I''m still concerned about one thing, and that is the idea of intentionally submitting a weaker robot at first, and then submitting a full power robot for the real compition. Everytime I think about how to make one (if I decide to compete) I think of the idea of hiding and special strategies and abilities by not including them in the first submission bot (I do have a list of things I don''t think [many] others would have thought of from my experience with battle bot programs). Such disabled abilities could easily be obscured in the source code by creating conditions for them to execute that never happen. Hence the bot has some killer stradgy, but the original version never uses it because some set of fuzzy conditions have been rigged to prevent it. Due to this method I doubt any but the most absolutely complete code inspections would find it, and then it''s a matter of whether it was intentional or simply an oversight/bug.
What I would suggest is a qualifier of sorts to be played by all bots submitted at the first date. It could be a single elimination tournament, with the winner getting some sort of bonus point in the actual competion (not something like enhanced health that would let them beat better opponents, but something outside the actual playing field, like an extra imaginary win). Something like this would put a big incentive to submit the best possible bot the first time around.
Also, you haven''t responded to a few of my questions in this thread, one I can think of is if the program will be able to be run from the command line, with DLL names passed as parameters and the result of the match written to a file in a parsable format (it might also be useable for such a mode to be able to disable 3D rendering). This would help in the development of neural net bots, since manually running the game hundreds of times would be difficult.
🎉 Celebrating 25 Years of GameDev.net! 🎉
Not many can claim 25 years on the Internet! Join us in celebrating this milestone. Learn more about our history, and thank you for being a part of our community!
GameDev: Arena
Hi Khawk,
Please will you consider a command line interface for the arena? I would like to try out some genetic AI bots, but without some automated method of running a competition again and again it would be almost impossible to get any sort of natural selection going on :-)
A simple:
C:\4E4.exe /nowindow /displaywinner
Kevin
C:\
I think might be enough. Unless of course you already have some kind of automated method to run lots of matches, or a better way (yay) - in which case I will retreat to the back of the class and turn my chair to face the wall :-)
Thanks,
Paul
Please will you consider a command line interface for the arena? I would like to try out some genetic AI bots, but without some automated method of running a competition again and again it would be almost impossible to get any sort of natural selection going on :-)
A simple:
C:\4E4.exe /nowindow /displaywinner
Kevin
C:\
I think might be enough. Unless of course you already have some kind of automated method to run lots of matches, or a better way (yay) - in which case I will retreat to the back of the class and turn my chair to face the wall :-)
Thanks,
Paul
quote: Original post by Khawkquote: Original post by liquidAirquote: Original post by Khawkquote: Original post by liquidAir
I downloaded the demo but it doesn''t work on my crappy system...
PIII 500MHz, 192MB RAM, Win98 SE, er... integrated video 4MB
Doesn''t work as in crashes? runs choppy? doesn''t do anything?
Doesn''t crash. Once it loads, it''s kinda like slow. It just shows a white screen with the status and menu bars, nothing else. Doesn''t handle mouse events immediately. When I click on the menu, takes like 5 to 10 seconds to come up.
I guess we found the minimum specs.. it requires a decent video card. If this becomes an issue, maybe I can create a simple 2D version.
It does the exact same thing on my machine. Plain white window with menu bar and status bar, menus take 2-3 seconds to respond. Selecting "go" does nothing, and its using up 99% processor time when I go to kill it in task manager.
And I dont think my machine qualifies as minimum spec just yet :-)
P4 1.7ghz
512MB RAM
Geforce 4 (128MB VRAM)
Win XP
DirectX 9
Alan
khawk:
So you plan to have a series of contests? Like an annual thing?
Can we perhaps enter 2 bots, one in each category?
If you are distributing the arena .exe does that mean two of us can challenge each others bots for practice - then people can give away as many secrets as they wish in return for info on other bots?
I like the idea of a specific forum for this BTW
[edited by - d000hg on July 14, 2003 12:05:50 PM]
So you plan to have a series of contests? Like an annual thing?
