Comparision of 1+1 and 5+3 for the IPON ... some statistics
Posted: Fri Dec 28, 2012 4:05 pm
Hello all,
I played the TOP20 of the IPON-RRRL without Fritz again. All is identical to th eIPON conditions but this time with a time control of 1 + 1 and not 5 + 3.
This is th eresult:
1+1 / 25650 games
29 games lost on time / 0.11%
[tt]
Rank Name Elo + - games score oppo. draws
1 Houdini 3 STD 3100 12 12 2700 84% 2812 22% + 9
2 Komodo 5 3007 11 11 2700 74% 2817 30% + 1
3 Critter 1.4a 2994 11 11 2700 73% 2818 34% + 9
4 Deep Rybka 4.1 2972 11 11 2700 70% 2819 34% + 9
5 Stockfish 2.2.2 JA 2956 10 10 2700 68% 2820 37% -12
6 Chiron 1.5 2844 10 10 2700 52% 2827 37% - 8
7 Naum 4.2 2826 10 10 2700 50% 2827 36% -15
8 HIARCS 14 WCSC 32b 2812 10 10 2700 48% 2828 34% -13
9 Deep Shredder 12 2800 10 10 2700 46% 2829 33% + 0
10 Gull 1.2 2798 10 10 2700 46% 2829 32% - 3
11 Hannibal 1.2 2790 10 10 2700 45% 2830 36% -11
12 spark-1.0 2775 10 10 2700 43% 2830 36% + 4
13 Deep Sjeng c't 2010 32b 2773 10 10 2700 43% 2830 34% -20
14 Spike 1.4 32b 2751 10 10 2700 40% 2832 34% -29
15 Protector 1.4.0 2735 10 11 2700 37% 2833 33% -28
16 Quazar 0.4 2714 11 10 2700 35% 2834 32% -27
17 Deep Junior 13.3 2709 11 11 2700 35% 2834 29% -45
18 Zappa Mexico II 2688 11 11 2700 31% 2835 32% -33
19 MinkoChess 1.3 2676 11 11 2700 30% 2836 31% -23[/tt]
Average draw rate: 33%
Average elo deviation to 5+3: -12 Elo
5+3 / 25650 games
49 games lost on time / 0.19%
[tt]Rank Name Elo + - games score oppo. draws
1 Houdini 3 STD 3091 11 11 2700 82% 2826 24% - 9
2 Komodo 5 3006 10 10 2700 73% 2831 35% - 1
3 Critter 1.4a 2985 10 10 2700 70% 2832 37% - 9
4 Stockfish 2.2.2 JA 2968 10 10 2700 68% 2833 40% +12
5 Deep Rybka 4.1 2963 10 10 2700 68% 2833 40% - 9
6 Chiron 1.5 2852 10 10 2700 52% 2839 42% + 8
7 Naum 4.2 2841 10 10 2700 50% 2840 41% +15
8 HIARCS 14 WCSC 32b 2825 10 10 2700 48% 2841 40% +13
9 Hannibal 1.2 2801 10 10 2700 45% 2842 41% +11
10 Gull 1.2 2801 10 10 2700 45% 2842 38% + 3
11 Deep Shredder 12 2800 10 10 2700 45% 2842 40% + 0
12 Deep Sjeng c't 2010 32b 2793 10 10 2700 43% 2842 40% +20
13 Spike 1.4 32b 2780 10 10 2700 42% 2843 40% +29
14 spark-1.0 2771 10 10 2700 40% 2844 39% - 4
15 Protector 1.4.0 2763 10 10 2700 39% 2844 39% +28
16 Deep Junior 13.3 2754 10 10 2700 38% 2845 33% +45
17 Quazar 0.4 2742 10 10 2700 37% 2845 37% +27
18 Zappa Mexico II 2721 10 10 2700 34% 2846 36% +33
19 MinkoChess 1.3 2699 10 10 2700 31% 2848 35% +23[/tt]
Average draw rate: 38%
Average elo deviation to 5+3: +12
Average Elo deviation is 12 so the 1+1 Rating has to be increased by 12 to get a propper comparison
