Re: WBEC Ridderkerk EloStat result.

[Jose Carlos]

Geschrieben von: / Posted by: Jose Carlos at 02 February 2004 18:07:33:
Als Antwort auf: / In reply to: WBEC Ridderkerk EloStat result. geschrieben von: / posted by: Leo Dijksman at 02 February 2004 17:31:47:

For who is interrested I have created a ratinglist using EloStat v1.2b from all WBEC Ridderkerk games played sofar (20399 games, 498 players (new engine qualify's are not included)), start rating 2150, minimal 15 games/player:
Program Elo + - Games Score Av.Op. Draws
1 Ruffian 1.0.1 : 2677 48 92 100 79.0 % 2447 22.0 %
2 The King 3.23 : 2601 72 68 68 63.2 % 2507 35.3 %
3 Ruffian 0.76 : 2597 77 108 52 69.2 % 2456 19.2 %
4 Ruffian 1.0.5 : 2592 62 60 84 66.1 % 2476 39.3 %
5 DeepSjeng 1.5 : 2589 34 43 236 69.3 % 2448 31.8 %
6 Ruffian 2.0.0 : 2586 74 62 68 61.0 % 2508 39.7 %
7 SmarThink 0.17a : 2566 77 60 68 58.1 % 2509 39.7 %
8 Yace 0.99.77 : 2556 57 63 100 64.5 % 2452 29.0 %
9 Aristarch 4.21 : 2549 37 43 236 64.0 % 2450 25.4 %
10 Gandalf 4.32h : 2549 34 35 288 60.9 % 2471 30.2 %
11 Gandalf 5.1 : 2548 84 80 52 62.5 % 2460 32.7 %
12 Crafty 19.03 : 2541 68 49 84 58.9 % 2478 46.4 %
13 Kaissa 1.7 : 2535 54 172 68 88.2 % 2185 11.8 %
14 LittleGoliath 3.9po : 2534 52 42 152 55.9 % 2493 35.5 %
15 WARP 0.37 : 2528 45 40 184 59.0 % 2465 35.3 %
16 Crafty 19.01 : 2524 61 62 100 60.0 % 2453 26.0 %
17 Yace 0.99.56 : 2523 38 41 216 64.1 % 2423 32.9 %
18 Rebel 12 : 2512 69 69 68 50.0 % 2512 35.3 %
19 WBNimzo 2000b : 2510 35 29 340 55.4 % 2472 33.2 %
20 WARP 0.58 : 2508 61 86 68 49.3 % 2513 33.8 %
21 Averno 0.68a : 2507 120 265 20 85.0 % 2206 10.0 %

Is this true or am I dreaming?
In fact this strange things happens also in my tests. Every new version starts performing at very high level and goes down with more and more games. Anyway, after the bad performance in CCT this is a good new to make me a little bit happy.
José C.

[Uri Blass]

Geschrieben von: / Posted by: Uri Blass at 02 February 2004 21:20:57:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk EloStat result. geschrieben von: / posted by: Jose Carlos at 02 February 2004 18:07:33:

For who is interrested I have created a ratinglist using EloStat v1.2b from all WBEC Ridderkerk games played sofar (20399 games, 498 players (new engine qualify's are not included)), start rating 2150, minimal 15 games/player:
Program Elo + - Games Score Av.Op. Draws
1 Ruffian 1.0.1 : 2677 48 92 100 79.0 % 2447 22.0 %
2 The King 3.23 : 2601 72 68 68 63.2 % 2507 35.3 %
3 Ruffian 0.76 : 2597 77 108 52 69.2 % 2456 19.2 %
4 Ruffian 1.0.5 : 2592 62 60 84 66.1 % 2476 39.3 %
5 DeepSjeng 1.5 : 2589 34 43 236 69.3 % 2448 31.8 %
6 Ruffian 2.0.0 : 2586 74 62 68 61.0 % 2508 39.7 %
7 SmarThink 0.17a : 2566 77 60 68 58.1 % 2509 39.7 %
8 Yace 0.99.77 : 2556 57 63 100 64.5 % 2452 29.0 %
9 Aristarch 4.21 : 2549 37 43 236 64.0 % 2450 25.4 %
10 Gandalf 4.32h : 2549 34 35 288 60.9 % 2471 30.2 %
11 Gandalf 5.1 : 2548 84 80 52 62.5 % 2460 32.7 %
12 Crafty 19.03 : 2541 68 49 84 58.9 % 2478 46.4 %
13 Kaissa 1.7 : 2535 54 172 68 88.2 % 2185 11.8 %
14 LittleGoliath 3.9po : 2534 52 42 152 55.9 % 2493 35.5 %
15 WARP 0.37 : 2528 45 40 184 59.0 % 2465 35.3 %
16 Crafty 19.01 : 2524 61 62 100 60.0 % 2453 26.0 %
17 Yace 0.99.56 : 2523 38 41 216 64.1 % 2423 32.9 %
18 Rebel 12 : 2512 69 69 68 50.0 % 2512 35.3 %
19 WBNimzo 2000b : 2510 35 29 340 55.4 % 2472 33.2 %
20 WARP 0.58 : 2508 61 86 68 49.3 % 2513 33.8 %
21 Averno 0.68a : 2507 120 265 20 85.0 % 2206 10.0 %

