A complete 2000 lines of code engine

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Guenther
Posts: 3015
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: A complete 2000 lines of code engine

Post by Guenther » Tue Jul 09, 2019 1:05 pm

A little test with Minic on my old Quadcore vs. former #1 Rybka 1.0b-64.
400 games, tc 40m/40'', 256MB hash, randomly selected 6 plies openings from 1200 (quite balanced) with colors reversed.

(games are available in the unlikely case someone wants them, despite the tc and hardware)

Code: Select all

Minic076_Test
RWBC Win7U-64 Q8200 2.33Ghz Nvidia GT 710, 2019.07.08 - 2019.07.09
 1: Rybka_10b-64  258.0/400   +203  -87 =110
 2: Minic_076-64  142.0/400    +87 -203 =110
                              +153 -137 =110 (White POV)
 
Just for fun I computed the match standings and percentage after each game pair.
This shows how much the result still changes even for (still) not so close programs over hundreds of games.
(Caution: may be dangerous to the eyes of stats non-believers)

Code: Select all

Rybka	Minic	%Ryb	%Min	Games
1,0	1,0	50,00	50,00	(2)
2,0	2,0	50,00	50,00	(4)
3,0	3,0	50,00	50,00	(6)
3,0	5,0	37,50	62,50	(8)
3,5	6,5	35,00	65,00	(10)
5,0	7,0	41,67	58,33	(12)
5,0	9,0	35,71	64,29	(14)
6,5	9,5	40,63	59,38	(16)
8,0	10,0	44,44	55,56	(18)
9,5	10,5	47,50	52,50	(20)
10,0	12,0	45,45	54,55	(22)
10,5	13,5	43,75	56,25	(24)
12,0	14,0	46,15	53,85	(26)
13,0	15,0	46,43	53,57	(28)
14,5	15,5	48,33	51,67	(30)
15,5	16,5	48,44	51,56	(32)
17,5	16,5	51,47	48,53	(34)
18,5	17,5	51,39	48,61	(36)
20,5	17,5	53,95	46,05	(38)
22,0	18,0	55,00	45,00	(40)
22,0	20,0	52,38	47,62	(42)
22,0	22,0	50,00	50,00	(44)
24,0	22,0	52,17	47,83	(46)
26,0	22,0	54,17	45,83	(48)
27,5	22,5	55,00	45,00	(50)
28,5	23,5	54,81	45,19	(52)
29,0	25,0	53,70	46,30	(54)
31,0	25,0	55,36	44,64	(56)
32,0	26,0	55,17	44,83	(58)
33,5	26,5	55,83	44,17	(60)
35,5	26,5	57,26	42,74	(62)
36,5	27,5	57,03	42,97	(64)
38,0	28,0	57,58	42,42	(66)
39,5	28,5	58,09	41,91	(68)
40,5	29,5	57,86	42,14	(70)
40,5	31,5	56,25	43,75	(72)
42,0	32,0	56,76	43,24	(74)
42,0	34,0	55,26	44,74	(76)
43,0	35,0	55,13	44,87	(78)
44,5	35,5	55,63	44,38	(80)
46,0	36,0	56,10	43,90	(82)
46,0	38,0	54,76	45,24	(84)
46,5	39,5	54,07	45,93	(86)
48,5	39,5	55,11	44,89	(88)
50,0	40,0	55,56	44,44	(90)
51,0	41,0	55,43	44,57	(92)
53,0	41,0	56,38	43,62	(94)
54,5	41,5	56,77	43,23	(96)
56,0	42,0	57,14	42,86	(98)
