I appreciate your replies, but a two sentence report about your strategy isn't nearly as helpful as a replay of the match would be. I also want to study WHY that VR rush was successful in that game. I want to actually SEE your awesome micro winning the first battle without casualties.
I hear you, tigerija. I was planning on making the AI plan his tactics based on what he has seen through scouting. I don't want the AI to have any unfair advantage over a human player. Except maybe an unlimited APM, because who knows what a human can achieve in that area with enough practice.
And yes, if not enough information has been gathered, a random strategy is picked so the AI has something to do without being completely predictable. Scouting to gather required information is high on the list of course, but the AI can do something else in that time as well. Probably harassing an expansion or the economy of the opponent.
My research into the SC2 AI includes the strategies players are using to defeat computer controlled opponents.
So I would like to invite all of you to try and beat the Very Hard AI which is currently available in the beta on an official Blizzard map 1vs1.
I'm not asking you to try and beat an Insane AI because the only difference between Very Hard and Insane is that the latter is cheating (40% more resources gained). Props to you if you can beat an Insane, but I'm planning on making an AI that plays fair.
So please, send me your replays of your 1vs1 matches against the Very Hard or Insane AI, so I can analyze the strategies used.
Please mail to maarten AT 3tg.nl
That list was uploaded when there was no decent AI available in the beta. But now all AI difficulties are in so I am wondering if Blizzard is finished with the AI toolkit (at least until game release).
Does anyone here have the tools/knowledge/ability to check if the current list is still valid? Because I read somewhere they were discovered when picking apart the executable. Unfortunately that kind of hacking is beyond my current capabilities.
I will do anything in my power to make my AI unpredictable and as good as possible, but it's not possible to create an 'impossible to beat' AI that does not cheat. Don't underestimate the creativity of human players...
An AI that is impossible to beat for YOU would be a different story... ;)
But yes, economy harassment with air units is one of the things on my list. :)
I have decided to start my "adventure" with a formal introduction, so here goes:
As noted in the topic title, my name is Maarten Tops.
I am 27 years old and I live in Veldhoven, The Netherlands.
I am currently following the International Game Architecture & Design course at the NHTV in Breda.
This is an educational program that will teach you to be a professional game artist or programmer.
As I am entering my last year, all that is left for me is the specialization and graduation project.
I have chosen to specialize and graduate in the field of Artificial Intelligence.
As a long-term fan of RTS games, it has been my dream for at least 10 years to develop an AI for a RTS game.
And now the perfect opportunity has arrived. Starcraft 2 (easily the best RTS game ever) comes with community tools that support the building of a custom AI.
My teachers have accepted my plan to research and build an AI for Starcraft 2 as my specialization and graduation.
Of all the Starcraft 2 fan sites, SC2Mapster seems to provide the best location for me to work on this project. A wiki for the AI functions, a good sense for any noteworthy SC2 project and a nice community. So for the next 35/40 weeks, I will be around this site a lot, and I hope you will enjoy having me.
In the first 15 weeks I will be doing research. For example, one of the questions I need answered is what RTS players consider to be a good and fun AI? Another research object is the capabilities of the SC2 AI. The wiki on this site is vital for this part. Currently most of the AI functions listed there are not documented. To build the best AI I can manage, I need to know the tools, so I will work on documenting those functions to I have a better idea what is and is not possible.
When that is done, I will start building the AI itself. In 20 (full-time) weeks it will be my mission to realize my dream. An AI that is actually fun to play against.
This will not be a small undertaking, and I sincerely hope that I can find support and guidance in this community. Because any project manager needs an honest "this is crap" every now and then to prevent said crap from being released to the public... ;)
Thanks for reading this! :)
PS If you want to know more about me, my past projects or what I have learned at school, feel free to visit my portfolio: http://www.blackward.com/portfolio
0
I appreciate your replies, but a two sentence report about your strategy isn't nearly as helpful as a replay of the match would be. I also want to study WHY that VR rush was successful in that game. I want to actually SEE your awesome micro winning the first battle without casualties.
