AIIDE 2022 - what Steamhammer learned

Steamhammer did not record all games, only the last 200. I’ll mention my preparation for different opponents, because it makes a big difference.

Steamhammer’s gas steal achieved nothing significant against any opponent. Maybe next year I’ll disable it.

#1 bananabrain

opening	games	wins	wins	losses	first	last
10HatchHydra	2	0%	-	8:58	62	195
10HatchLing	1	0%	-	10:00	150	150
11Gas10PoolLurkerB	1	0%	-	20:12	32	32
11HatchTurtleHydra	1	0%	-	14:25	138	138
12-11Hatch	1	0%	-	11:11	47	47
12-12Hatch	1	0%	-	13:54	87	87
12Hatch12Pool	1	0%	-	15:02	137	137
12Hatch13Pool	1	0%	-	12:29	95	95
3HatchHydraExpo	1	0%	-	9:36	135	135
3HatchLing	1	0%	-	8:19	109	109
3HatchLingBust	47	26%	15:19	11:56	0	197
3HatchLingExpo	9	11%	7:58	8:56	3	154
4HatchBeforeLair	1	0%	-	8:34	7	7
4PoolHard	1	0%	-	8:00	159	159
5HatchBeforeGas	13	8%	12:11	11:56	4	161
5HatchPoolLing	1	0%	-	14:37	22	22
5HatchPoolLingBurrow	30	17%	22:38	10:23	50	187
5PoolHard2Player	1	0%	-	17:25	39	39
6PoolBurrow	1	0%	-	11:54	69	69
6PoolHide	2	0%	-	16:23	37	110
6PoolSpeed	1	0%	-	8:54	49	49
7-7HydraRush	1	0%	-	7:34	33	33
7PoolHard	1	0%	-	12:44	79	79
8Gas7PoolLurker B	1	0%	-	6:21	73	73
8Hatch7Pool	1	0%	-	13:32	106	106
8Pool	1	0%	-	11:18	44	44
9HatchExpo9Pool9Gas	1	0%	-	10:09	40	40
9HatchMain9Pool9Gas	19	16%	9:51	11:37	112	192
9Pool	2	0%	-	10:12	164	193
9PoolExpo	1	0%	-	11:38	74	74
9PoolHatch	1	0%	-	16:25	2	2
9PoolHatchBurrow	1	0%	-	14:41	57	57
9PoolHatchSpeed7Drone	1	0%	-	12:46	82	82
9PoolHatchSpeed7DroneB	1	0%	-	11:29	76	76
9PoolHatchSpeedSpire2	1	0%	-	15:09	55	55
9PoolHatchSpire	1	0%	-	8:10	99	99
9PoolLair	2	0%	-	11:31	52	105
9PoolLurker	1	0%	-	12:00	175	175
9PoolSpeed	1	0%	-	9:00	75	75
9PoolSpeedAllIn	2	0%	-	7:25	141	198
9PoolSpire	1	0%	-	6:43	143	143
9PoolSunkHatch	2	0%	-	13:15	98	111
9PoolSunkSpeed	1	0%	-	7:25	83	83
AntiFact_2Hatch	1	0%	-	6:49	199	199
AntiFactoryHydra	1	0%	-	11:02	129	129
AntiStyx_9Pool	2	0%	-	9:01	108	170
AntiStyx_Muta	1	0%	-	9:46	35	35
AntiZeal_12Hatch	1	0%	-	15:27	171	171
DefilerRush	1	0%	-	13:19	188	188
Over10Hatch	2	0%	-	11:31	86	165
Over10Hatch2Sunk	2	0%	-	17:36	30	42
Over10PoolHydra	2	0%	-	10:40	1	24
OverhatchExpoLing	2	0%	-	14:19	45	177
OverhatchExpoMuta	2	0%	-	12:15	80	184
OverhatchLing	2	0%	-	10:58	25	48
Overpool+1	1	0%	-	8:44	185	185
OverpoolHide	1	0%	-	9:38	145	145
OverpoolLurker	1	0%	-	6:57	162	162
OverpoolSpeed	1	0%	-	10:04	89	89
Phlegethon	10	20%	8:26	14:02	9	196
ZvP_2HatchFakeHydra	1	0%	-	9:20	78	78
ZvP_2HatchFakeMuta	1	0%	-	14:56	139	139
ZvP_3HatchPoolHydra	1	0%	-	16:34	26	26
ZvP_Overpool3Hatch	1	0%	-	12:40	46	46
ZvT_13Pool	1	0%	-	7:52	160	160
65 openings	200	12%	12:49	11:31

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	17	8%	6%	14	7%	7%	24%	47%
Heavy rush	87	44%	11%	79	40%	13%	75%	15%
Naked expand	9	4%	11%	4	2%	50%	0%	89%
Proxy	24	12%	12%	16	8%	0%	29%	17%
Safe expand	53	26%	17%	32	16%	12%	51%	9%
Turtle	10	5%	0%	15	8%	7%	10%	20%
Unknown		-	-	40	20%	15%	0%	0%

timing	#	median	early	late
my combat unit	200	3:09	1:49	7:43
my gas	183	3:27	1:15	11:15
enemy scout	200	1:44	0:57	6:47
enemy combat unit	200	2:53	2:18	6:02
enemy gas	187	6:18	2:48	10:52
enemy air unit	183	6:33	2:49	12:46
enemy cloaked unit	139	9:02	3:26	13:50
game duration	200	11:33	5:52	26:39

