Feel free to answer as many questions as you like, but it would be great if everyone answered everything! Please feel free to provide external references/links as necessary Bot Name: NUS-Bot Bot Race: Protoss Author Name(s): Gu Zhan, Francisco Liwa, Liu Jie, He Tao, Jinson Xu Guang Zu Affiliation(s): National University of Singapore Nationality(s): Gu Zhan (PR.China/Singapore Permanent Resident), Francisco Liwa (Philippines), Liu Jie (PR.China/Singapore Permanent Resident), He Tao (PR.China/Singapore Permanent Resident), Jinson Xu Guang Zu (Singapore) Occupation(s): IT professionals: Project Manager, System Analyst, Programmer, etc https://code.google.com/p/nus-bot/ Q: How did you become interested in Starcraft AI? Most of us were StarCraft players during our school days (We are now at our 30s). And we happened to notice BWAPI lib from internet. Then we were amazed to see AI competitions. Since Jan 2013, we are doing our 2.5 years part-time postgraduate program 'Master of Technology' at Institute of Systems Science, National University of Singapore. This NUS-Bot served as our Final Year Project. Q: How long have you been working on your bot? 7 months (2014 Feb-Aug) Q: About how many lines of code is your bot? Around 1500 lines of code enhancement upon UAlbertaBot version 2013 Q: Why did you choose the race of your bot? We enhanced it from UAlbertaBot version 2013, which uses Protoss. Q: Did you use any existing code as the basis for your bot? If so, why, and what did you change? We enhanced it from UAlbertaBot version 2013, which uses Protoss. Q: What is the overall strategy/strategies of your bot? Why did you choose them? The NUS-Bot practices 'Intelligence Preparation of the Battlefield' framework adapted from a real world USA military operation manual, emphasizing on spatial and temporal reasoning. We also enhanced UAlbertaBot's MapGrid.cpp to provide more (map related) information to better support spatial reasoning. Our motivation is to practice 'Knowledge Engineering' upon the real world USA military operation manual, and see any significance in virtual RTS war game, a test bed. FM34-130 IPB manual: http://fas.org/irp/doddir/army/fm34-130.pdf Q: Do you incorporate learning of any form in your bot? If so, how was it accomplished? No learning capability now. (or same as UOABot) Q: Do you use any interesting AI techniques or algorithms in your bot? If so, which? Simplified Layered Influence Map Q: What do you feel are the strongest and weakest parts of your bot's overall performance? Advantage: Need to analyze competition replays to further investigate. Weakness: Strategy is not adaptive. The potential gain form adaptive spacial reasoning might not sufficiently impact the result/performance of out bot, due to congenital/deterministic settings. Q: If you competed in previous tournaments, what did you change for this year's entry? N.A. Q: Have you tested your bot against humans? If so, how did it go? We ourselves played with the bot, we found it relatively strong (credits to UOABot =). We found enjoyable to play with it for several rounds of games, but not for more than 20 rounds. (The strategy is less dynamic) Q: Any fun or interesting stories about the development / testing of your bot? Debugging is a headache =) Q: Any other projects you're working on that you'd like to advertise? We haven't developed all planned functions under 'Intelligence Preparation of the Battlefield' framework, especially the expected/resulted intelligent behaviors of using Transport Units. Our (possible) future release of NUS-Bot will focus more on spatial intelligence and to operate on island maps based on its own autonomous reasoning. Optional Opinion Questions: Q: What is your opinion on the current state of StarCraft AI? How long do you think before computers can beat humans in a best-of-7 match? 1)The mirco management of Bot is great, other than that, I don’t see other major competencies over Human. The current Bot is not good even at tactical reasoning, i.e.: target picking in different situations, and Bots are not able to strategically utilize terrain/map. (Some Terran blocked the choke, but I think it’s pre-configurated, instead of knowing why it need to do so. I think Risk Mgmt here is needed). Our NUS-Bot attempts to use human experience (the IPB manual) to straighten this spatial reasoning part as well as ‘Threat Evaluation (Risk Mgmt)’ part. 2)If no replays can be watched/studied by Human player during the ‘best-of-7 Bot-vs-Human match’, plus the game time is restricted at 15 minutes (basically only allow one round of major military contact. After 15 minutes, games stops and the win/loss is calculated based on scores), I guess it can take 5 years for such a Bot to beat a professional Human player. If Human can watch/study replay after each game, then the Bot need encode more (fixed) strategies, better opponent modeling/prediction and adopt more probability in strategy section, this can take extra 3 years. If there is no limit on game time, I am not optimistic that Bot could win a human player within 10 years. Q: What do you feel is the biggest hurdle (technological or otherwise) in improving your bot's AI? What are the features/states/factors to filter for reasoning in different circumstances. We have read a lot of research papers in SratCraft area, and couldn't find/generate a comprehensively integrated approach/framework to address this. Q: Which bots are the most interesting to you and why? AIUR. Although it encodes very limited, and almost fixed strategies, due to its randomness, it worked very well against other stereotyped bots (especially in the long run, as NUS-Bot vs AIUR trend showed it clearly =). And when competing with Humans, this is also extremely important, as professional human players spend a lot of training time to practice ‘several optimized strategies’ (numerous Case based reasoning at various scales/levels). Since StarCraft is like ‘scissor paper stone’ game, without dominate strategy, the opponent modeling and prediction (with ‘timely’ belief updating), are put under spot. Plus, risk taking/management (to human is psychology and mental element) can also influence the balance of 2 players in/between game(s).