,2485,90'+30'',05/07/2017,0-1
13,Time_bandit,,2500,CoffeeOne,,2606,90'+30'',05/09/2017,0-1
14,Sunnytown,,2365,RJN,,2477,90'+30'',05/10/2017,0-1
14,OlivierEvan,,2405,MiG29,,2643,90'+30'',05/10/2017,0-1
15,ChessLion,,2440,ArKheiN,,2400,90'+30'',05/12/2017,0-1
16,Sunnytown,,2365,fati64,,2510,90'+30'',05/13/2017,0-1
18,ChessLion,,2440,PublicEnemyN1,,2554,90'+30'',05/16/2017,0-1      MOUSE SLIP + BLUNDER (and bad opening), discarded for performance calcualtions
19,TheForce,,2450,deepthroat,,2647,90'+30'',05/17/2017,0-1          BOOK LOSS, lost line played out off book, known as loosing in 30 games
20,Coco69,,2375,Thomas_A_Anderson,,2648,90'+30'',05/19/2017,0-1
20,OlivierEvan,,2405,FireFather,,2357,90'+30'',05/19/2017,0-1       MOUSE SLIP
20,Sunnytown,,2365,MiG29,,2643,90'+30'',05/19/2017,0-1
20,KhalidOmar,,2450,hafaba,,2444,90'+30'',05/19/2017,0-1
21,MGR,,2468,TheForce,,2450,90'+30'',05/20/2017,0-1
21,FireFather,,2357,kobudera,,2450,90'+30'',05/20/2017,0-1
21,ChessLion,,2440,SkyPilot,,2477,90'+30'',05/20/2017,0-1

Irregular games won: 19
Number of game results changed to draw, for performance calcualtions: 7
Number of games discarded for performance calculations: book loses + 1 = 12
To properly calculate the performance, every game should be analyzed, but I already did that in 2014, and I can only recall one instance where both players made mistakes. If one player makes a mistake, almost surely, that will result in a loss, so checking decided games may be all we need:

Code: Select all

#  PLAYER            RATING   POINTS PLAYED (%)   W    L    D(%)
1  thomas_a_anderson 2585     11.5   21   54.8    2    0    90.5
2  etaoinshrdlu      2582     11.5   21   54.8    2    0    90.5
3  coffeeone         2573     11.0   20   55.0    3    1    80.0
4  zor               2569     10.0   19   52.6    1    0    94.7
5  regina_h.milch    2568     10.5   20   52.5    1    0    95.0
6  woodrun           2566     11.5   21   54.8    3    1    81.0
7  akhtar            2564     11.0   21   52.4    1    0    95.2
8  ozymandias        2560     11.0   21   52.4    1    0    95.2
9  mcdermont         2559     11.0   21   52.4    1    0    95.2
10 deepthroat        2559     10.5   20   52.5    1    0    95.0
11 jpn               2557     11.0   21   52.4    1    0    95.2
12 victorious        2556     11.0   21   52.4    1    0    95.2
13 theforce          2553     10.5   19   55.3    2    0    89.5
14 kingslayer        2547     11.0   21   52.4    1    0    95.2
15 gilgamesch        2543     11.0   21   52.4    1    0    95.2
16 paul              2542     11.0   21   52.4    1    0    95.2
17 soloman           2541      9.0   18   50.0    0    0   100.0
18 mateinfour        2535     11.0   21   52.4    1    0    95.2
19 knightmare        2531     11.0   21   52.4    1    0    95.2
20 kobudera          2529     11.5   21   54.8    2    0    90.5
21 maximus           2529      9.5   19   50.0    1    1    89.5
22 jkec              2528     11.5   21   54.8    2    0    90.5
23 medusa            2526     10.5   20   52.5    1    0    95.0
24 izanagi           2525     11.0   21   52.4    1    0    95.2
25 harry_potter      2524     11.0   21   52.4    1    0    95.2
26 team_komodo       2523     11.0   21   52.4    1    0    95.2
27 murx              2523     10.5   21   50.0    1    1    90.5
28 spaghetti_chess   2522     10.5   21   50.0    0    0   100.0
29 kullani           2521     10.5   20   52.5    2    1    85.0
30 fati64            2519     10.5   21   50.0    2    2    81.0
31 kraken            2514     10.5   20   52.5    1    0    95.0
32 dovahkiin         2513     10.5   21   50.0    1    1    90.5
33 kolelo            2511     10.0   20   50.0    1    1    90.0
34 a_h_a_n_i_b_a_l   2510     10.0   20   50.0    0    0   100.0
35 mgr               2510     10.5   21   50.0    1    1    90.5
36 brtroy            2509     11.0   21   52.4    1    0    95.2
37 coco69            2507      9.5   20   47.5    2    3    75.0
38 lordsirknight     2506     10.5   21   50.0    2    2    81.0
39 newton296         2504     10.5   21   50.0    1    1    90.5
40 katzenmaier       2503     10.0   21   47.6    1    2    85.7
41 biertrinker       2502      9.5   20   47.5    0    1    95.0
42 prepared          2502     10.5   21   50.0    1    1    90.5
43 miguello          2501     11.0   21   52.4    1    0    95.2
44 skypilot          2500     11.0   21   52.4    1    0    95.2
45 publicenemyn1     2496      9.5   20   47.5    0    1    95.0
46 xeonrocks         2496     10.5   21   50.0    1    1    90.5
47 gamelover         2496     10.5   21   50.0    1    1    90.5
48 rjn               2496     10.5   21   50.0    1    1    90.5
49 phyrros           2495     10.0   20   50.0    0    0   100.0
50 ocirema           2490     10.5   21   50.0    0    0   100.0
51 suzy              2489      9.5   19   50.0    0    0   100.0
52 mrscience         2488     10.5   20   52.5    2    1    85.0
53 mig29             2488     10.5   20   52.5    2    1    85.0
54 time_bandit       2486      8.5   19   44.7    0    2    89.5
55 arkhein           2483     11.0   21   52.4    2    1    85.7
56 sirocco           2482     10.0   21   47.6    0    1    95.2
57 the_machine       2480     10.5   21   50.0    1    1    90.5
58 firefather        2476     10.0   21   47.6    1    2    85.7
59 retep1            2474      9.5   20   47.5    0    1    95.0
60 xxaaxx            2472     10.0   21   47.6    0    1    95.2
61 maniac            2471     10.5   21   50.0    0    0   100.0
62 ugurpc            2470     10.5   20   52.5    1    0    95.0
63 marsell           2457     10.0   21   47.6    1    2    85.7
64 wolle             2456     10.0   20   50.0    1    1    90.0
65 longin            2454      9.0   20   45.0    0    2    90.0
66 hafaba            2437      9.0   18   50.0    1    1    88.9
67 khalidomar        2436      9.0   20   45.0    0    2    90.0
68 jernjoffen        2416      9.5   21   45.2    2    4    71.4
69 chesslion         2394      8.5   20   42.5    1    4    75.0
70 sunnytown         2353      7.0   19   36.8    1    6    63.2
71 claexit           2342      6.5   19   34.2    0    6    68.4
72 olivierevan       2307      6.5   21   31.0    0    8    61.9
* Only the players who finished the tournament are on the list, so you have to start counting from Jernjoffen, and include Ironman, to get the list of six player I'm talking about. They don't appear, but their games are part of the calculations (notice the number of games played by the opponents opponents).

Thomas A. Anderson
Posts: 18
Joined: Tue Feb 23, 2016 5:57 pm

Re: $20,000 ICUC - a report on wins

Post by Thomas A. Anderson » Mon May 29, 2017 3:31 pm

Interesting stats, Juan. In my oppinion, the main reason for this extremly high draw rates is the stubborn adherence to drawish lines, when the bots playing itselfs. Just did a quick check and even in recent (2017) correspondence games we don't get such high rates (maximum was arround 87% at levels between 2400 and 2500). Have you also get a statistic of games with at least one centaurs in play? The real interesting stuff might be a statistics where the evaluation data/curve during the course of a game is considered. We know that centaurs tried hard for getting some imbalance in the games, while engines are usually satisfied with evals of 0.0x during the entire game. While most of the centaur games ended up with draw as well, this difference isn't well reflected in a statistics that is only feeded by the resulst of the games. Anyway, thanks for the work you and Thomas have already done, I was asking myself many times how a cleaned-up statistics would look like.
cu

jefk
Posts: 390
Joined: Sun Jul 25, 2010 8:07 pm
Location: the Netherlands
Contact:

Re: $20,000 ICUC - a report on wins

Post by jefk » Wed May 31, 2017 5:11 pm

Thomas A. Anderson wrote:Interesting stats, Juan. In my opinion, the main reason for this extremly high draw rates is the stubborn adherence to drawish lines, when the bots playing itselfs.
maybe, but what do you mean with 'drawish'? Nowadays -for example- the 'slow Italian' is often played, and if an suboptimal book move is made (somewhere in the modern long lines in this opening), you often get unbalanced positions; yet with the Cerebellum books often used in this tourn (Brainfish, and probably also often with Raubfisch) such mistakes are rare.

Because/as result of the prize money i get the impression the level was high, although there still were some blunders obviously, mouseslips, bad book moves, and sometimes crappy openings.

How to avoid 'drawish'lines ? Well playing Sicilian with Black is possible i think but (also) requires a very accurate book; ideally even better than Cerebellum level, at least the April 2017 Cerebellum level. Maybe eg 1.g4 is not a drawish line but it's simply not wise for White to play it. Then there are ofcourse many c4 d4 or Nf3 lines, sometimes transposed, but are they less 'drawish' ? Maybe, if White tries to maintain an advantage, but in this tourn winning with Black was already quite difficult and i also speak from my own experience (won twice with White, but only in 2nd half of the tourn after i had improved my book and switched to centaur mode..).

As for the engines, there's indeed still a lot to improve, and i think engine programmers could still learn from some of these (slow) games, if they would study them carefully (and have some own chess knowledge as well). As for eg the Elo of SF vs Kom10 or 11 in fast games, this doesn't say much about the performance during such slow games. Even the socalled corfish has not the software (eval/heuristic) routines which could benefit from such slow time controls; hint it has to do with planning ofcourse, when does a pawn avalanche work, when it doesnt, how to attack based on the criteria of a certain pawnstructure, and so on; having defeated one Brainfisch and one Raubfisch, needless to say i wasn't so impressed with the performance of the fishes, even although they usuallyl showed impressive search depths like >40 (a reason why i started using some faster hardware as centaur, and in my won games i didn't use a fish at all; but i did use my own book which i had checked against cerebellum, a book of which i gradually became less impressed with (some indeed say it gives 'drawish'lines but then i'm going into circular reasoning
:)
jefk

Thomas A. Anderson
Posts: 18
Joined: Tue Feb 23, 2016 5:57 pm

Re: $20,000 ICUC - a report on wins

Post by Thomas A. Anderson » Wed May 31, 2017 7:52 pm

Maybe I should have called them the bullet-prooved drawish lines. Of course there are many, many drawish opening lines all over the place. But the bots played almost everytime the same lines up to the 20th-30th move, with very low variaty. And they ended up in positions, that they proved not to be able to win in many thousand of games at IC tournaments. I have to agree, that this Italian game lines are very hard to push out of the draw bandwidth. But top centaurs demonstrated how it can work out. But the engines at their own played the positions very unambitioned, ending up with many, many draws. I'm sure that the cerebellum book, as well as all other books, are still full of dubious lines. But if you play 90% of the time the best discovered Italian game opening line, its simply very unlikely that you will disclose such a book inaccurancy.
cu

User avatar
Leto
Posts: 2028
Joined: Thu May 04, 2006 1:40 am
Location: Dune

Re: $20,000 ICUC - a report on wins

Post by Leto » Thu Jun 01, 2017 3:17 pm

Which one was the Brainfish team?

User avatar
Ozymandias
Posts: 978
Joined: Sun Oct 25, 2009 12:30 am

Re: $20,000 ICUC - a report on wins

Post by Ozymandias » Sat Jun 03, 2017 9:20 am

Thomas A. Anderson wrote:Have you also get a statistic of games with at least one centaurs in play?
Just one thing to add, to the answer I already gave him on the Rybka Forum, his nick is thomas_a_anderson, #1 on the performance list.
jefk wrote:what do you mean with 'drawish'? Nowadays -for example- the 'slow Italian' is often played, and if an suboptimal book move is made (somewhere in the modern long lines in this opening), you often get unbalanced positions; yet with the Cerebellum books often used in this tour (Brainfish, and probably also often with Raubfisch) such mistakes are rare.
[...]
How to avoid 'drawish'lines ? Well playing Sicilian with Black is possible i think but (also) requires a very accurate book; ideally even better than Cerebellum level, at least the April 2017 Cerebellum level. Maybe eg 1.g4 is not a drawish line but it's simply not wise for White to play it. Then there are ofcourse many c4 d4 or Nf3 lines, sometimes transposed, but are they less 'drawish' ? Maybe, if White tries to maintain an advantage, but in this tour winning with Black was already quite difficult and i also speak from my own experience (won twice with White, but only in 2nd half of the tour after i had improved my book and switched to centaur mode..)
There's drawish lines and risky lines for white, there's drawish lines and suicide lines for black. That may sound like an oversimplification, but I guess I get my point across with such a hyperbole. Giuoco Pianissimo not only is drawish, but favourable to black if played correctly from both sides, and yet, it was the line that scored better for white in this tournament. When I say "favourable", I mean that we get a draw if white avoids black's traps. White also has a few, but not as many, and in any case, all of them are already known and the decided games only highlight my point from the first post: it doesn't matter what you play, it matters what your opponent plays. Unbalanced positions don't mean much, if engines are able to find a draw in the end.

Playing anything other than the Marsall Attack with black, is playing with fire, it may still be possible, but I doubt anyone is going to benefit (win) from doing so, against a well booked opponent. Again, is the opponent that matters. If white wants to avoid playing against it (or the semi-slav, for that matter), that's when things can get interesting. Most of the times they don't, because many sidelines (I consider the Giuoco Pianissimo a sideline) are a draw, but sometimes white may even go for sidelines of sidelines. Case in point, in my last game as white, I tried an opening with a very low draw rate, but black played correctly and I had to fight for a draw. It's sad, but that's currently the state of affairs. That opening had been previously tried in this tournament, with Alvin losing the white side, despite black not playing accurately, which again, gets back to my point: everything is a draw, as long as your opponent doesn't screw-up.

As for your wins, tell me your knick and I'll tell you why you won.
Thomas A. Anderson wrote:I have to agree, that this Italian game lines are very hard to push out of the draw bandwidth. But top centaurs demonstrated how it can work out.
The stats you requested, and which I provided, seem to indicate otherwise. Draw rate is almost identical, whether you have centaurs or engines involved in the game. That's basically because of the Marshall and the semi-slav. It's as if you had two walls, one to the left and another one to the right, no matter where you turn your head to, you hit it against one.
Leto wrote:Which one was the Brainfish team?
gilgamesch

Thomas A. Anderson
Posts: 18
Joined: Tue Feb 23, 2016 5:57 pm

Re: $20,000 ICUC - a report on wins

Post by Thomas A. Anderson » Sat Jun 03, 2017 6:05 pm

Ozymandias wrote:it doesn't matter what you play, it matters what your opponent plays.
I fear you are very right on that. I remember the stats in 2014 and they have been already significant. But now in 2017, we are at such high draw rates, that blunders became an even more important, if not decisive factor for this kind of chess tournaments. We urgently need to find something to stop this development.
cu

