AIIDE 2020 - what DaQin learned

Holdover bot DaQin is based on Locutus and writes game records in a format close to old-style Steamhammer game records. The same script parses both Locutus and DaQin files, and is only slightly different from my original Steamhammer code. But DaQin plays a more restricted set of builds.

#1 stardust

opening	games	wins
2GateDT	6	0%
3GateDT	133	8%
4GateGoon	11	0%
3 openings	150	7%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
DarkTemplar rush	112	75%	7%	89	59%	3%	58%	1%
Fast rush	11	7%	0%	16	11%	6%	0%	0%
Not fast rush	26	17%	12%	44	29%	16%	23%	0%
Unknown	1	1%	0%	1	1%	0%	0%	0%

timing	#	median	early	late
gas steal attempt	63	1:43	0:46	1:47
gas steal success	2	-	-	-
enemy scout	150	1:55	1:17	2:34
enemy combat units	150	3:07	2:22	5:42
enemy air units	126	8:02	5:35	12:26
enemy cloaked units	11	8:13	7:06	22:38

It’s interesting that DaQin settled on a dark templar strategy. DaQin seems poor at recognizing the enemy strategy. Given these classes, I think all games would have been best classified as Not fast rush.

#2 purplewave

opening	games	wins
2GateDT	7	43%
3GateDT	3	0%
4GateGoon	140	11%
3 openings	150	13%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
DarkTemplar rush	12	8%	33%	13	9%	23%	50%	0%
Fast rush	137	91%	11%	136	91%	11%	95%	0%
Naked expand		-	-	1	1%	100%	0%	0%
Unknown	1	1%	0%		-	-	0%	0%

timing	#	median	early	late
gas steal attempt	8	0:46	0:45	0:47
gas steal success	5	-	-	-
enemy scout	150	2:07	1:18	3:47
enemy combat units	150	3:02	2:21	7:50
enemy air units	14	6:07	3:41	19:42
enemy cloaked units	93	6:47	5:14	20:05

BananaBrain chose 4 gate goon as best against both Stardust and PurpleWave. DaQin liked the dragoons only against PurpleWave. This table doesn’t say so, but 2GateDT was tried on the tournament’s 3rd round out of 150 and won then, while 4GateGoon lost on its first outing. I would have to read the code to decipher the strange seeming preference for dragoons.

#3 bananabrain

opening	games	wins
2GateDT	14	36%
3GateDT	89	40%
4GateGoon	47	21%
3 openings	150	34%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
DarkTemplar rush	65	43%	34%	72	48%	40%	43%	2%
Fast rush	79	53%	35%	63	42%	25%	41%	3%
Naked expand	2	1%	0%	3	2%	0%	0%	0%
Not fast rush	1	1%	0%	4	3%	75%	0%	100%
Proxy	2	1%	50%	3	2%	0%	0%	0%
Safe expand		-	-	1	1%	100%	0%	0%
Unknown	1	1%	0%	4	3%	50%	0%	0%

timing	#	median	early	late
gas steal attempt	60	1:42	0:46	1:47
gas steal success	9	-	-	-
enemy scout	150	1:58	0:53	3:11
enemy combat units	150	2:57	2:18	5:50
enemy air units	103	7:45	3:45	15:31
enemy cloaked units	38	6:11	5:29	22:03

#4 dragon

opening	games	wins
12NexusCarriers	106	58%
3GateDT	9	33%
4GateGoon	27	15%
DTDrop	8	25%
4 openings	150	47%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Factory	41	27%	56%	29	19%	62%	44%	15%
Naked expand	10	7%	30%	15	10%	67%	10%	10%
Not fast rush	35	23%	63%	24	16%	12%	3%	20%
Proxy	26	17%	15%	24	16%	21%	0%	8%
Safe expand	35	23%	46%	31	21%	48%	26%	20%
Unknown	2	1%	100%	24	16%	67%	0%	50%
Worker rush	1	1%	0%	3	2%	100%	0%	0%

timing	#	median	early	late
gas steal attempt	122	2:18	0:46	3:18
gas steal success	9	-	-	-
enemy scout	141	2:15	1:06	19:29
enemy combat units	148	2:59	2:30	6:37
enemy air units	132	9:29	7:58	14:01
enemy cloaked units	127	9:55	8:18	17:26

The carriers were successful versus Dragon. DaQin attempted to steal gas in almost all games, and rarely succeeded. The tables don’t show enough information to tell whether the attempts were worth it.