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	53	2:14	1:58	3:46	15%	6:27
gas steal success	38	2:20	2:08	3:43	16%	6:28
none or failed	162	-	-	-	11%	6:12
gas steal killed	38	3:05	2:51	20:12

On BASIL, Steamhammer had been scoring best against BananaBrain with the Phlegethon ling-flood build. If that worked, then I reasoned (without testing) that the similar but more efficient 3HatchLingBust should be better. I gave Steamhammer hints to try both, and I was right. It wins sometimes when BananaBrain incautiously expands too early. Since no opening was good, Steamhammer experimented widely and found a lot of inferior tries.

#2 stardust

opening	games	wins	wins	losses	first	last
11Gas10PoolLurker	1	0%	-	11:22	34	34
11HatchTurtleHydra	2	0%	-	10:54	3	4
11HatchTurtleMuta	13	8%	13:04	12:38	13	153
12HatchTurtle	1	0%	-	12:21	104	104
2HatchLurkerAllIn	2	0%	-	11:36	36	73
2HatchMutaPure	11	36%	0:01	10:45	165	195
2x10HatchAllIn	9	11%	0:01	10:00	55	194
2x10HatchSlow	1	0%	-	9:59	10	10
3GasMuta	34	24%	7:39	12:03	18	198
3HatchHydraExpo	2	0%	-	13:44	5	109
3HatchLing	1	0%	-	9:49	1	1
3HatchLingExpo	1	0%	-	9:55	190	190
3HatchLurker	1	0%	-	9:55	146	146
4HatchBeforeGas	1	0%	-	10:29	126	126
4PoolSoft	1	0%	-	9:54	17	17
5HatchPoolLing	1	0%	-	11:54	63	63
5PoolHard2Player	1	0%	-	10:03	169	169
6PoolSpeed	7	0%	-	10:04	9	97
7PoolMid	1	0%	-	9:24	94	94
8-8HydraRush	11	9%	7:03	8:01	60	189
8Hatch7PoolBurrowB	1	0%	-	9:16	123	123
9GasLair	1	0%	-	8:18	74	74
9PoolFastLair	1	0%	-	8:40	156	156
9PoolFastLurker	1	0%	-	12:37	80	80
9PoolFastSpireB	1	0%	-	9:03	24	24
9PoolHatch	1	0%	-	9:25	100	100
9PoolLair	1	0%	-	13:53	199	199
9PoolLurker	1	0%	-	11:44	81	81
9PoolSpeed	1	0%	-	13:58	159	159
9PoolSpeedAllIn	1	0%	-	12:02	6	6
9PoolSpire	1	0%	-	9:58	136	136
9PoolSunkBurrow	8	12%	0:01	9:46	66	162
9PoolSunkHatch	1	0%	-	13:06	103	103
9PoolSunkSpeed	1	0%	-	9:29	37	37
AntiZeal_12Hatch	1	0%	-	9:56	27	27
DefilerRush	1	0%	-	26:43	99	99
HiveRush	1	0%	-	22:44	186	186
Over10Hatch1Sunk	3	0%	-	10:26	2	107
Over10Hatch2Sunk	2	0%	-	11:25	50	111
Over10HatchBust	2	0%	-	9:20	0	52
Over10HatchHydra	1	0%	-	11:45	44	44
Over10HatchSlowLings	1	0%	-	12:30	39	39
Over10PoolMuta	2	0%	-	8:05	108	140
OverhatchExpoLing	1	0%	-	9:18	29	29
OverpoolFastLair	1	0%	-	10:35	130	130
OverpoolSpeed	1	0%	-	8:54	8	8
OverpoolTurtle 0	1	0%	-	12:07	11	11
Overpool_3HatchLing	16	31%	0:01	10:40	43	157
Overpool_3HatchSpeed	1	0%	-	9:42	40	40
Sparkle 1HatchMuta	34	47%	17:34	15:45	112	197
Sparkle 3HatchMuta	1	0%	-	9:05	168	168
ZvP_Overpool3Hatch	1	0%	-	8:43	171	171
ZvT_2HatchMuta	1	0%	-	15:44	90	90
ZvT_3HatchMuta	1	0%	-	11:28	187	187
ZvZ_12Gas11Pool	1	0%	-	7:32	192	192
ZvZ_12HatchExpo	1	0%	-	9:19	22	22
ZvZ_12HatchMain	1	0%	-	10:21	163	163
ZvZ_Overpool9Gas	1	0%	-	9:55	30	30
58 openings	200	18%	15:17	10:48