57,5	42,5	57,50	42,50	(100)
59,0	43,0	57,84	42,16	(102)
61,0	43,0	58,65	41,35	(104)
62,0	44,0	58,49	41,51	(106)
63,5	44,5	58,80	41,20	(108)
65,0	45,0	59,09	40,91	(110)
67,0	45,0	59,82	40,18	(112)
68,5	45,5	60,09	39,91	(114)
70,0	46,0	60,34	39,66	(116)
71,0	47,0	60,17	39,83	(118)
72,0	48,0	60,00	40,00	(120)
74,0	48,0	60,66	39,34	(122)
75,5	48,5	60,89	39,11	(124)
77,5	48,5	61,51	38,49	(126)
78,0	50,0	60,94	39,06	(128)
79,0	51,0	60,77	39,23	(130)
80,0	52,0	60,61	39,39	(132)
81,0	53,0	60,45	39,55	(134)
82,5	53,5	60,66	39,34	(136)
83,5	54,5	60,51	39,49	(138)
84,0	56,0	60,00	40,00	(140)
84,0	58,0	59,15	40,85	(142)
84,0	60,0	58,33	41,67	(144)
86,0	60,0	58,90	41,10	(146)
88,0	60,0	59,46	40,54	(148)
88,5	61,5	59,00	41,00	(150)
90,0	62,0	59,21	40,79	(152)
91,5	62,5	59,42	40,58	(154)
93,0	63,0	59,62	40,38	(156)
94,5	63,5	59,81	40,19	(158)
96,0	64,0	60,00	40,00	(160)
97,5	64,5	60,19	39,81	(162)
98,5	65,5	60,06	39,94	(164)
99,0	67,0	59,64	40,36	(166)
100,0	68,0	59,52	40,48	(168)
101,0	69,0	59,41	40,59	(170)
103,0	69,0	59,88	40,12	(172)
105,0	69,0	60,34	39,66	(174)
105,5	70,5	59,94	40,06	(176)
106,5	71,5	59,83	40,17	(178)
107,0	73,0	59,44	40,56	(180)
108,5	73,5	59,62	40,38	(182)
110,5	73,5	60,05	39,95	(184)
112,0	74,0	60,22	39,78	(186)
114,0	74,0	60,64	39,36	(188)
116,0	74,0	61,05	38,95	(190)
117,5	74,5	61,20	38,80	(192)
117,5	76,5	60,57	39,43	(194)
119,0	77,0	60,71	39,29	(196)
120,0	78,0	60,61	39,39	(198)
121,5	78,5	60,75	39,25	(200)
122,5	79,5	60,64	39,36	(202)
124,5	79,5	61,03	38,97	(204)
125,0	81,0	60,68	39,32	(206)
126,0	82,0	60,58	39,42	(208)
127,0	83,0	60,48	39,52	(210)
128,5	83,5	60,61	39,39	(212)
129,5	84,5	60,51	39,49	(214)
131,5	84,5	60,88	39,12	(216)
133,5	84,5	61,24	38,76	(218)
135,5	84,5	61,59	38,41	(220)
137,5	84,5	61,94	38,06	(222)
139,5	84,5	62,28	37,72	(224)
140,5	85,5	62,17	37,83	(226)
142,0	86,0	62,28	37,72	(228)
143,0	87,0	62,17	37,83	(230)
145,0	87,0	62,50	37,50	(232)
146,0	88,0	62,39	37,61	(234)
147,5	88,5	62,50	37,50	(236)
148,5	89,5	62,39	37,61	(238)
150,5	89,5	62,71	37,29	(240)
152,0	90,0	62,81	37,19	(242)
153,5	90,5	62,91	37,09	(244)
155,5	90,5	63,21	36,79	(246)
156,5	91,5	63,10	36,90	(248)
158,5	91,5	63,40	36,60	(250)
159,5	92,5	63,29	36,71	(252)
160,5	93,5	63,19	36,81	(254)
161,5	94,5	63,09	36,91	(256)
162,5	95,5	62,98	37,02	(258)
163,0	97,0	62,69	37,31	(260)
165,0	97,0	62,98	37,02	(262)
166,0	98,0	62,88	37,12	(264)