Can we perhaps enter 2 bots, one in each category?
If you are distributing the arena .exe does that mean two of us can challenge each others bots for practice - then people can give away as many secrets as they wish in return for info on other bots?
I like the idea of a specific forum for this BTW
[edited by - d000hg on July 14, 2003 12:05:50 PM]
Wow... Can't believe it work on my sitty video card...
I have a p3 733mhz, 384 sdram, S3 Savage 2000
The only problem is that it work correctly ONLY when I maximize the window, otherwise, everything is flashing...
But there is a bug, when the windows is maximized, I can no more see the time and the HP :I guess it would be easy to fix Kevin ?
edit: Everything was correct even in the normal window size when I reopened it !
Nice job!
btw, Bots are stupids !!
Last bug I just noticed, the bots can be bugged in a tree, he can try to move in any direction but he is locked :\
[edited by - Hedos on July 14, 2003 12:14:15 PM]
I have a p3 733mhz, 384 sdram, S3 Savage 2000
The only problem is that it work correctly ONLY when I maximize the window, otherwise, everything is flashing...
But there is a bug, when the windows is maximized, I can no more see the time and the HP :I guess it would be easy to fix Kevin ?
edit: Everything was correct even in the normal window size when I reopened it !
Nice job!
btw, Bots are stupids !!
Last bug I just noticed, the bots can be bugged in a tree, he can try to move in any direction but he is locked :\
[edited by - Hedos on July 14, 2003 12:14:15 PM]
Ouch, the demo doesn''t work too well on my comp. It does work, well enough to see what the bots do, but I get some major flicker. Strange thing is, the flicker isn''t the standard v-sync flicker, because it only affects the terrain (not the trees or rocks or anything, just the floor), and it is all over the place.
My system: 1 Ghz, 128 Megs, TNT2
I was going to try and get a screenshot of the wierd flicker (not sure if it is possible though), but the program crashed when one bot came into the other''s FOV.
Another comment: The FOV is triangle shaped. Shouldn''t it be more like a pie slice? Is this just how it is drawn, or is it really what the bot can see?
My system: 1 Ghz, 128 Megs, TNT2
I was going to try and get a screenshot of the wierd flicker (not sure if it is possible though), but the program crashed when one bot came into the other''s FOV.
Another comment: The FOV is triangle shaped. Shouldn''t it be more like a pie slice? Is this just how it is drawn, or is it really what the bot can see?
quote: Original post by Michalson
I''m still concerned about one thing, and that is the idea of intentionally submitting a weaker robot at first, and then submitting a full power robot for the real compition. Everytime I think about how to make one (if I decide to compete) I think of the idea of hiding and special strategies and abilities by not including them in the first submission bot (I do have a list of things I don''t think [many] others would have thought of from my experience with battle bot programs). Such disabled abilities could easily be obscured in the source code by creating conditions for them to execute that never happen. Hence the bot has some killer stradgy, but the original version never uses it because some set of fuzzy conditions have been rigged to prevent it. Due to this method I doubt any but the most absolutely complete code inspections would find it, and then it''s a matter of whether it was intentional or simply an oversight/bug.
What I would suggest is a qualifier of sorts to be played by all bots submitted at the first date. It could be a single elimination tournament, with the winner getting some sort of bonus point in the actual competion (not something like enhanced health that would let them beat better opponents, but something outside the actual playing field, like an extra imaginary win). Something like this would put a big incentive to submit the best possible bot the first time around.
Also, you haven''t responded to a few of my questions in this thread, one I can think of is if the program will be able to be run from the command line, with DLL names passed as parameters and the result of the match written to a file in a parsable format (it might also be useable for such a mode to be able to disable 3D rendering). This would help in the development of neural net bots, since manually running the game hundreds of times would be difficult.