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Heavy rush	191	96%	18%	56	28%	5%	29%	70%
Turtle	9	4%	33%	2	1%	0%	0%	89%
Unknown		-	-	142	71%	24%	0%	0%

timing	#	median	early	late
my combat unit	186	3:10	1:49	5:19
my gas	185	3:01	1:18	7:25
enemy scout	186	1:57	1:18	6:01
enemy combat unit	186	2:46	2:21	5:14
enemy gas	186	3:57	3:05	7:53
enemy air unit	186	4:05	3:06	10:13
enemy cloaked unit	62	11:06	6:57	17:09
game duration	200	10:56	0:01	26:43

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	12	2:04	1:53	2:15	8%	3:56
gas steal success	6	2:04	1:56	2:21	0%	4:17
none or failed	194	-	-	-	19%	3:56
gas steal killed	6	2:45	2:36	2:57

I expected to score zero against Stardust, or next to it. As it turned out, mutalisk openings held out some hope, and not all wins were due to Stardust crashing (the “wins” timing column gives the median duration of winning games).

#3 dragon

opening	games	wins	wins	losses	first	last
3HatchPoolMuta	7	57%	14:29	26:35	67	177
5HatchPool	51	90%	14:20	28:30	0	194
9PoolFastLurker	10	70%	10:36	31:33	109	191
9PoolHatchSpeed	1	0%	-	16:12	7	7
9PoolSpire	1	0%	-	16:30	54	54
AntiFact_13Pool	15	60%	17:22	22:12	2	180
AntiFact_Overpool11Hatch	59	90%	16:51	24:11	20	196
OverpoolLurker	1	0%	-	21:47	74	74
OverpoolSpeed	1	0%	-	15:01	45	45
ZvP_4HatchPoolHydra	16	75%	14:25	27:12	4	197
ZvT_3HatchMutaExpo	38	84%	16:25	23:31	3	199
11 openings	200	82%	15:45	23:35

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Factory	24	12%	75%	20	10%	75%	21%	29%
Heavy rush	3	2%	67%	6	3%	100%	0%	67%
Naked expand	14	7%	79%	16	8%	81%	7%	43%
Safe expand	90	45%	80%	46	23%	83%	30%	34%
Turtle		-	-	3	2%	100%	0%	0%
Unknown		-	-	67	34%	79%	0%	0%
Worker rush	69	34%	87%	42	21%	83%	43%	30%

timing	#	median	early	late
my combat unit	200	3:10	2:13	8:17
my gas	168	3:39	1:46	17:58
enemy scout	193	2:14	1:03	13:37
enemy combat unit	169	2:43	2:21	8:46
enemy gas	165	6:06	2:34	14:09
enemy air unit	163	9:29	5:33	18:13
enemy cloaked unit	115	10:50	7:22	24:22
game duration	200	16:24	3:42	51:19

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	10	2:44	2:24	3:06	80%	6:29
gas steal success	0	-	-	-	-	-
none or failed	200	-	-	-	82%	6:06
gas steal killed	0	-	-	-

Since Dragon was a carryover bot, and Steamhammer was not much changed from last year, my preparation was to drop in last year’s learning file unchanged. The Steamhammer fixes I did make improved defense substantially, though, so Steamhammer in most cases scored better playing the same openings. It discovered AntiFact_Overpool11Hatch on its own, though.

Dragon never allowed a gas steal. And it played the longest games.

#5 purplewave

opening	games	wins	wins	losses	first	last
10Hatch	1	0%	-	12:13	73	73
10HatchBurrow	1	0%	-	11:24	24	24
11HatchTurtleMuta	52	69%	20:14	16:15	68	194
11Pool	1	0%	-	28:54	106	106
2.5HatchMuta	1	0%	-	6:46	172	172
6PoolHide	1	0%	-	31:13	37	37
7-7HydraRush	1	0%	-	10:23	149	149
7Pool12Hatch	11	73%	10:13	21:27	167	198
9GasLair	8	25%	24:21	11:51	66	164
9PoolHatchSpeedAllIn	1	0%	-	9:27	17	17
9PoolSpeedAllIn	26	42%	24:40	13:19	29	192
9PoolSunkHatch	13	23%	25:24	16:08	11	147
AntiZeal_12Hatch	6	50%	21:43	14:02	161	184
Over10Hatch	34	35%	25:13	22:41	1	199
Over10Hatch1Sunk	1	0%	-	43:24	4	4
Over10Hatch2Sunk	2	0%	-	21:01	3	196
Over10HatchBurrow	1	0%	-	15:52	97	97
Over10HatchBust	18	44%	7:59	12:13	0	166
Over10HatchSlowLings	1	0%	-	10:48	7	7
OverhatchLateGas	9	33%	24:14	13:28	135	195
OverpoolBurrow	1	0%	-	10:42	152	152
OverpoolHide	6	17%	8:39	24:27	59	176
OverpoolTurtle 0	1	0%	-	11:23	133	133
UltraRush	1	0%	-	9:28	25	25
ZvP_3BaseSpire+Den	1	0%	-	19:12	27	27
ZvT_12PoolMuta	1	0%	-	21:27	158	158
26 openings	200	44%	21:45	15:45