167,5	98,5	62,97	37,03	(266)
169,0	99,0	63,06	36,94	(268)
169,5	100,5	62,78	37,22	(270)
171,5	100,5	63,05	36,95	(272)
172,5	101,5	62,96	37,04	(274)
173,5	102,5	62,86	37,14	(276)
174,0	104,0	62,59	37,41	(278)
176,0	104,0	62,86	37,14	(280)
178,0	104,0	63,12	36,88	(282)
179,5	104,5	63,20	36,80	(284)
180,0	106,0	62,94	37,06	(286)
181,5	106,5	63,02	36,98	(288)
183,5	106,5	63,28	36,72	(290)
185,0	107,0	63,36	36,64	(292)
187,0	107,0	63,61	36,39	(294)
189,0	107,0	63,85	36,15	(296)
191,0	107,0	64,09	35,91	(298)
191,0	109,0	63,67	36,33	(300)
192,0	110,0	63,58	36,42	(302)
192,5	111,5	63,32	36,68	(304)
193,5	112,5	63,24	36,76	(306)
193,5	114,5	62,82	37,18	(308)
195,0	115,0	62,90	37,10	(310)
196,0	116,0	62,82	37,18	(312)
197,0	117,0	62,74	37,26	(314)
198,5	117,5	62,82	37,18	(316)
200,0	118,0	62,89	37,11	(318)
201,0	119,0	62,81	37,19	(320)
202,5	119,5	62,89	37,11	(322)
203,0	121,0	62,65	37,35	(324)
205,0	121,0	62,88	37,12	(326)
205,0	123,0	62,50	37,50	(328)
205,5	124,5	62,27	37,73	(330)
206,5	125,5	62,20	37,80	(332)
208,5	125,5	62,43	37,57	(334)
209,0	127,0	62,20	37,80	(336)
210,5	127,5	62,28	37,72	(338)
211,5	128,5	62,21	37,79	(340)
212,5	129,5	62,13	37,87	(342)
214,0	130,0	62,21	37,79	(344)
216,0	130,0	62,43	37,57	(346)
217,5	130,5	62,50	37,50	(348)
219,0	131,0	62,57	37,43	(350)
220,5	131,5	62,64	37,36	(352)
222,0	132,0	62,71	37,29	(354)
223,0	133,0	62,64	37,36	(356)
224,0	134,0	62,57	37,43	(358)
226,0	134,0	62,78	37,22	(360)
227,0	135,0	62,71	37,29	(362)
229,0	135,0	62,91	37,09	(364)
230,5	135,5	62,98	37,02	(366)
232,5	135,5	63,18	36,82	(368)
234,0	136,0	63,24	36,76	(370)
235,5	136,5	63,31	36,69	(372)
237,5	136,5	63,50	36,50	(374)
239,0	137,0	63,56	36,44	(376)
240,0	138,0	63,49	36,51	(378)
242,0	138,0	63,68	36,32	(380)
243,0	139,0	63,61	36,39	(382)
244,5	139,5	63,67	36,33	(384)
246,0	140,0	63,73	36,27	(386)
248,0	140,0	63,92	36,08	(388)
249,5	140,5	63,97	36,03	(390)
251,0	141,0	64,03	35,97	(392)
252,5	141,5	64,09	35,91	(394)
254,5	141,5	64,27	35,73	(396)
256,0	142,0	64,32	35,68	(398)
258,0	142,0	64,50	35,50	(400)
Here some insights in depth/time usage for several stages in all games done with a not yet released (AFAIK?)
version of another great tool by Ferdinand.
The final table computes all moves in all games, for the other ones just note the move window numbers.
(For those who don't remember, you can safely add 5 plies to Rybkas depth count)