At one point I asked what would be expected in a results file, because the most I''ll give is the results of the battle in terms of winner/loser, remaining HP, and remaining time. Anything else would be giving more information than is available to the bot during execution, which is why if the bot wants to output data from its battles, it will need to do so on its own.
From feedback here, some changes will need to be made to enable a non-graphics and/or 2D mode.
quote: Original post by paulsdsrubbish
Hi Khawk,
Please will you consider a command line interface for the arena? I would like to try out some genetic AI bots, but without some automated method of running a competition again and again it would be almost impossible to get any sort of natural selection going on :-)
A simple:
C:\4E4.exe /nowindow /displaywinner
Kevin
C:\
I think might be enough. Unless of course you already have some kind of automated method to run lots of matches, or a better way (yay) - in which case I will retreat to the back of the class and turn my chair to face the wall :-)
Thanks,
Paul
I''ll consider it, but I can''t guarantee it yet.
quote: Original post by Khawkquote: Original post by Michalson
I''m still concerned about one thing, and that is the idea of intentionally submitting a weaker robot at first, and then submitting a full power robot for the real compition. Everytime I think about how to make one (if I decide to compete) I think of the idea of hiding and special strategies and abilities by not including them in the first submission bot (I do have a list of things I don''t think [many] others would have thought of from my experience with battle bot programs). Such disabled abilities could easily be obscured in the source code by creating conditions for them to execute that never happen. Hence the bot has some killer stradgy, but the original version never uses it because some set of fuzzy conditions have been rigged to prevent it. Due to this method I doubt any but the most absolutely complete code inspections would find it, and then it''s a matter of whether it was intentional or simply an oversight/bug.
What I would suggest is a qualifier of sorts to be played by all bots submitted at the first date. It could be a single elimination tournament, with the winner getting some sort of bonus point in the actual competion (not something like enhanced health that would let them beat better opponents, but something outside the actual playing field, like an extra imaginary win). Something like this would put a big incentive to submit the best possible bot the first time around.
Also, you haven''t responded to a few of my questions in this thread, one I can think of is if the program will be able to be run from the command line, with DLL names passed as parameters and the result of the match written to a file in a parsable format (it might also be useable for such a mode to be able to disable 3D rendering). This would help in the development of neural net bots, since manually running the game hundreds of times would be difficult.
At one point I asked what would be expected in a results file, because the most I''ll give is the results of the battle in terms of winner/loser, remaining HP, and remaining time. Anything else would be giving more information than is available to the bot during execution, which is why if the bot wants to output data from its battles, it will need to do so on its own.
From feedback here, some changes will need to be made to enable a non-graphics and/or 2D mode.
Match outcome (which could be expressed simply as the final health of both robots, and perhaps the time it took for the match to complete) is all that is required.
quote: Original post by AlanKempquote: Original post by Khawkquote: Original post by liquidAirquote: Original post by Khawkquote: Original post by liquidAir
I downloaded the demo but it doesn''t work on my crappy system...
PIII 500MHz, 192MB RAM, Win98 SE, er... integrated video 4MB
Doesn''t work as in crashes? runs choppy? doesn''t do anything?
Doesn''t crash. Once it loads, it''s kinda like slow. It just shows a white screen with the status and menu bars, nothing else. Doesn''t handle mouse events immediately. When I click on the menu, takes like 5 to 10 seconds to come up.
I guess we found the minimum specs.. it requires a decent video card. If this becomes an issue, maybe I can create a simple 2D version.
It does the exact same thing on my machine. Plain white window with menu bar and status bar, menus take 2-3 seconds to respond. Selecting "go" does nothing, and its using up 99% processor time when I go to kill it in task manager.
And I dont think my machine qualifies as minimum spec just yet :-)
P4 1.7ghz
512MB RAM
Geforce 4 (128MB VRAM)
Win XP
DirectX 9
Alan
I think I know what''s causing this issue, but I don''t know why it''s doing it on your machine. I''ll try some other algorithms..
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement