Feel free to answer as many questions as you like, but it would be great if everyone answered everything! These survey answers will be included in the 2015 competition report, so the more info the better. Please feel free to provide external references/links as necessary Bot Name: Overkill Bot Race: zerg Author Name(s): sijia xu Affiliation(s): Independent Nationality(s): china Occupation(s): data engineer (These will be listed on the competition website) Bot URL: https://github.com/sijiaxu/Overkill Personal URL: https://github.com/sijiaxu Affiliation URL: Q: How did you become interested in Starcraft AI? I love video games, especially Blizzard's games, and as a data mining programmer, I also interest in AI-related things. so starcraft AI seems to be the exactly suitable thing for my interests. by the way, working with so many people which has the same interest and improving skills together is also a great thing. Q: How long have you been working on your bot? about half year. Q: About how many lines of code is your bot? about 12000 lines. Q: Why did you choose the race of your bot? I love queen! Q: Did you use any existing code as the basis for your bot? If so, why, and what did you change? yes, I use ualbertabot as the basis, it document well and has good coding style :) now I have changed most part of it, mostly because I use the previous ualbertabot version and it don't contain zerg-related work. besides, a little suggestion is some part of ualbertabot may be over-design, including too many layers. sometimes simpler is better. Q: What is the overall strategy/strategies of your bot? Why did you choose them? main strategy is mutalisk harass, zergling/hydrisk pushing are just working as supplement to mutalisk strategy. I think the high mobility of mutalisk is a big advantage comparing with other races. Q: Do you incorporate learning of any form in your bot? If so, how was it accomplished? yes. currently just using UCB1 to selecting the opening strategy based on previous match results. Besides, I also use record simulation to accelerate UCB1’s learning speed(for example, if detecting early rush and 10-hatch opening fail, simulating a record that 12-hatch also fail). Q: Do you use any interesting AI techniques or algorithms in your bot? If so, which? just a A-star path finding and influence map, not so interesting.. Q: What do you feel are the strongest and weakest parts of your bot's overall performance? weak part is so many... the weakest part I think is that bot is not adaptively at the strategy level. it including two things: 1. bot not contain sufficient basic strategy. for example, droping, lurker attack.. 2. bot can not adaptively select strategy base on current situation. currently just hard-coding some poor conditions. Q: If you competed in previous tournaments, what did you change for this year's entry? yes, add the ucb1 opening selection to counter the early rush, faster A-star and many little improvement. Q: Have you tested your bot against humans? If so, how did it go? no Q: Any fun or interesting stories about the development / testing of your bot? development itself is exactly a happy thing for me. Q: Any other projects you're working on that you'd like to advertise? no.. 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? I think the gap is still exist. although the bot's micro-mangement is better than human at sometimes, but at the strategy level is still poor. I think maybe need about two more years before computers can beat humans. Q: What do you feel is the biggest hurdle (technological or otherwise) in improving your bot's AI? I think the biggest hurdle for me is utilizing the starcraft's replay, including do the basic battle data recording, making the features and than using the data to train the strategy selecting model. among all the things, the basic data recording process maybe the biggest hurdle. Q: Which bots are the most interesting to you and why? tscmoo. it has good tactic management and different types of unit can cooperate well. I have tried to read the source code, but it is so unreadable...