Code: Select all

B. Time and Depth average:

move window  : [4, 16]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:23:41  13.87  00:00:966
  2                     Rybka_10b-64      400  64.5  00:01:02:40   9.02  00:00:723

move window  : [17, 28]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:27:53  15.08  00:01:108
  2                     Rybka_10b-64      400  64.5  00:00:59:48   9.39  00:00:756

move window  : [29, 40]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:22:28  16.51  00:01:077
  2                     Rybka_10b-64      400  64.5  00:01:08:24  10.46  00:00:902

move window  : [41, 52]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:21:18  19.13  00:01:186
  2                     Rybka_10b-64      400  64.5  00:01:14:33  11.70  00:01:110

move window  : [53, 64]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:00:58:19  22.10  00:01:074
  2                     Rybka_10b-64      400  64.5  00:00:49:43  12.47  00:00:941

move window  : [4, 400]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:08:04:44  19.15  00:01:055
  2                     Rybka_10b-64      400  64.5  00:06:45:45  11.28  00:00:898
Current foe list count : [93]
http://rwbc-chess.de/chronology.htm

User avatar
xr_a_y
Posts: 733
Joined: Sat Nov 25, 2017 1:28 pm
Location: France

Re: A complete 2000 lines of code engine

Post by xr_a_y » Tue Jul 09, 2019 6:08 pm

Guenther wrote:
Tue Jul 09, 2019 1:05 pm
A little test with Minic on my old Quadcore vs. former #1 Rybka 1.0b-64.
400 games, tc 40m/40'', 256MB hash, randomly selected 6 plies openings from 1200 (quite balanced) with colors reversed.