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Heavy rush	200	100%	44%	155	78%	44%	78%	9%
Naked expand		-	-	16	8%	69%	0%	0%
Safe expand		-	-	2	1%	50%	0%	0%
Turtle		-	-	9	4%	11%	0%	0%
Unknown		-	-	18	9%	33%	0%	0%

timing	#	median	early	late
my combat unit	200	3:03	1:58	4:02
my gas	199	4:02	1:19	7:09
enemy scout	199	1:35	1:18	8:06
enemy combat unit	200	2:43	2:22	3:46
enemy gas	190	6:29	3:54	9:46
enemy air unit	173	9:19	6:30	15:02
enemy cloaked unit	144	12:13	6:42	17:31
game duration	200	19:33	5:38	60:00

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	41	2:35	2:05	3:15	41%	6:18
gas steal success	31	2:40	2:17	3:20	39%	6:13
none or failed	169	-	-	-	44%	6:33
gas steal killed	31	3:17	3:03	4:06

For PurpleWave, I gave hints that 3HatchHydraExpo might lose and Over10Hatch might win. On BASIL, PurpleWave had been playing the same opening over and over versus Steamhammer, and I doubted it would do the same in the tournament, so I didn’t have much to go on. Steamhammer made interesting and surprising choices. I would not have guessed that the crazy 7Pool12Hatch would succeed. It’s too bad that the bot discovered it so late in the tournament.

Steamhammer tended to get gas late, a sign that it was either under pressure or else putting on its own pressure instead of teching.

#6 mcrave

opening	games	wins	wins	losses	first	last
10Hatch	1	0%	-	5:57	45	45
11Gas10PoolMuta	1	0%	-	6:23	14	14
11HatchTurtleLurker	1	0%	-	9:11	192	192
12Hatch13Pool	2	0%	-	5:53	28	157
2HatchLingAllInSpire	1	0%	-	5:35	118	118
3HatchHydraBust	1	0%	-	5:25	40	40
3HatchLingBurrow	1	0%	-	5:37	50	50
3HatchPoolHydraAllIn	1	0%	-	5:47	57	57
4PoolHard	1	0%	-	8:48	188	188
4PoolSoft	1	0%	-	7:58	21	21
5PoolHard2Player	1	0%	-	8:36	114	114
5Scout	7	29%	5:35	9:01	31	162
7Pool12Hatch	1	0%	-	8:58	47	47
7Pool6GasLurker A	1	0%	-	5:23	143	143
7Pool6GasLurker B	2	0%	-	5:23	38	184
8Hatch7PoolBurrowB	1	0%	-	5:33	78	78
9Hatch8Pool	1	0%	-	5:25	121	121
9Pool9Hatch	1	0%	-	5:50	195	195
9PoolBurrow	1	0%	-	9:54	102	102
9PoolHatchSpeed7Drone	1	0%	-	5:29	142	142
9PoolHatchSpeedAllIn	34	53%	5:48	9:01	15	199
9PoolHatchSpeedLurker	1	0%	-	9:59	126	126
9PoolSunkHatch	2	0%	-	8:49	12	124
AntiFact_Overpool9Gas	1	0%	-	5:15	89	89
Over10Hatch	10	10%	5:48	5:38	83	133
Over10Hatch11Pool	1	0%	-	5:25	80	80
Over10Hatch2Sunk	1	0%	-	12:07	103	103
Over10HatchBurrow	1	0%	-	8:18	27	27
Overgas+1	1	0%	-	5:26	98	98
Overpool+1	1	0%	-	6:15	146	146
OverpoolTurtle 0	1	0%	-	14:10	30	30
ZvP_4HatchPoolHydra	1	0%	-	5:39	90	90
ZvZ_12Gas11Pool	23	35%	11:38	5:30	62	182
ZvZ_12HatchMain	1	0%	-	8:38	153	153
ZvZ_12PoolLing	1	0%	-	9:28	163	163
ZvZ_Overgas9Pool	14	29%	10:47	5:41	2	194
ZvZ_OverpoolTurtle	78	67%	9:34	12:39	0	198
37 openings	200	42%	9:12	8:18

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	73	36%	44%	16	8%	50%	7%	79%
Turtle	127	64%	42%	29	14%	48%	15%	76%
Unknown		-	-	155	78%	41%	0%	0%

timing	#	median	early	late
my combat unit	196	2:26	1:47	3:45
my gas	183	2:06	1:30	7:06
enemy scout	183	2:53	0:42	8:23
enemy combat unit	200	2:27	0:41	3:42
enemy gas	195	3:09	2:32	6:03
enemy air unit	187	4:35	3:38	6:22
enemy cloaked unit	0	-	-	-
game duration	200	8:44	5:12	19:17

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	14	2:11	0:45	3:21	21%	3:28
gas steal success	1	0:49	0:49	0:49	100%	-
none or failed	199	-	-	-	42%	3:09
gas steal killed	1	4:33	4:33	4:33

For McRave I gave hints suggesting 5Scout and ZvZ_OverpoolTurtle. 5Scout seriously has scored well against McRave, but not this time—though it does seem to have brought one winning gas steal.