#5 mcrave

opening	games	wins
ForgeExpand5GateGoon	3	0%
ForgeExpandSpeedlots	147	36%
2 openings	150	35%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Heavy rush	14	9%	43%	28	19%	57%	14%	7%
Not fast rush	132	88%	36%	117	78%	31%	78%	2%
Unknown	4	3%	0%	5	3%	20%	0%	25%

timing	#	median	early	late
gas steal attempt	0	-	-	-
gas steal success	0	-	-	-
enemy scout	150	2:39	0:53	5:35
enemy combat units	150	3:02	2:25	6:17
enemy air units	150	6:21	5:23	7:01
enemy cloaked units	131	11:02	9:30	15:13

#6 microwave

opening	games	wins
4GateGoon	11	73%
ForgeExpand5GateGoon	3	0%
ForgeExpandSpeedlots	136	12%
3 openings	150	17%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush	1	1%	0%	1	1%	100%	0%	0%
Heavy rush	13	9%	8%	30	20%	17%	15%	0%
Not fast rush	124	83%	13%	103	69%	17%	69%	6%
Proxy	11	7%	73%	8	5%	12%	18%	0%
Unknown	1	1%	0%	8	5%	12%	0%	0%

timing	#	median	early	late
gas steal attempt	0	-	-	-
gas steal success	0	-	-	-
enemy scout	146	2:42	1:23	10:18
enemy combat units	150	3:56	2:47	6:45
enemy air units	133	6:17	4:43	16:55
enemy cloaked units	50	10:06	5:49	11:58

Another puzzling choice of a seemingly less-successful strategy...

#7 steamhammer

opening	games	wins
ForgeExpand5GateGoon	136	79%
ForgeExpandSpeedlots	14	71%
2 openings	150	78%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush		-	-	4	3%	100%	0%	0%
Heavy rush	27	18%	89%	33	22%	85%	15%	4%
Hydra bust	8	5%	62%	2	1%	100%	0%	0%
Not fast rush	114	76%	77%	104	69%	74%	68%	5%
Unknown	1	1%	0%	7	5%	86%	0%	0%

timing	#	median	early	late
gas steal attempt	0	-	-	-
gas steal success	0	-	-	-
enemy scout	127	2:11	0:59	36:17
enemy combat units	150	3:15	1:55	6:29
enemy air units	33	8:06	4:39	27:42
enemy cloaked units	33	12:02	4:54	23:23

... but not against Steamhammer. Steamhammer is far more skillful at fighting zealots than dragoons, and DaQin’s choice was correct here. (I expect that Steamhammer would have scored well against the speed zealots if DaQin had stuck to them.) Was Steamhammer’s poor showing relative to Microwave partly due to a mistake by DaQin that only happened versus Microwave?

#9 zzzkbot

opening	games	wins
ForgeExpand5GateGoon	147	10%
ForgeExpandSpeedlots	3	0%
2 openings	150	9%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Fast rush		-	-	5	3%	100%	0%	0%
Heavy rush	143	95%	6%	144	96%	6%	99%	1%
Unknown	7	5%	71%	1	1%	0%	0%	0%

timing	#	median	early	late
gas steal attempt	0	-	-	-
gas steal success	0	-	-	-
enemy scout	147	3:45	0:51	5:39
enemy combat units	150	2:47	1:46	3:49
enemy air units	9	7:47	7:43	8:21
enemy cloaked units	0	-	-	-

#10 ualbertabot

opening	games	wins
12NexusCarriers	1	0%
3GateDT	34	68%
4GateGoon	32	53%
DTDrop	1	0%
ForgeExpand5GateGoon	81	78%
5 openings	149	69%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
DarkTemplar rush	36	24%	61%	35	23%	60%	22%	0%
Factory	1	1%	100%	16	11%	94%	100%	0%
Fast rush	81	54%	78%	52	35%	69%	37%	1%
Heavy rush	3	2%	67%	6	4%	83%	0%	0%
Hydra bust		-	-	1	1%	100%	0%	0%
Naked expand		-	-	1	1%	100%	0%	0%
Not fast rush	19	13%	68%	26	17%	73%	5%	5%
Proxy	8	5%	25%	10	7%	30%	0%	0%
Unknown	1	1%	0%	2	1%	100%	0%	0%

timing	#	median	early	late
gas steal attempt	39	1:42	0:46	1:48
gas steal success	11	-	-	-
enemy scout	138	1:41	1:14	8:25
enemy combat units	147	3:57	1:34	6:54
enemy air units	7	6:54	5:50	15:14
enemy cloaked units	19	5:07	4:31	5:15