(games are available in the unlikely case someone wants them, despite the tc and hardware)

Code: Select all

Minic076_Test
RWBC Win7U-64 Q8200 2.33Ghz Nvidia GT 710, 2019.07.08 - 2019.07.09
 1: Rybka_10b-64  258.0/400   +203  -87 =110
 2: Minic_076-64  142.0/400    +87 -203 =110
                              +153 -137 =110 (White POV)
 
Just for fun I computed the match standings and percentage after each game pair.
This shows how much the result still changes even for (still) not so close programs over hundreds of games.
(Caution: may be dangerous to the eyes of stats non-believers)

Code: Select all

Rybka	Minic	%Ryb	%Min	Games
1,0	1,0	50,00	50,00	(2)
2,0	2,0	50,00	50,00	(4)
3,0	3,0	50,00	50,00	(6)
3,0	5,0	37,50	62,50	(8)
3,5	6,5	35,00	65,00	(10)
5,0	7,0	41,67	58,33	(12)
5,0	9,0	35,71	64,29	(14)
6,5	9,5	40,63	59,38	(16)
8,0	10,0	44,44	55,56	(18)
9,5	10,5	47,50	52,50	(20)
10,0	12,0	45,45	54,55	(22)
10,5	13,5	43,75	56,25	(24)
12,0	14,0	46,15	53,85	(26)
13,0	15,0	46,43	53,57	(28)
14,5	15,5	48,33	51,67	(30)
15,5	16,5	48,44	51,56	(32)
17,5	16,5	51,47	48,53	(34)
18,5	17,5	51,39	48,61	(36)
20,5	17,5	53,95	46,05	(38)
22,0	18,0	55,00	45,00	(40)
22,0	20,0	52,38	47,62	(42)
22,0	22,0	50,00	50,00	(44)
24,0	22,0	52,17	47,83	(46)
26,0	22,0	54,17	45,83	(48)
27,5	22,5	55,00	45,00	(50)
28,5	23,5	54,81	45,19	(52)
29,0	25,0	53,70	46,30	(54)
31,0	25,0	55,36	44,64	(56)
32,0	26,0	55,17	44,83	(58)
33,5	26,5	55,83	44,17	(60)
35,5	26,5	57,26	42,74	(62)
36,5	27,5	57,03	42,97	(64)
38,0	28,0	57,58	42,42	(66)
39,5	28,5	58,09	41,91	(68)
40,5	29,5	57,86	42,14	(70)
40,5	31,5	56,25	43,75	(72)
42,0	32,0	56,76	43,24	(74)
42,0	34,0	55,26	44,74	(76)
43,0	35,0	55,13	44,87	(78)
44,5	35,5	55,63	44,38	(80)
46,0	36,0	56,10	43,90	(82)
46,0	38,0	54,76	45,24	(84)
46,5	39,5	54,07	45,93	(86)
48,5	39,5	55,11	44,89	(88)
50,0	40,0	55,56	44,44	(90)
51,0	41,0	55,43	44,57	(92)
53,0	41,0	56,38	43,62	(94)
54,5	41,5	56,77	43,23	(96)
56,0	42,0	57,14	42,86	(98)
57,5	42,5	57,50	42,50	(100)
59,0	43,0	57,84	42,16	(102)
61,0	43,0	58,65	41,35	(104)
62,0	44,0	58,49	41,51	(106)
63,5	44,5	58,80	41,20	(108)
65,0	45,0	59,09	40,91	(110)
67,0	45,0	59,82	40,18	(112)
68,5	45,5	60,09	39,91	(114)
70,0	46,0	60,34	39,66	(116)
71,0	47,0	60,17	39,83	(118)
72,0	48,0	60,00	40,00	(120)
74,0	48,0	60,66	39,34	(122)
75,5	48,5	60,89	39,11	(124)
77,5	48,5	61,51	38,49	(126)
78,0	50,0	60,94	39,06	(128)
79,0	51,0	60,77	39,23	(130)
80,0	52,0	60,61	39,39	(132)
81,0	53,0	60,45	39,55	(134)
82,5	53,5	60,66	39,34	(136)
83,5	54,5	60,51	39,49	(138)
84,0	56,0	60,00	40,00	(140)
84,0	58,0	59,15	40,85	(142)
84,0	60,0	58,33	41,67	(144)
86,0	60,0	58,90	41,10	(146)
88,0	60,0	59,46	40,54	(148)
88,5	61,5	59,00	41,00	(150)
90,0	62,0	59,21	40,79	(152)
91,5	62,5	59,42	40,58	(154)
93,0	63,0	59,62	40,38	(156)
94,5	63,5	59,81	40,19	(158)
96,0	64,0	60,00	40,00	(160)
97,5	64,5	60,19	39,81	(162)
98,5	65,5	60,06	39,94	(164)
99,0	67,0	59,64	40,36	(166)
100,0	68,0	59,52	40,48	(168)
101,0	69,0	59,41	40,59	(170)
103,0	69,0	59,88	40,12	(172)
105,0	69,0	60,34	39,66	(174)
105,5	70,5	59,94	40,06	(176)
106,5	71,5	59,83	40,17	(178)
107,0	73,0	59,44	40,56	(180)
108,5	73,5	59,62	40,38	(182)
110,5	73,5	60,05	39,95	(184)
112,0	74,0	60,22	39,78	(186)
114,0	74,0	60,64	39,36	(188)
116,0	74,0	61,05	38,95	(190)
117,5	74,5	61,20	38,80	(192)
117,5	76,5	60,57	39,43	(194)
119,0	77,0	60,71	39,29	(196)
120,0	78,0	60,61	39,39	(198)
