The Distributed Chess Project

Creating Chess-Playing Artificial Neural Networks with Distributed Evolutionary Algorithms

Since 06/23/2002 a cumulated computing time of 3 years 15 days 2 hours 2 min 35 sec has been contributed.

98 participants have returned at least one candidate neural network for this problem set. The results of the best ones are ranked according to test performance, corresponding training performance and contributed computing time in the following table. The problems come from the collection '1001 Brilliant Ways to Checkmate' by Fred Reinfeld, 1955, ISBN: 0-87980-110-7. Both the training and the test set consist of 50 problems from this collection.

Rank	User-ID	Best Performance on Test Set (%)	Best Performance on Training Set (%)	Cumulated Computing Time (d:h:m:s)
1	esa.elovaara	78	88	87:18:13:47
2	jason	78	88	7:3:22:3
3	peter.k.campbell	78	86	70:22:30:18
4	henkf	76	94	8:11:10:50
5	midmalex	76	92	65:0:49:6
6	regloh.lliw	76	92	11:4:24:8
7	ravik	76	92	7:8:7:3
8	ssieh	76	92	1:2:9:47
9	freemac_99	76	90	50:10:1:26
10	scotte	76	90	9:23:21:39
11	pt.schwarz	76	90	7:23:16:44
12	john.obrien	76	90	7:13:24:58
13	thornton	76	90	5:13:7:57
14	vishvananda	76	88	33:8:55:11
15	stefan.bengtsson	76	88	2:10:55:26
16	lithron	76	88	1:22:39:52
17	W	76	86	14:11:55:25
18	gipe	76	86	2:22:34:39
19	sr	76	82	14:1:4:23
20	amiguel	74	94	11:22:16:33
21	m_diosi	74	92	32:10:9:6
22	tran_tang	74	92	21:12:11:31
23	ad_lord\hotmail.com	74	92	17:12:17:24
24	atalmadge	74	92	13:4:22:25
25	NeuroBernd	74	92	5:20:23:15
26	u37206657	74	90	32:12:47:4
27	simon	74	90	30:7:54:55
28	richbcla	74	90	22:3:1:17
29	njenson	74	90	17:7:5:26
30	david7091	74	90	9:8:34:20
31	5xd7g251lq2vs001	74	90	7:19:49:30
32	gillrich	74	90	3:13:12:34
33	ddcc	74	90	2:21:10:12
34	shop	74	90	2:4:18:42
35	qwerty4	74	88	15:6:31:45
36	jarekfil	74	88	3:21:51:7
37	regloh.lliw	74	88	2:12:22:10
38	mitchellonline	74	86	8:0:19:27
39	jean_efpraxiadis	74	84	16:15:9:32
40	rich_luce	74	84	8:20:6:46
41	arnowa	72	94	6:14:56:2
42	thunder124	72	92	15:6:57:29
43	ssmythe	72	92	8:12:34:10
44	zanglerska	72	92	3:21:14:39
45	cglose	72	92	2:16:49:59
46	santzi	72	92	2:13:44:23
47	toddkloos	72	92	1:18:49:49
48	Sebastian	72	92	1:9:25:17
49	llclarisse	72	90	38:6:25:28
50	organizer	72	90	27:19:22:12
51	ic	72	90	26:2:31:58
52	jwa	72	90	20:19:32:33
53	wingy	72	90	20:2:5:9
54	mutefly	72	90	15:9:50:32
55	joachim	72	90	10:17:26:20
56	ish_alviso	72	90	10:11:53:17
57	idiott47	72	90	8:12:17:43
58	kannan	72	90	7:10:12:14
59	alasdair	72	90	6:1:17:58
60	kennynthebest	72	90	6:0:19:37
61	jopicard	72	90	4:20:22:12
62	azrael	72	90	4:18:7:10
63	fronte	72	90	3:7:31:4
64	doo	72	90	2:16:45:29
65	frank	72	90	2:4:59:41
66	sliver	72	90	1:21:28:23
67	vince.mele	72	90	1:18:2:21
68	aaron	72	90	0:23:47:58
69	holle	72	88	17:9:56:33
70	mentis	72	86	20:1:9:18
71	a.wagner1	72	86	5:7:23:57
72	petera	72	86	0:5:7:53
73	afmomania	72	86	0:3:27:12
74	mhond99	72	84	6:8:55:35
75	turrinipa	72	82	10:0:29:33
76	layton97	72	80	3:17:16:44
77	amanwithattitude	70	94	1:23:3:7
78	rdmltrs	70	92	9:20:46:55
79	kurt	70	92	4:10:23:8
80	arno	70	92	3:10:10:46
81	lshome	70	92	3:3:59:49
82	shambalamaster	70	92	3:0:57:36
83	troebie	70	92	1:23:8:35
84	davidroel	70	90	5:19:51:36
85	NNBernd	70	90	3:7:49:43
86	paul.abrams	70	90	1:20:1:41
87	joe	70	90	1:13:57:13
88	stjernbecker	70	90	0:22:20:39
89	berndt	70	86	2:5:4:32
90	chess	68	94	5:12:34:34
91	Nico.DeBecker	68	94	3:5:5:53
92	agarcia	68	92	14:14:36:18
93	mljwing	68	92	12:20:16:30
94	Barrios_Blue	68	92	6:1:41:49
95	stuart	68	90	7:13:26:46
96	ttf11	68	90	2:12:35:23
97	pouilley	68	88	2:15:47:56
98	jperkins	68	78	3:3:3:44

4 participants have returned at least one candidate neural network for this problem set. The results of the best ones are ranked according to test performance, corresponding training performance and contributed computing time in the following table. This problem set comes from a collection of more than 3000 endgame problems which can be found here.

Both the training and the test set consist of 200 problems each which I have randomly selected from the 3000+ problems. If one were to pick the solutions for those 400 problems randomly, one could expect a performance of 9.7% on average. Therefore, any significantly higher performance can be considered as a success.