#7 microwave

opening	games	wins	wins	losses	first	last
10HatchLing	9	56%	6:37	12:38	154	185
12-13Hatch	1	0%	-	4:39	79	79
12HatchTurtle	1	0%	-	14:19	99	99
2HatchLingAllInSpire	14	64%	11:32	6:14	128	180
3HatchLurker	1	0%	-	4:40	117	117
4HatchBeforeGas	1	0%	-	7:11	124	124
5PoolHard	1	0%	-	10:25	80	80
6PoolSpeedBurrow	1	0%	-	6:32	187	187
9Pool8Hatch	1	0%	-	10:31	70	70
9PoolHatchSpeed7Drone	8	100%	6:10	-	192	199
9PoolHatchSpeedAllIn	24	67%	6:15	11:06	52	179
9PoolHatchSpeedAllInB	1	0%	-	12:22	190	190
OverhatchMuta	21	67%	9:58	5:04	0	177
OverpoolSpeed	25	72%	12:40	10:36	2	189
ZvT_3HatchMutaExpo	1	0%	-	5:34	119	119
ZvZ_12Gas11Pool	1	0%	-	5:20	89	89
ZvZ_12HatchMain	10	50%	12:38	14:42	20	191
ZvZ_Overpool9Gas	79	91%	13:52	22:10	3	186
18 openings	200	74%	11:47	10:52

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	46	23%	83%	26	13%	77%	9%	48%
Heavy rush	15	8%	80%	13	6%	92%	0%	60%
Naked expand	107	54%	68%	52	26%	77%	27%	38%
Turtle	32	16%	75%	15	8%	47%	3%	69%
Unknown		-	-	94	47%	72%	0%	0%

timing	#	median	early	late
my combat unit	196	2:26	1:51	4:07
my gas	195	2:06	1:49	5:23
enemy scout	194	2:26	1:17	8:55
enemy combat unit	200	2:34	1:01	3:25
enemy gas	157	4:04	2:15	11:59
enemy air unit	127	5:57	3:34	14:17
enemy cloaked unit	3	4:57	4:54	7:38
game duration	200	11:27	4:06	37:54

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	10	2:11	1:55	2:48	40%	4:28
gas steal success	0	-	-	-	-	-
none or failed	200	-	-	-	74%	4:04
gas steal killed	0	-	-	-

For Microwave I prepared four different ZvZ openings that were successful on BASIL. All were reasonably effective, and ZvZ_Overpool9Gas was excellent. Steamhammer discovered 9PoolHatchSpeed7Drone on its own, very late in the tournament, and won the last 8 games straight with it. Overall, zerglings mostly worked out better than mutalisks, but the median game duration suggests that the win often came after switching to mutas.

#8 ualbertabot

opening	games	wins	wins	losses	first	last
OverpoolTurtle	200	96%	5:55	20:16	0	199
1 openings	200	96%	5:55	20:16

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Factory	28	14%	100%	31	16%	100%	7%	21%
Fast rush	21	10%	95%	29	14%	100%	5%	33%
Heavy rush	129	64%	96%	65	32%	98%	33%	25%
Naked expand	22	11%	95%	24	12%	100%	18%	27%
Unknown		-	-	51	26%	88%	0%	0%

timing	#	median	early	late
my combat unit	200	2:26	2:22	2:37
my gas	200	2:58	2:55	4:25
enemy scout	179	2:10	1:18	6:38
enemy combat unit	169	2:46	1:46	4:14
enemy gas	184	3:30	2:31	14:48
enemy air unit	33	13:33	5:51	16:47
enemy cloaked unit	19	14:55	2:38	17:21
game duration	200	5:57	3:31	34:14

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	39	1:56	1:45	2:41	100%	3:33
gas steal success	15	2:06	1:58	2:41	100%	11:43
none or failed	185	-	-	-	96%	3:29
gas steal killed	15	3:03	2:49	6:56

For UAlbertaBot I did not provide hints, but configured Steamhammer to always play its best known choice. It’s an extreme turtle build that holds on against all of UAlbertaBot’s rushes and then strikes back. Any experimentation would have lowered the win rate.

#9 pylonpuller

opening	games	wins	wins	losses	first	last
12HatchTurtle	24	88%	11:16	8:56	22	165
12Hatch_4HatchLing	27	96%	6:45	8:30	170	197
3HatchHydra	27	67%	11:17	8:31	0	130
9HatchExpo9Pool9Gas	19	74%	6:13	11:16	82	156
9PoolHatch	24	79%	8:00	9:40	9	199
AntiFact_2Hatch	5	40%	12:59	8:14	97	117
AntiZeal_12Hatch	40	88%	11:00	8:43	6	193
Over10HatchBurrow	1	0%	-	12:38	129	129
ZvZ_OverpoolTurtle	33	85%	12:01	12:29	132	169
9 openings	200	82%	10:23	9:37