121,5	78,5	60,75	39,25	(200)
122,5	79,5	60,64	39,36	(202)
124,5	79,5	61,03	38,97	(204)
125,0	81,0	60,68	39,32	(206)
126,0	82,0	60,58	39,42	(208)
127,0	83,0	60,48	39,52	(210)
128,5	83,5	60,61	39,39	(212)
129,5	84,5	60,51	39,49	(214)
131,5	84,5	60,88	39,12	(216)
133,5	84,5	61,24	38,76	(218)
135,5	84,5	61,59	38,41	(220)
137,5	84,5	61,94	38,06	(222)
139,5	84,5	62,28	37,72	(224)
140,5	85,5	62,17	37,83	(226)
142,0	86,0	62,28	37,72	(228)
143,0	87,0	62,17	37,83	(230)
145,0	87,0	62,50	37,50	(232)
146,0	88,0	62,39	37,61	(234)
147,5	88,5	62,50	37,50	(236)
148,5	89,5	62,39	37,61	(238)
150,5	89,5	62,71	37,29	(240)
152,0	90,0	62,81	37,19	(242)
153,5	90,5	62,91	37,09	(244)
155,5	90,5	63,21	36,79	(246)
156,5	91,5	63,10	36,90	(248)
158,5	91,5	63,40	36,60	(250)
159,5	92,5	63,29	36,71	(252)
160,5	93,5	63,19	36,81	(254)
161,5	94,5	63,09	36,91	(256)
162,5	95,5	62,98	37,02	(258)
163,0	97,0	62,69	37,31	(260)
165,0	97,0	62,98	37,02	(262)
166,0	98,0	62,88	37,12	(264)
167,5	98,5	62,97	37,03	(266)
169,0	99,0	63,06	36,94	(268)
169,5	100,5	62,78	37,22	(270)
171,5	100,5	63,05	36,95	(272)
172,5	101,5	62,96	37,04	(274)
173,5	102,5	62,86	37,14	(276)
174,0	104,0	62,59	37,41	(278)
176,0	104,0	62,86	37,14	(280)
178,0	104,0	63,12	36,88	(282)
179,5	104,5	63,20	36,80	(284)
180,0	106,0	62,94	37,06	(286)
181,5	106,5	63,02	36,98	(288)
183,5	106,5	63,28	36,72	(290)
185,0	107,0	63,36	36,64	(292)
187,0	107,0	63,61	36,39	(294)
189,0	107,0	63,85	36,15	(296)
191,0	107,0	64,09	35,91	(298)
191,0	109,0	63,67	36,33	(300)
192,0	110,0	63,58	36,42	(302)
192,5	111,5	63,32	36,68	(304)
193,5	112,5	63,24	36,76	(306)
193,5	114,5	62,82	37,18	(308)
195,0	115,0	62,90	37,10	(310)
196,0	116,0	62,82	37,18	(312)
197,0	117,0	62,74	37,26	(314)
198,5	117,5	62,82	37,18	(316)
200,0	118,0	62,89	37,11	(318)
201,0	119,0	62,81	37,19	(320)
202,5	119,5	62,89	37,11	(322)
203,0	121,0	62,65	37,35	(324)
205,0	121,0	62,88	37,12	(326)
205,0	123,0	62,50	37,50	(328)
205,5	124,5	62,27	37,73	(330)
206,5	125,5	62,20	37,80	(332)
208,5	125,5	62,43	37,57	(334)
209,0	127,0	62,20	37,80	(336)
210,5	127,5	62,28	37,72	(338)
211,5	128,5	62,21	37,79	(340)
212,5	129,5	62,13	37,87	(342)
214,0	130,0	62,21	37,79	(344)
216,0	130,0	62,43	37,57	(346)
217,5	130,5	62,50	37,50	(348)
219,0	131,0	62,57	37,43	(350)
220,5	131,5	62,64	37,36	(352)
222,0	132,0	62,71	37,29	(354)
223,0	133,0	62,64	37,36	(356)
224,0	134,0	62,57	37,43	(358)
226,0	134,0	62,78	37,22	(360)
227,0	135,0	62,71	37,29	(362)
229,0	135,0	62,91	37,09	(364)
230,5	135,5	62,98	37,02	(366)
232,5	135,5	63,18	36,82	(368)
234,0	136,0	63,24	36,76	(370)
235,5	136,5	63,31	36,69	(372)
237,5	136,5	63,50	36,50	(374)
239,0	137,0	63,56	36,44	(376)
240,0	138,0	63,49	36,51	(378)
242,0	138,0	63,68	36,32	(380)
243,0	139,0	63,61	36,39	(382)
244,5	139,5	63,67	36,33	(384)
246,0	140,0	63,73	36,27	(386)
248,0	140,0	63,92	36,08	(388)
249,5	140,5	63,97	36,03	(390)
251,0	141,0	64,03	35,97	(392)
252,5	141,5	64,09	35,91	(394)
254,5	141,5	64,27	35,73	(396)
256,0	142,0	64,32	35,68	(398)
258,0	142,0	64,50	35,50	(400)
Here some insights in depth/time usage for several stages in all games done with a not yet released (AFAIK?)
version of another great tool by Ferdinand.
The final table computes all moves in all games, for the other ones just note the move window numbers.
(For those who don't remember, you can safely add 5 plies to Rybkas depth count)