0
Then please help me to create a fun AI and send me a replay. ;)
0
@tigerija: Go
I hear you, tigerija. I was planning on making the AI plan his tactics based on what he has seen through scouting. I don't want the AI to have any unfair advantage over a human player. Except maybe an unlimited APM, because who knows what a human can achieve in that area with enough practice.
And yes, if not enough information has been gathered, a random strategy is picked so the AI has something to do without being completely predictable. Scouting to gather required information is high on the list of course, but the AI can do something else in that time as well. Probably harassing an expansion or the economy of the opponent.
0
My research into the SC2 AI includes the strategies players are using to defeat computer controlled opponents.
So I would like to invite all of you to try and beat the Very Hard AI which is currently available in the beta on an official Blizzard map 1vs1. I'm not asking you to try and beat an Insane AI because the only difference between Very Hard and Insane is that the latter is cheating (40% more resources gained). Props to you if you can beat an Insane, but I'm planning on making an AI that plays fair.
So please, send me your replays of your 1vs1 matches against the Very Hard or Insane AI, so I can analyze the strategies used. Please mail to maarten AT 3tg.nl
0
Since I will be researching the AI toolkit in the next couple of months, I was wondering if the function list at http://www.sc2mapster.com/api-docs/functions/ is up to date.
That list was uploaded when there was no decent AI available in the beta. But now all AI difficulties are in so I am wondering if Blizzard is finished with the AI toolkit (at least until game release).
Does anyone here have the tools/knowledge/ability to check if the current list is still valid? Because I read somewhere they were discovered when picking apart the executable. Unfortunately that kind of hacking is beyond my current capabilities.
0
Sorry, but he doesn't work at the IGAD course, so I'm afraid not... ;)
@ezbeats: Go
I will do anything in my power to make my AI unpredictable and as good as possible, but it's not possible to create an 'impossible to beat' AI that does not cheat. Don't underestimate the creativity of human players...
An AI that is impossible to beat for YOU would be a different story... ;)
But yes, economy harassment with air units is one of the things on my list. :)
0
For melee games (at first at least). Adding support for coop play is planned.
0
...and I will be here for a while.
I have decided to start my "adventure" with a formal introduction, so here goes:
As noted in the topic title, my name is Maarten Tops.
I am 27 years old and I live in Veldhoven, The Netherlands.
I am currently following the International Game Architecture & Design course at the NHTV in Breda.
This is an educational program that will teach you to be a professional game artist or programmer.
As I am entering my last year, all that is left for me is the specialization and graduation project.
I have chosen to specialize and graduate in the field of Artificial Intelligence.
As a long-term fan of RTS games, it has been my dream for at least 10 years to develop an AI for a RTS game.
And now the perfect opportunity has arrived. Starcraft 2 (easily the best RTS game ever) comes with community tools that support the building of a custom AI.
My teachers have accepted my plan to research and build an AI for Starcraft 2 as my specialization and graduation.
Of all the Starcraft 2 fan sites, SC2Mapster seems to provide the best location for me to work on this project. A wiki for the AI functions, a good sense for any noteworthy SC2 project and a nice community. So for the next 35/40 weeks, I will be around this site a lot, and I hope you will enjoy having me.
In the first 15 weeks I will be doing research. For example, one of the questions I need answered is what RTS players consider to be a good and fun AI? Another research object is the capabilities of the SC2 AI. The wiki on this site is vital for this part. Currently most of the AI functions listed there are not documented. To build the best AI I can manage, I need to know the tools, so I will work on documenting those functions to I have a better idea what is and is not possible.
When that is done, I will start building the AI itself. In 20 (full-time) weeks it will be my mission to realize my dream. An AI that is actually fun to play against.
This will not be a small undertaking, and I sincerely hope that I can find support and guidance in this community. Because any project manager needs an honest "this is crap" every now and then to prevent said crap from being released to the public... ;)
Thanks for reading this! :)
PS If you want to know more about me, my past projects or what I have learned at school, feel free to visit my portfolio: http://www.blackward.com/portfolio