(Actually it has to be something with the square of the difference ... too long ago, I think the difference is marginal and this will do):
1+1 adjusted:
[tt]Rank Name Elo + - games score oppo. draws
1 Houdini 3 STD 3112 12 12 2700 84% 2812 22% +21
2 Komodo 5 3019 11 11 2700 74% 2817 30% +13
3 Critter 1.4a 3006 11 11 2700 73% 2818 34% +21
4 Deep Rybka 4.1 2984 11 11 2700 70% 2819 34% +21
5 Stockfish 2.2.2 JA 2968 10 10 2700 68% 2820 37% + 0
6 Chiron 1.5 2856 10 10 2700 52% 2827 37% + 4
7 Naum 4.2 2838 10 10 2700 50% 2827 36% - 3
8 HIARCS 14 WCSC 32b 2824 10 10 2700 48% 2828 34% - 1
9 Deep Shredder 12 2812 10 10 2700 46% 2829 33% +12
10 Gull 1.2 2810 10 10 2700 46% 2829 32% + 9
11 Hannibal 1.2 2802 10 10 2700 45% 2830 36% + 1
12 spark-1.0 2787 10 10 2700 43% 2830 36% +16
13 Deep Sjeng c't 2010 32b 2785 10 10 2700 43% 2830 34% - 8
14 Spike 1.4 32b 2763 10 10 2700 40% 2832 34% -17
15 Protector 1.4.0 2747 10 11 2700 37% 2833 33% -16
16 Quazar 0.4 2726 11 10 2700 35% 2834 32% -15
17 Deep Junior 13.3 2721 11 11 2700 35% 2834 29% -33
18 Zappa Mexico II 2700 11 11 2700 31% 2835 32% -21
19 MinkoChess 1.3 2688 11 11 2700 30% 2836 31% -11[/tt]
While no engine is in another ball park (interesting expression) the maximum additional difference is 54 Elo which is quite a high number!
I have to think about that!
Bye
Ingo
PS: Results, Bayes and Elostat files in Archive at http://www.inwoba.de
I played the TOP20 of the IPON-RRRL without Fritz again. All is identical to th eIPON conditions but this time with a time control of 1 + 1 and not 5 + 3.
This is th eresult:
1+1 / 25650 games
29 games lost on time / 0.11%
[tt]
Rank Name Elo + - games score oppo. draws
1 Houdini 3 STD 3100 12 12 2700 84% 2812 22% + 9
2 Komodo 5 3007 11 11 2700 74% 2817 30% + 1
3 Critter 1.4a 2994 11 11 2700 73% 2818 34% + 9
4 Deep Rybka 4.1 2972 11 11 2700 70% 2819 34% + 9
5 Stockfish 2.2.2 JA 2956 10 10 2700 68% 2820 37% -12
6 Chiron 1.5 2844 10 10 2700 52% 2827 37% - 8
7 Naum 4.2 2826 10 10 2700 50% 2827 36% -15
8 HIARCS 14 WCSC 32b 2812 10 10 2700 48% 2828 34% -13
9 Deep Shredder 12 2800 10 10 2700 46% 2829 33% + 0
10 Gull 1.2 2798 10 10 2700 46% 2829 32% - 3
11 Hannibal 1.2 2790 10 10 2700 45% 2830 36% -11
12 spark-1.0 2775 10 10 2700 43% 2830 36% + 4
13 Deep Sjeng c't 2010 32b 2773 10 10 2700 43% 2830 34% -20
14 Spike 1.4 32b 2751 10 10 2700 40% 2832 34% -29
15 Protector 1.4.0 2735 10 11 2700 37% 2833 33% -28
16 Quazar 0.4 2714 11 10 2700 35% 2834 32% -27
17 Deep Junior 13.3 2709 11 11 2700 35% 2834 29% -45
18 Zappa Mexico II 2688 11 11 2700 31% 2835 32% -33
19 MinkoChess 1.3 2676 11 11 2700 30% 2836 31% -23[/tt]
Average draw rate: 33%