Is this true or am I dreaming?
In fact this strange things happens also in my tests. Every new version starts performing at very high level and goes down with more and more games. Anyway, after the bad performance in CCT this is a good new to make me a little bit happy.
José C.

Averno did not play enough games and played against too weak opponents.
Note that even if a program get 85% in hundreds of games against average of 2206 then it means that we need also games against stronger opponents to know the real strength of it because it is possible to have 2 programs of equal strength when one of them tends to do more draws against stronger and weaker opponents because of differerent style or differewnt opening book.

Uri

[Jose Carlos]

Geschrieben von: / Posted by: Jose Carlos at 02 February 2004 21:53:19:
Als Antwort auf: / In reply to: Re: WBEC Ridderkerk EloStat result. geschrieben von: / posted by: Uri Blass at 02 February 2004 21:20:57:

For who is interrested I have created a ratinglist using EloStat v1.2b from all WBEC Ridderkerk games played sofar (20399 games, 498 players (new engine qualify's are not included)), start rating 2150, minimal 15 games/player:
Program Elo + - Games Score Av.Op. Draws
1 Ruffian 1.0.1 : 2677 48 92 100 79.0 % 2447 22.0 %
2 The King 3.23 : 2601 72 68 68 63.2 % 2507 35.3 %
3 Ruffian 0.76 : 2597 77 108 52 69.2 % 2456 19.2 %
4 Ruffian 1.0.5 : 2592 62 60 84 66.1 % 2476 39.3 %
5 DeepSjeng 1.5 : 2589 34 43 236 69.3 % 2448 31.8 %
6 Ruffian 2.0.0 : 2586 74 62 68 61.0 % 2508 39.7 %
7 SmarThink 0.17a : 2566 77 60 68 58.1 % 2509 39.7 %
8 Yace 0.99.77 : 2556 57 63 100 64.5 % 2452 29.0 %
9 Aristarch 4.21 : 2549 37 43 236 64.0 % 2450 25.4 %
10 Gandalf 4.32h : 2549 34 35 288 60.9 % 2471 30.2 %
11 Gandalf 5.1 : 2548 84 80 52 62.5 % 2460 32.7 %
12 Crafty 19.03 : 2541 68 49 84 58.9 % 2478 46.4 %
13 Kaissa 1.7 : 2535 54 172 68 88.2 % 2185 11.8 %
14 LittleGoliath 3.9po : 2534 52 42 152 55.9 % 2493 35.5 %
15 WARP 0.37 : 2528 45 40 184 59.0 % 2465 35.3 %
16 Crafty 19.01 : 2524 61 62 100 60.0 % 2453 26.0 %
17 Yace 0.99.56 : 2523 38 41 216 64.1 % 2423 32.9 %
18 Rebel 12 : 2512 69 69 68 50.0 % 2512 35.3 %
19 WBNimzo 2000b : 2510 35 29 340 55.4 % 2472 33.2 %
20 WARP 0.58 : 2508 61 86 68 49.3 % 2513 33.8 %
21 Averno 0.68a : 2507 120 265 20 85.0 % 2206 10.0 %

Is this true or am I dreaming?
In fact this strange things happens also in my tests. Every new version starts performing at very high level and goes down with more and more games. Anyway, after the bad performance in CCT this is a good new to make me a little bit happy.
José C.

Averno did not play enough games and played against too weak opponents.
Note that even if a program get 85% in hundreds of games against average of 2206 then it means that we need also games against stronger opponents to know the real strength of it because it is possible to have 2 programs of equal strength when one of them tends to do more draws against stronger and weaker opponents because of different style or differewnt opening book.

Uri

Absolutely. But don't forget that next best averno has 2241 so playing against 2206 would give just a little more than 50%.
Also I'm having a bad personal moment last weeks. Let me be happy until number of games grow
PS.: Leo, don't play more games with Averno. Let it memory rest in peaeace at 2507 for eternity...