User avatar
Ozymandias
Posts: 978
Joined: Sun Oct 25, 2009 12:30 am

Re: $20,000 ICUC - a report on wins

Post by Ozymandias » Sun Jun 04, 2017 12:45 pm

Thomas A. Anderson wrote:We urgently need to find something to stop this development.
The main problem, is that some players, like Jef, still believe they can score better trough their own play, and most of the players who already acknowledge reality, remain silent.
From Thomas, who helped gathering the info, to many others who shared their concerns with me, via PM or on the tournament's private chat, all we're getting is silence, and you know what happens when just a small minority voice concerns? Nothing.

Next time, we'll have something like a 95% filtered draw rate, and in the not so distant future, a player like Olivier Evans, will become the most deciding factor, no matter what tournament format you choose, because the only wins left, will be against players like him. Arno himself singled him out, more than a year ago, when he removed his games from the stats of the Centaur Grand Prix, but he's still to make the connection between polluted stats and a polluted tournament.

Thomas A. Anderson
Posts: 18
Joined: Tue Feb 23, 2016 5:57 pm

Re: $20,000 ICUC - a report on wins

Post by Thomas A. Anderson » Sun Jun 04, 2017 6:58 pm

Before the tournament started I asked TD what he thinks about having the number of "technical" wins as the first tie-breaker (the player with the most "wins by dc etc, get lowest rating). My oppinion is that for that type of hyper-modern-chess-tournaments we need nowadays adjusted tie-breakers and not the old-fashioned BH/SB stuff. I after the tournament I try to keep contact with TD and hopefully we can establish something modernized for future tournaments. This might make things a little bit better, but surely will not eliminate all problems. For example, what I'm totally clueless about is, how to deal with things like we've seen in case of CoffeeOne's book weakness in the tour. We can't and do not want to prohibit disclosures of book weaknesses during a tournament. But what has to been adressed is, what happened after a disclosure is public? In the ICUC tour we've seen many players trying to hit the variation in the following rounds. In most cases the books got adjusted until the very next rounds and everything was fine. Ironically it looks like all attempts to hit the hole ended sucessless (or even fired back against centaurs), but the player CoffeeOne itself, alhtough regognizing the problem instantly, didn't fix his book and loose two more games with that variation against Zor (very next round) and Soloman (8 rounds later!). Needless to say, what this meant for the tour standings at the end. So we might find solution for some problems, other are harder to avoid, if not impossible.
cu

User avatar
Ozymandias
Posts: 978
Joined: Sun Oct 25, 2009 12:30 am

Re: $20,000 ICUC - a report on wins

Post by Ozymandias » Sun Jun 04, 2017 10:15 pm

Thomas A. Anderson wrote:we need nowadays adjusted tie-breakers and not the old-fashioned BH/SB stuff. [...] We can't and do not want to prohibit disclosures of book weaknesses during a tournament.
There're many things about how chess tournaments are run, that no longer apply. We're dealing with a different game now, one that's been proven in practice, to be a draw. Considering what Thomas and I labelled as "irregular wins", as the first tie-break rule, is a no-brainier, but I think we should be more ambitious, about the changes needed, and tackle the problem head on, at a much earlier stage.

Besides disconnections, mouse slips and blunders, book losses were the single most destabilising factor, accounting for almost as many irregularities, as all the other causes put together. And we're just talking about the most blatant of cases; other strange openings, which couldn't be directly pointed as the cause of a loss, certainly gave the winning side, ample advantage. The only way to weed out (most of) those weak books, is by having at least two categories of play. They would still be in play, but on the lower category. Having players on separate categories, would also obviously deal with most of the disconnections, mouse slips and blunders, as they have a tendency to affect repeatedly, the usual suspects.

In essence, this would reinforce claims, about the upper league being the best chess ever played, or any statement to that effect. When you throw all players into the mix, the same can't be said without serious reservations.

Post Reply