#11 willyt

opening	games	wins
12NexusCarriers	3	67%
3GateDT	147	97%
2 openings	150	96%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Factory		-	-	2	1%	100%	0%	0%
Heavy rush	11	7%	82%	8	5%	100%	0%	27%
Not fast rush	10	7%	100%	8	5%	100%	0%	40%
Safe expand	123	82%	98%	90	60%	94%	60%	27%
Unknown	6	4%	83%	42	28%	98%	0%	33%

timing	#	median	early	late
gas steal attempt	139	1:43	1:39	2:21
gas steal success	53	-	-	-
enemy scout	149	1:53	1:43	3:55
enemy combat units	150	3:05	2:38	5:06
enemy air units	65	17:30	8:27	29:58
enemy cloaked units	37	15:07	11:43	21:49

#12 ecgberht

opening	games	wins
12NexusCarriers	2	50%
3GateDT	148	100%
2 openings	150	99%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
Factory	88	59%	100%	54	36%	100%	23%	14%
Fast rush	24	16%	100%	27	18%	96%	17%	0%
Naked expand		-	-	2	1%	100%	0%	0%
Not fast rush	5	3%	100%	14	9%	100%	20%	0%
Proxy		-	-	1	1%	100%	0%	0%
Safe expand	31	21%	100%	37	25%	100%	19%	10%
Unknown	2	1%	50%	15	10%	100%	0%	0%

timing	#	median	early	late
gas steal attempt	126	1:43	1:38	2:21
gas steal success	29	-	-	-
enemy scout	140	1:35	0:34	3:55
enemy combat units	150	3:43	2:05	5:45
enemy air units	46	7:29	5:59	9:31
enemy cloaked units	15	8:41	7:35	9:54

#13 eggbot

opening	games	wins
3GateDT	16	94%
4GateGoon	134	97%
2 openings	150	97%

plan	predicted			recognized			accuracy
plan	count	games	wins	count	games	wins	good	?
DarkTemplar rush	104	69%	95%	90	60%	98%	58%	0%
Proxy	42	28%	100%	60	40%	95%	36%	0%
Unknown	4	3%	100%		-	-	0%	0%

timing	#	median	early	late
gas steal attempt	58	0:46	0:46	0:49
gas steal success	54	-	-	-
enemy scout	121	0:46	0:35	3:41
enemy combat units	29	6:34	4:39	22:57
enemy air units	0	-	-	-
enemy cloaked units	0	-	-	-

overall

	total		PvT		PvP		PvZ		PvR
opening	games	wins	games	wins	games	wins	games	wins	games	wins
12NexusCarriers	112	57%	111	58%					1	0%
2GateDT	27	30%			27	30%
3GateDT	579	65%	304	96%	241	26%			34	68%
4GateGoon	402	46%	27	15%	332	47%	11	73%	32	53%
DTDrop	9	22%	8	25%					1	0%
ForgeExpand5GateGoon	370	50%					289	42%	81	78%
ForgeExpandSpeedlots	300	27%					300	27%
total	1799	50%	450	81%	600	38%	600	35%	149	69%
openings played	7		4		3		3		5

Trackbacks

No Trackbacks

Comments

Bruce on Saturday, November 7. 2020:

If Stardust scouts an enemy two-gate early, it will build two gates before core, and a reaction to 4-gate is to hunker down and quickly build DTs, so the strategies observed by DaQin do somewhat make sense. Though of course they would not help at all for opening selection in later games!

I take full responsibility for DaQin’s poor opening selection in some of the matches - Locutus’ opening selection was basically just hacked together for the first tournament and not really changed much afterwards, which I think was a mistake (it’s a major weakness).

Jay Scott on Saturday, November 7. 2020:

Ah, I haven’t seen that reaction. I underestimated Stardust’s adaptivity.

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