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	2	1%	100%	12	6%	8%	0%	50%
Heavy rush	137	68%	80%	84	42%	79%	49%	25%
Turtle	59	30%	85%	40	20%	90%	22%	44%
Unknown		-	-	62	31%	94%	0%	0%
Worker rush	2	1%	100%	2	1%	100%	0%	50%

timing	#	median	early	late
my combat unit	200	3:09	2:13	8:13
my gas	197	3:41	2:01	6:05
enemy scout	125	3:55	1:22	9:09
enemy combat unit	178	3:27	2:21	7:23
enemy gas	87	4:21	2:52	14:14
enemy air unit	55	4:29	2:53	14:15
enemy cloaked unit	33	5:42	3:18	11:45
game duration	200	10:17	4:19	24:00

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	71	2:08	1:46	3:07	90%	3:53
gas steal success	40	2:19	2:05	3:00	88%	8:35
none or failed	160	-	-	-	80%	4:18
gas steal killed	40	5:15	2:57	17:53

I made no prep. There wasn’t data, the opponent was being updated too often.

#10 styx

opening	games	wins	wins	losses	first	last
9PoolSunkHatch	99	96%	5:17	7:09	0	199
9PoolSunkSpeed	22	91%	4:43	10:18	148	194
Over10PoolMuta	9	78%	6:57	5:24	39	198
OverpoolSunk	17	94%	5:32	5:59	8	172
ZvT_12PoolMuta	28	86%	6:12	5:23	57	196
ZvZ_12PoolLing	25	80%	5:57	6:27	7	161
6 openings	200	91%	5:27	6:26

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	182	91%	91%	103	52%	97%	51%	29%
Heavy rush	18	9%	89%	40	20%	88%	17%	22%
Unknown		-	-	57	28%	82%	0%	0%

timing	#	median	early	late
my combat unit	199	2:17	2:13	2:49
my gas	189	4:06	2:06	5:50
enemy scout	173	2:31	0:37	5:39
enemy combat unit	188	2:13	1:13	3:43
enemy gas	0	-	-	-
enemy air unit	0	-	-	-
enemy cloaked unit	0	-	-	-
game duration	200	5:37	3:25	14:12

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	0	-	-	-	-	-
gas steal success	0	-	-	-	-	-
none or failed	200	-	-	-	91%	-
gas steal killed	0	-	-	-

No prep needed.

#11 cunybot

opening	games	wins	wins	losses	first	last
11Gas10PoolMuta	10	100%	7:59	-	45	194
4PoolHard	4	75%	4:54	4:22	9	132
4PoolSoft	5	80%	6:48	6:54	54	134
5PoolHard	8	100%	5:25	-	82	193
5PoolHard2Player	2	50%	5:01	9:42	78	113
5PoolSoft	5	100%	5:11	-	110	181
6PoolSpeed	10	100%	4:55	-	8	138
9PoolHatchSpeedAllIn	6	100%	5:14	-	144	196
9PoolHatchSpeedAllInB	14	100%	5:12	-	5	178
9PoolSunkHatch	20	100%	7:18	-	0	177
9PoolSunkSpeed	12	100%	5:56	-	31	190
Over10PoolLing	2	100%	6:22	-	160	182
OverhatchLing	14	100%	5:47	-	3	170
OverhatchMuta	16	94%	6:21	4:50	15	192
OverpoolSunk	11	100%	5:20	-	1	198
ZvZ_12HatchExpo	12	100%	6:37	-	12	191
ZvZ_12HatchMain	14	100%	8:04	-	33	197
ZvZ_12PoolLing	6	100%	5:49	-	95	165
ZvZ_12PoolMain	10	100%	6:39	-	4	199
ZvZ_Overgas9Pool	9	89%	7:43	15:50	85	189
ZvZ_OverpoolTurtle	10	90%	10:30	9:46	13	107
21 openings	200	97%	6:27	8:18

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	52	26%	96%	48	24%	92%	25%	17%
Heavy rush	17	8%	100%	28	14%	100%	12%	29%
Naked expand	131	66%	97%	70	35%	99%	29%	31%
Unknown		-	-	54	27%	98%	0%	0%

timing	#	median	early	late
my combat unit	200	2:27	1:47	3:17
my gas	182	2:41	1:30	7:51
enemy scout	189	2:59	1:31	6:43
enemy combat unit	198	2:34	1:02	4:42
enemy gas	62	3:45	2:43	11:53
enemy air unit	19	4:27	3:51	14:19
enemy cloaked unit	0	-	-	-
game duration	200	6:27	3:37	23:51

gas steal	#	median	early	late	wins	enemy gas
gas steal decision	0	-	-	-	-	-
gas steal success	0	-	-	-	-	-
none or failed	200	-	-	-	97%	3:45
gas steal killed	0	-	-	-

No prep needed. The win percentages say that Steamhammer immediately switched away from every opening that lost once.

