AIIDE 2021 - what Dragon learned

Dragon records for each game only its own build and win/loss, so the information is sparse. It has a total of 7 builds. Dragon is a carryover bot, and I analyzed its game records from AIIDE 2020 last year. Dragon considers that “the opening” is a very brief phase of the game: It quickly adapts to what it sees of the opponent’s play, and the opening build fades out of view. Last year I found that, against many opponents, the choice of opening build made little difference; the game was decided later.

#1 stardust

opening	games	wins	first	last
1rax fe	24	0%	5	156
2rax bio	35	6%	1	153
2rax mech	18	0%	2	146
bio	20	5%	3	144
dirty worker rush	19	0%	0	133
mass vulture	20	0%	4	152
siege expand	21	0%	6	150
7 openings	157	2%

It’s interesting that the only openings to make a dent were “2rax bio” and “bio”. Was Stardust surprised by marines? If Stardust made zealots to get units faster, that may have backfired.

#2 bananabrain

opening	games	wins	first	last
1rax fe	8	0%	6	82
2rax bio	30	37%	1	156
2rax mech	8	0%	7	109
bio	13	8%	8	140
dirty worker rush	8	0%	5	148
mass vulture	36	25%	2	151
siege expand	54	35%	0	145
7 openings	157	25%

Again, the marines were a relatively successful choice against protoss. It’s a surprise. #8 DaQin below is different. Yesterday we saw that BananaBrain liked zealot openings against Dragon, and it’s true that marines with good micro can hold their own against zealots. Maybe BananaBrain liked zealots because they upset Dragon’s tech builds, and Dragon found that marines answered best, so that the two settled into this equilibrium with neither bot able to 100% counter the other. It’s a nice story, at least.

#4 steamhammer

opening	games	wins	first	last
1rax fe	53	45%	8	124
2rax bio	7	14%	0	153
2rax mech	12	25%	4	121
bio	17	29%	6	135
dirty worker rush	47	40%	2	156
mass vulture	15	27%	10	117
siege expand	6	0%	5	122
7 openings	157	36%

Switching between opposite builds like fast expand (“1rax fe”) and worker rush (“dirty worker rush”) is not a bad plan for defeating Steamhammer. I don’t think Dragon did it on purpose, though. Most openings scored about equal.

#5 mcrave

opening	games	wins	first	last
1rax fe	20	80%	34	153
2rax bio	40	65%	0	109
2rax mech	17	65%	23	90
bio	24	62%	64	100
dirty worker rush	2	0%	20	98
mass vulture	6	33%	63	137
siege expand	46	72%	11	154
7 openings	155	66%

Again, most were about equal, with only a couple of exceptions. “1rax fe” was also best against last year’s McRave, even though it played rather differently.

#6 willyt

opening	games	wins	first	last
2rax bio	1	0%	53	53
mass vulture	154	98%	0	154
2 openings	155	97%

Last year Dragon chose “2rax mech” as its build to trample on WillyT (winning fewer, 94%, even though this year’s WillyT is substantially stronger). I think it found something that worked and felt no need to experiment any further.

#7 microwave

opening	games	wins	first	last
1rax fe	24	67%	1	120
2rax bio	28	61%	31	152
2rax mech	66	74%	0	149
bio	26	62%	46	147
dirty worker rush	2	0%	8	124
mass vulture	5	40%	12	156
siege expand	6	33%	43	153
7 openings	157	65%

#8 daqin

opening	games	wins	first	last
1rax fe	98	56%	8	156
2rax bio	6	17%	14	104
2rax mech	11	36%	1	105
bio	13	31%	7	136
dirty worker rush	5	0%	5	153
mass vulture	20	50%	9	103
siege expand	4	0%	0	132
7 openings	157	47%

Best was the fast expansion. (“1rax fe” is faster than “siege expand”.) That makes sense against DaQin’s style of play.

#9 freshmeat

opening	games	wins	first	last
1rax fe	12	25%	4	128
2rax bio	12	33%	18	143
2rax mech	25	36%	0	141
bio	10	30%	9	139
dirty worker rush	34	35%	7	155
mass vulture	49	53%	10	156
siege expand	15	27%	15	142
7 openings	157	39%

Mostly about equal again. At some point I’ll look at the games and see how FreshMeat upset Dragon.

#10 ualbertabot

opening	games	wins	first	last
1rax fe	40	92%	116	155
2rax bio	6	67%	54	73
2rax mech	2	50%	114	115
bio	13	77%	36	49
dirty worker rush	17	76%	50	74
mass vulture	76	84%	0	113
siege expand	2	50%	85	110
7 openings	156	83%

Trackbacks

No Trackbacks

Comments

Jay Scott on Wednesday, October 20. 2021:

I watched the replays of Dragon’s 3 wins over Stardust, since there are so few. One was a Stardust crash, but it didn’t matter because Dragon was winning. Two were on Destination, where the 2 bridges to each side’s natural are easy to defend. Dragon got ahead when Stardust made repeated tactical misplays, like trying to squeeze its army past a strong terran ball. Dragon’s opening build didn’t seem to figure in. The other was a successful bunker rush of a fast expansion, putting Dragon way ahead. All were long games.

Tully Elliston on Thursday, October 21. 2021:

A fuzzy chance of selecting an opening with the specific criteria that it requires a drastically different response from your opponent than the previous opening seems like a prime way to mess with a learning bot (or meatbag).

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