Average elo deviation to 5+3: -12 Elo
5+3 / 25650 games
49 games lost on time / 0.19%
[tt]Rank Name Elo + - games score oppo. draws
1 Houdini 3 STD 3091 11 11 2700 82% 2826 24% - 9
2 Komodo 5 3006 10 10 2700 73% 2831 35% - 1
3 Critter 1.4a 2985 10 10 2700 70% 2832 37% - 9
4 Stockfish 2.2.2 JA 2968 10 10 2700 68% 2833 40% +12
5 Deep Rybka 4.1 2963 10 10 2700 68% 2833 40% - 9
6 Chiron 1.5 2852 10 10 2700 52% 2839 42% + 8
7 Naum 4.2 2841 10 10 2700 50% 2840 41% +15
8 HIARCS 14 WCSC 32b 2825 10 10 2700 48% 2841 40% +13
9 Hannibal 1.2 2801 10 10 2700 45% 2842 41% +11
10 Gull 1.2 2801 10 10 2700 45% 2842 38% + 3
11 Deep Shredder 12 2800 10 10 2700 45% 2842 40% + 0
12 Deep Sjeng c't 2010 32b 2793 10 10 2700 43% 2842 40% +20
13 Spike 1.4 32b 2780 10 10 2700 42% 2843 40% +29
14 spark-1.0 2771 10 10 2700 40% 2844 39% - 4
15 Protector 1.4.0 2763 10 10 2700 39% 2844 39% +28
16 Deep Junior 13.3 2754 10 10 2700 38% 2845 33% +45
17 Quazar 0.4 2742 10 10 2700 37% 2845 37% +27
18 Zappa Mexico II 2721 10 10 2700 34% 2846 36% +33
19 MinkoChess 1.3 2699 10 10 2700 31% 2848 35% +23[/tt]
Average draw rate: 38%
Average elo deviation to 5+3: +12
Average Elo deviation is 12 so the 1+1 Rating has to be increased by 12 to get a propper comparison
(Actually it has to be something with the square of the difference ... too long ago, I think the difference is marginal and this will do):
1+1 adjusted:
[tt]Rank Name Elo + - games score oppo. draws
1 Houdini 3 STD 3112 12 12 2700 84% 2812 22% +21
2 Komodo 5 3019 11 11 2700 74% 2817 30% +13
3 Critter 1.4a 3006 11 11 2700 73% 2818 34% +21
4 Deep Rybka 4.1 2984 11 11 2700 70% 2819 34% +21
5 Stockfish 2.2.2 JA 2968 10 10 2700 68% 2820 37% + 0
6 Chiron 1.5 2856 10 10 2700 52% 2827 37% + 4
7 Naum 4.2 2838 10 10 2700 50% 2827 36% - 3
8 HIARCS 14 WCSC 32b 2824 10 10 2700 48% 2828 34% - 1
9 Deep Shredder 12 2812 10 10 2700 46% 2829 33% +12
10 Gull 1.2 2810 10 10 2700 46% 2829 32% + 9
11 Hannibal 1.2 2802 10 10 2700 45% 2830 36% + 1
12 spark-1.0 2787 10 10 2700 43% 2830 36% +16
13 Deep Sjeng c't 2010 32b 2785 10 10 2700 43% 2830 34% - 8
14 Spike 1.4 32b 2763 10 10 2700 40% 2832 34% -17
15 Protector 1.4.0 2747 10 11 2700 37% 2833 33% -16
16 Quazar 0.4 2726 11 10 2700 35% 2834 32% -15
17 Deep Junior 13.3 2721 11 11 2700 35% 2834 29% -33
18 Zappa Mexico II 2700 11 11 2700 31% 2835 32% -21
19 MinkoChess 1.3 2688 11 11 2700 30% 2836 31% -11[/tt]
While no engine is in another ball park (interesting expression) the maximum additional difference is 54 Elo which is quite a high number!
I have to think about that!
Bye
Ingo
PS: Results, Bayes and Elostat files in Archive at http://www.inwoba.de