overall

	total		ZvT		ZvP		ZvZ		ZvR
opening	games	wins	games	wins	games	wins	games	wins	games	wins
10Hatch	2	0%			1	0%	1	0%
10HatchBurrow	1	0%			1	0%
10HatchHydra	2	0%			2	0%
10HatchLing	10	50%			1	0%	9	56%
11Gas10PoolLurker	1	0%			1	0%
11Gas10PoolLurkerB	1	0%			1	0%
11Gas10PoolMuta	11	91%					11	91%
11HatchTurtleHydra	3	0%			3	0%
11HatchTurtleLurker	1	0%					1	0%
11HatchTurtleMuta	65	57%			65	57%
11Pool	1	0%			1	0%
12-11Hatch	1	0%			1	0%
12-12Hatch	1	0%			1	0%
12-13Hatch	1	0%					1	0%
12Hatch12Pool	1	0%			1	0%
12Hatch13Pool	3	0%			1	0%	2	0%
12HatchTurtle	26	81%			25	84%	1	0%
12Hatch_4HatchLing	27	96%			27	96%
2.5HatchMuta	1	0%			1	0%
2HatchLingAllInSpire	15	60%					15	60%
2HatchLurkerAllIn	2	0%			2	0%
2HatchMutaPure	11	36%			11	36%
2x10HatchAllIn	9	11%			9	11%
2x10HatchSlow	1	0%			1	0%
3GasMuta	34	24%			34	24%
3HatchHydra	27	67%			27	67%
3HatchHydraBust	1	0%					1	0%
3HatchHydraExpo	3	0%			3	0%
3HatchLing	2	0%			2	0%
3HatchLingBurrow	1	0%					1	0%
3HatchLingBust	47	26%			47	26%
3HatchLingExpo	10	10%			10	10%
3HatchLurker	2	0%			1	0%	1	0%
3HatchPoolHydraAllIn	1	0%					1	0%
3HatchPoolMuta	7	57%	7	57%
4HatchBeforeGas	2	0%			1	0%	1	0%
4HatchBeforeLair	1	0%			1	0%
4PoolHard	6	50%			1	0%	5	60%
4PoolSoft	7	57%			1	0%	6	67%
5HatchBeforeGas	13	8%			13	8%
5HatchPool	51	90%	51	90%
5HatchPoolLing	2	0%			2	0%
5HatchPoolLingBurrow	30	17%			30	17%
5PoolHard	9	89%					9	89%
5PoolHard2Player	5	20%			2	0%	3	33%
5PoolSoft	5	100%					5	100%
5Scout	7	29%					7	29%
6PoolBurrow	1	0%			1	0%
6PoolHide	3	0%			3	0%
6PoolSpeed	18	56%			8	0%	10	100%
6PoolSpeedBurrow	1	0%					1	0%
7-7HydraRush	2	0%			2	0%
7Pool12Hatch	12	67%			11	73%	1	0%
7Pool6GasLurker A	1	0%					1	0%
7Pool6GasLurker B	2	0%					2	0%
7PoolHard	1	0%			1	0%
7PoolMid	1	0%			1	0%
8-8HydraRush	11	9%			11	9%
8Gas7PoolLurker B	1	0%			1	0%
8Hatch7Pool	1	0%			1	0%
8Hatch7PoolBurrowB	2	0%			1	0%	1	0%
8Pool	1	0%			1	0%
9GasLair	9	22%			9	22%
9Hatch8Pool	1	0%					1	0%
9HatchExpo9Pool9Gas	20	70%			20	70%
9HatchMain9Pool9Gas	19	16%			19	16%
9Pool	2	0%			2	0%
9Pool8Hatch	1	0%					1	0%
9Pool9Hatch	1	0%					1	0%
9PoolBurrow	1	0%					1	0%
9PoolExpo	1	0%			1	0%
9PoolFastLair	1	0%			1	0%
9PoolFastLurker	11	64%	10	70%	1	0%
9PoolFastSpireB	1	0%			1	0%
9PoolHatch	26	73%			26	73%
9PoolHatchBurrow	1	0%			1	0%
9PoolHatchSpeed	1	0%	1	0%
9PoolHatchSpeed7Drone	10	80%			1	0%	9	89%
9PoolHatchSpeed7DroneB	1	0%			1	0%
9PoolHatchSpeedAllIn	65	62%			1	0%	64	62%
9PoolHatchSpeedAllInB	15	93%					15	93%
9PoolHatchSpeedLurker	1	0%					1	0%
9PoolHatchSpeedSpire2	1	0%			1	0%
9PoolHatchSpire	1	0%			1	0%
9PoolLair	3	0%			3	0%
9PoolLurker	2	0%			2	0%
9PoolSpeed	2	0%			2	0%
9PoolSpeedAllIn	29	38%			29	38%
9PoolSpire	3	0%	1	0%	2	0%
9PoolSunkBurrow	8	12%			8	12%
9PoolSunkHatch	137	86%			16	19%	121	95%
9PoolSunkSpeed	36	89%			2	0%	34	94%
AntiFact_13Pool	15	60%	15	60%
AntiFact_2Hatch	6	33%			6	33%
AntiFact_Overpool11Hatch	59	90%	59	90%
AntiFact_Overpool9Gas	1	0%					1	0%
AntiFactoryHydra	1	0%			1	0%
AntiStyx_9Pool	2	0%			2	0%
AntiStyx_Muta	1	0%			1	0%
AntiZeal_12Hatch	48	79%			48	79%
DefilerRush	2	0%			2	0%
HiveRush	1	0%			1	0%
Over10Hatch	46	28%			36	33%	10	10%
Over10Hatch11Pool	1	0%					1	0%
Over10Hatch1Sunk	4	0%			4	0%
Over10Hatch2Sunk	7	0%			6	0%	1	0%
Over10HatchBurrow	3	0%			2	0%	1	0%
Over10HatchBust	20	40%			20	40%
Over10HatchHydra	1	0%			1	0%
Over10HatchSlowLings	2	0%			2	0%
Over10PoolHydra	2	0%			2	0%
Over10PoolLing	2	100%					2	100%
Over10PoolMuta	11	64%			2	0%	9	78%
Overgas+1	1	0%					1	0%
OverhatchExpoLing	3	0%			3	0%
OverhatchExpoMuta	2	0%			2	0%
OverhatchLateGas	9	33%			9	33%
OverhatchLing	16	88%			2	0%	14	100%
OverhatchMuta	37	78%					37	78%
Overpool+1	2	0%			1	0%	1	0%
OverpoolBurrow	1	0%			1	0%
OverpoolFastLair	1	0%			1	0%
OverpoolHide	7	14%			7	14%
OverpoolLurker	2	0%	1	0%	1	0%
OverpoolSpeed	28	64%	1	0%	2	0%	25	72%
OverpoolSunk	28	96%					28	96%
OverpoolTurtle	200	96%							200	96%
OverpoolTurtle 0	3	0%			2	0%	1	0%
Overpool_3HatchLing	16	31%			16	31%
Overpool_3HatchSpeed	1	0%			1	0%
Phlegethon	10	20%			10	20%
Sparkle 1HatchMuta	34	47%			34	47%
Sparkle 3HatchMuta	1	0%			1	0%
UltraRush	1	0%			1	0%
ZvP_2HatchFakeHydra	1	0%			1	0%
ZvP_2HatchFakeMuta	1	0%			1	0%
ZvP_3BaseSpire+Den	1	0%			1	0%
ZvP_3HatchPoolHydra	1	0%			1	0%
ZvP_4HatchPoolHydra	17	71%	16	75%			1	0%
ZvP_Overpool3Hatch	2	0%			2	0%
ZvT_12PoolMuta	29	83%			1	0%	28	86%
ZvT_13Pool	1	0%			1	0%
ZvT_2HatchMuta	1	0%			1	0%
ZvT_3HatchMuta	1	0%			1	0%
ZvT_3HatchMutaExpo	39	82%	38	84%			1	0%
ZvZ_12Gas11Pool	25	32%			1	0%	24	33%
ZvZ_12HatchExpo	13	92%			1	0%	12	100%
ZvZ_12HatchMain	26	73%			1	0%	25	76%
ZvZ_12PoolLing	32	81%					32	81%
ZvZ_12PoolMain	10	100%					10	100%
ZvZ_Overgas9Pool	23	52%					23	52%
ZvZ_Overpool9Gas	80	90%			1	0%	79	91%
ZvZ_OverpoolTurtle	121	74%			33	85%	88	69%
total	2000	64%	200	82%	800	39%	800	76%	200	96%
openings played	153		11		117		60		1

