14,425 comparisons among 10 agents.
One of 3 tables from the same source. The others are listed under Related tables.
| Description | Conference on Games (CoG) StarCraft AI Competition 2026: 14,425 one-on-one games of StarCraft: Brood War between 10 bot entries, one complete round robin run in August 2026 on six maps. The CoG competition had not been held for some years; David Churchill, who runs the AIIDE competition, took it over for 2026, so this is the first edition of a second annual series and its entrants overlap aiide_starcraft_2017_2025 without matching it. One row per game, written winner first as the source writes it, so agent_a wins in every row and winner is the constant "agent_a". score_a/score_b are the two bots' in-game scores, which decide a game only when it reaches the 86,400-frame limit. No homefield column: both bots are given a start location on a shared map, and the source records which LAN machine ran each bot rather than which start position, so there is no side to hold an advantage. Also carries round, map, end_frame (game length in frames, roughly 24 a second), game_end_type, crashed and timeout naming the bot that crashed or lost on frame time, and race_a/race_b -- the race each bot REGISTERED, so a bot registered Random played an unrecorded race per game. |
|---|---|
| Reference | Churchill, D. (2026). CoG StarCraft AI Competition 2026 results [Data set]. https://davechurchill.ca/starcraft/cog/results/2026/ |
| DOI | 10.1109/TG.2018.2883499 |
| Licence | Custom |
| Source data | https://davechurchill.ca/starcraft/cog/results/2026/ |
| Comparisons | 14,425 |
|---|---|
| Agents | 10 |
agent_aagent_bcompetitioncrasheddateend_framegame_end_typemaprace_arace_broundscore_ascore_btimeoutwinner
Custom licence — check the terms before reuse: Published on the four rules above, not under a licence of IRW's choosing: no CC licence was named and none of the four is expressible in one, so both licence cells read Custom as with ogs_go_2005_2021. Attribute the competition, do not sell it, cite the original source if you re-host it, and do not train an LLM on it.
Or load it directly in R (no account needed with irw 1.4.0 or later) or Python (needs a free Redivis account; the first call opens a browser to sign in):
# R
install.packages("irw")
library(irw)
df <- irw_fetch("cog_starcraft_2026", source = "comp")
# Python
pip install irw
import irw
df = irw.fetch("cog_starcraft_2026", source="comp")
Or in Stata, which reads the same no-account CSV as the button above:
* Stata
net install irw, from("https://raw.githubusercontent.com/itemresponsewarehouse/Stata-pkg/main/")
irw_fetch cog_starcraft_2026, clear source(comp)
@article{Certicky_2019, title={StarCraft AI Competitions, Bots, and Tournament Manager Software}, volume={11}, ISSN={2475-1510}, url={http://dx.doi.org/10.1109/TG.2018.2883499}, DOI={10.1109/tg.2018.2883499}, number={3}, journal={IEEE Transactions on Games}, publisher={Institute of Electrical and Electronics Engineers (IEEE)}, author={Certicky, Michal and Churchill, David and Kim, Kyung-Joong and Certicky, Martin and Kelly, Richard}, year={2019}, month=Sept, pages={227–237} }
| IRW version | v712 |
|---|---|
| Redivis dataset | irw_competitions v7.1 |
| Redivis dataset DOI | 10.57761/3r1w-t660 |
| Manifest pin for this IRW version | v7.1 |
| Metadata source | irw_meta v68.0 |