Code: Select all

B. Time and Depth average:

move window  : [4, 16]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:23:41  13.87  00:00:966
  2                     Rybka_10b-64      400  64.5  00:01:02:40   9.02  00:00:723

move window  : [17, 28]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:27:53  15.08  00:01:108
  2                     Rybka_10b-64      400  64.5  00:00:59:48   9.39  00:00:756

move window  : [29, 40]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:22:28  16.51  00:01:077
  2                     Rybka_10b-64      400  64.5  00:01:08:24  10.46  00:00:902

move window  : [41, 52]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:01:21:18  19.13  00:01:186
  2                     Rybka_10b-64      400  64.5  00:01:14:33  11.70  00:01:110

move window  : [53, 64]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:00:58:19  22.10  00:01:074
  2                     Rybka_10b-64      400  64.5  00:00:49:43  12.47  00:00:941

move window  : [4, 400]

 nr                           player    games  pts%      sumTime aveDep    aveTime
  1                     Minic_076-64      400  35.5  00:08:04:44  19.15  00:01:055
  2                     Rybka_10b-64      400  64.5  00:06:45:45  11.28  00:00:898


Very good informations ! thanks. Indeed, less than 1000 games shall never be trusted. Often see some 10-0 became 15-30 or so ...

Another good news is that is it a 100 elo difference versus Rybka1 so Minic is around 2725 as expected ;-)

User avatar
Guenther
Posts: 3015
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: A complete 2000 lines of code engine

Post by Guenther » Tue Jul 09, 2019 6:23 pm

xr_a_y wrote:
Tue Jul 09, 2019 6:08 pm
Guenther wrote:
Tue Jul 09, 2019 1:05 pm
A little test with Minic on my old Quadcore vs. former #1 Rybka 1.0b-64.
400 games, tc 40m/40'', 256MB hash, randomly selected 6 plies openings from 1200 (quite balanced) with colors reversed.

(games are available in the unlikely case someone wants them, despite the tc and hardware)

Code: Select all

Minic076_Test
RWBC Win7U-64 Q8200 2.33Ghz Nvidia GT 710, 2019.07.08 - 2019.07.09
 1: Rybka_10b-64  258.0/400   +203  -87 =110
 2: Minic_076-64  142.0/400    +87 -203 =110
                              +153 -137 =110 (White POV)
 
Very good informations ! thanks. Indeed, less than 1000 games shall never be trusted. Often see some 10-0 became 15-30 or so ...

Another good news is that is it a 100 elo difference versus Rybka1 so Minic is around 2725 as expected ;-)
In CCRL 40/4 Rybka 1.0b 64 bit is even 2853.

Ordo with 1000 simuls says this then:

Code: Select all

   # PLAYER          :   RATING  ERROR  POINTS  PLAYED    (%)    W    D    L
   1 Rybka_10b-64    :  2853.00  15.31   258.0     400  64.50  203  110   87
   2 Minic_076-64    :  2748.16  15.31   142.0     400  35.50   87  110  203

White advantage = 15.32 +/- 14.88
Draw rate (equal opponents) = 29.31 % +/- 2.30
From watching a few dozens of games inbetween I can say that Rybka just relentlessly tries to
create an unstoppable passer and succeeds quite often in this vs. Minic.
Current foe list count : [93]
http://rwbc-chess.de/chronology.htm

User avatar
xr_a_y
Posts: 733
Joined: Sat Nov 25, 2017 1:28 pm
Location: France

Re: A complete 2000 lines of code engine

Post by xr_a_y » Tue Jul 09, 2019 7:15 pm

Guenther wrote:
Tue Jul 09, 2019 6:23 pm
From watching a few dozens of games inbetween I can say that Rybka just relentlessly tries to
create an unstoppable passer and succeeds quite often in this vs. Minic.
Very good hint !
Maybe I should add some code for candidates.
Passer is already there, and quite tuned.
There is an addition bonus for free passer and connected passer.
Another bonus in endgame is rook is behind.
A malus for isolated, doubled and backward.

I'll check if I can build some easy candidate code ...

User avatar
Guenther
Posts: 3015
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: A complete 2000 lines of code engine

Post by Guenther » Tue Jul 09, 2019 7:22 pm

xr_a_y wrote:
Tue Jul 09, 2019 7:15 pm
Guenther wrote:
Tue Jul 09, 2019 6:23 pm
From watching a few dozens of games inbetween I can say that Rybka just relentlessly tries to
create an unstoppable passer and succeeds quite often in this vs. Minic.
Very good hint !
Maybe I should add some code for candidates.
Passer is already there, and quite tuned.
There is an addition bonus for free passer and connected passer.
Another bonus in endgame is rook is behind.
A malus for isolated, doubled and backward.

I'll check if I can build some easy candidate code ...
Ok it seems you ask for another hint :)
A few times I saw Minics Q grab some b2/a2 pawns and then being locked there, while Rybka attacked the K-side or even trapped Minics Q.
(Ofc this was at the depths of the fast tc as shown in the original post - but I guess knowing this can also help at higher depths/tc)

Two examples below. Note that with current analysing with Minic just from FEN/EPD the moves cannot be completly verified for
that depth/time, as obviously hash entries from previous searches tend to favour those moves.
(Still it can be seen that those moves need probably more depth to be resolved as for other programs at that strength)