Trackbacks

No Trackbacks

Comments

Dan on Monday, November 7. 2022:

I'm gonna hazard a guess based on etymology that Phlegethon is nine pool into two hatch speedling.

Jay Scott on Monday, November 7. 2022:

Phlegethon is 10 hatch 10 pool third hatchery into speedlings. I did add a link to my analysis showing that it is inferior to standard three hatch ling.

Dan on Monday, November 7. 2022:

Ah, I missed it in the post. Thanks!

Dan on Monday, November 7. 2022:

You were correct in anticipating PurpleWave would do something different in the tournament. I've long considered its PvZ strategies band-aids until I had competent FFE/GE, a goal which continues its viscous buttery habit of sliding down my priority list whenever it gets too close to the top. So I rewrote the PvZ strategy about 24 hours before the deadline after the last band-aid strategy had grown too complex to debug without undue attention. It's a bit of an indictment of one's old strategy when the new one, barely tested, outperforms it.

Jay Scott on Monday, November 7. 2022:

Ha ha, viscous buttery habit! If it keeps dropping, then don’t put it on your toast!

Dan on Monday, November 7. 2022:

Exactly what a Zerg author allergic to buttered toast might say!

Tully Elliston on Sunday, November 13. 2022:

For defeats, is the game length tracked?

For the opponents that SH loses too most the time, builds that last 15-20 minutes (eg. 11Gas10PoolLurkerB vs BanannaBrain) are likely to actually be good builds. The learning dismissing the build and moving to the next in these cases might be missing a trick.

Jay Scott on Sunday, November 13. 2022:

I plan to solve that not with the weak signal of game length but with the strong signal from my ML system, which is ready to integrate into the code.

Add Comment

Name*

Homepage

Comment*

In reply to

E-Mail addresses will not be displayed and will only be used for E-Mail notifications.

To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly.
CAPTCHA