Current foe list count : [93]
http://rwbc-chess.de/chronology.htm

User avatar
xr_a_y
Posts: 733
Joined: Sat Nov 25, 2017 1:28 pm
Location: France

Re: A complete 2000 lines of code engine

Post by xr_a_y » Tue Jul 09, 2019 8:18 pm

Guenther wrote:
Tue Jul 09, 2019 7:22 pm
xr_a_y wrote:
Tue Jul 09, 2019 7:15 pm
Guenther wrote:
Tue Jul 09, 2019 6:23 pm
From watching a few dozens of games inbetween I can say that Rybka just relentlessly tries to
create an unstoppable passer and succeeds quite often in this vs. Minic.
Very good hint !
Maybe I should add some code for candidates.
Passer is already there, and quite tuned.
There is an addition bonus for free passer and connected passer.
Another bonus in endgame is rook is behind.
A malus for isolated, doubled and backward.

I'll check if I can build some easy candidate code ...
Ok it seems you ask for another hint :)
A few times I saw Minics Q grab some b2/a2 pawns and then being locked there, while Rybka attacked the K-side or even trapped Minics Q.
(Ofc this was at the depths of the fast tc as shown in the original post - but I guess knowing this can also help at higher depths/tc)

Two examples below. Note that with current analysing with Minic just from FEN/EPD the moves cannot be completly verified for
that depth/time, as obviously hash entries from previous searches tend to favour those moves.
(Still it can be seen that those moves need probably more depth to be resolved as for other programs at that strength)

Some kind of Queen safety was on my todo list ;-)
I have also some trapped pièces code not activated yet.

User avatar
xr_a_y
Posts: 733
Joined: Sat Nov 25, 2017 1:28 pm
Location: France

Re: A complete 2000 lines of code engine

Post by xr_a_y » Wed Jul 17, 2019 6:35 pm

Code: Select all

Score of minic_0.77 vs Winter0.6: 193 - 549 - 258 [0.322]
Elo difference: -129.3 +/- 19.3, LOS: 0.0 %, DrawRatio: 25.8 %

1000 of 1000 games finished.
Not that bad ... :)

User avatar
xr_a_y
Posts: 733
Joined: Sat Nov 25, 2017 1:28 pm
Location: France

Re: A complete 2000 lines of code engine

Post by xr_a_y » Sat Jul 27, 2019 7:24 am

Minic 0.82 is out with some new features :
- pondering
- analysis mode
- some minor fix and tweak
- same level as TCEC version
- a kind a level thing (level 0 being a random mover), to be continued ...

Code: Select all

Rank Name                          Elo     +/-   Games   Score   Draws
   1 minic_devl10                  621      41    1525   97.3%    4.5%
   2 minic_devl9                   387      26    1525   90.3%    7.6%
   3 minic_devl8                   235      20    1527   79.5%    7.7%
   4 minic_devl7                   158      18    1527   71.3%    6.4%
   5 minic_devl6                    74      17    1527   60.5%    3.9%
   6 minic_devl5                    -6      17    1528   49.1%    3.4%
   7 minic_devl4                   -75      17    1529   39.4%    4.6%
   8 minic_devl3                  -151      18    1529   29.5%    6.3%
   9 minic_devl2                  -264      21    1529   18.0%    6.7%
  10 minic_devl1                  -298      23    1529   15.3%    6.0%
  11 minic_devl0                 -1114     235    1529    0.2%    0.3%

User avatar
xr_a_y
Posts: 733
Joined: Sat Nov 25, 2017 1:28 pm
Location: France

Re: A complete 2000 lines of code engine

Post by xr_a_y » Sat Jul 27, 2019 4:51 pm

:!: :!: :!: :!: :!: :!: :!: :!:

WARNING : Dist directory as been REMOVED from the repository because it was starting to be too big. Unofficial releases are not available anymore. This operation has changed Minic git history, so you shall probably re-clone a clean repo I you have a Minic copy.

:!: :!: :!: :!: :!: :!: :!: :!:

User avatar
M ANSARI
Posts: 3408
Joined: Thu Mar 16, 2006 6:10 pm

Re: A complete 2000 lines of code engine

Post by M ANSARI » Mon Jul 29, 2019 9:13 am

Can I ask what the present number of line the code is now? Would be interesting to see if it can still do well without going over 2000 lines or if 2000 is a number that hurts progress significantly.

Post Reply