| Bot | ||
| SandboxDT | deathcon, | |
| Shadow | deathcon,Nat | |
| Phoenix | Nat, |
You really would not like to see the DT source - it has not been cleaned up since SandboxLump - its ugly, ugly ugly. -- Paul
Send it over and I'll help clean it up ;-) -- Kawigi
He all will help you. -deathcon
I'm mostly interested in the movement. And that couldn't be the same since Lump, can it? =) -- PEZ
Actually, if there's one set of movement code I'd like to see and try to understand, it would be SandboxDT 1.61's. #2 would probably be either HawkOnFire or GlowBlowMelee. -- Kawigi
What about other games, you guys like to beat them yourselves, or just rush to find a walkthrough at gamefaqs.com? ;) Seriously, I always tried to resist the temptation to read other people's bots source code, I consider them to be "spoilers" to some degree, the real fun is in trying to find your own personal way of solving the puzzle. Otherwise wouldn't we end up with a "DT clone competition" with everybody trying small tweaks to the same basic idea? -- ABC
For a short time, yes, but with that many people all tweaking the same bot, we would soon reach the maximum possible and people would have to start being original. That originality could then be based on experience learned while moding DT. -- Tango
Yes, I definately agree. Look at Raiko. It's hardly a Tityus clone anymore. At least not more than T is Raiko clone. =) Jamougha brought some refreshing and needed new thinking to my bot. It might do wonders to DT and Shadow to open source them. In fact it probably would. -- PEZ
Ok, HawkOnFire is open source now. I hope there will be more melee bots in future. -- rozu
*Tango applauds*
:-D - Kawigi
It looks like the one you just uploaded still doesn't have source (?) -- Kawigi
I uploaded it separately, it isn't in the package. There should be a button with "Download Source". -- rozu
Ah, I see. I've hidden some source code in those links before, but usually found that others don't. I hope you're ready to race with Shiz for the top, à la Jamougha and PEZ. -- Kawigi
*jaw drops* Wow, HawkOnFire OpenSource... it's like I won the lottery or something, I've just been bashing my head against the wall trying to understand what could make a movement this good in melee.. awesome.. Oh, and yes, ABC, I do consider source to be something of a spoiler, but sometimes I need to look up the solution to one puzzle on gamefaqs ;) -- Kuuran
I have mixed feelings about that move Rozu :\, I really hope people try to improve it instead of just using it as an easy way into the top of the melee ranking... -- ABC
Ok, ABC just to satisfy a curiosity i've had for ages is there any chance you could plug the HawkOnFire movement into either Shadow or Tron?? pretty please... -- Brainfade
Ah, a killer Melee movement. I wonder what could be done with that... =) -- PEZ
Well, ABC, I think that Rozu and I are more optimistic than you. I show that by having open sourced Coriantumr and Shiz, which, with HawkOnFire now, are the three most effective open-source melee bots. My hope is, as well, that we can work together (in friendly competition) to make an even more outstanding bot and push the state of the art. I faced something of a dilemma when I released Coriantumr - whether to include source in the jar or not (David as my witness, we talked about it for awhile), because I suspected it would be the most effective melee bot that ever released source. In the end, I decided I wanted others to follow my example and delve into melee (you haven't done robocode until you've tried to build a melee strategy that works). My hope is that people working on melee-capable bots will read the MinimumRiskMovement page (which could use some more helpful content), devise their own solution to the problem, and then take a look at implementations like mine and Rozu's and benefit from our observations and experience, while also letting us benefit from their new ideas. -- Kawigi
Ok, here it is, abc.tron2.Hawk sports HawkOnFire's melee movement and Shadow's radar+gun+1on1 movement:
Now you know what I meant by "best melee movement I've ever seen"... I hope you are right Kawigi, the melee rumble could use a little more competition, it's just that this is a big spoiler. I won't release Hawk, but will surely take the challenge of trying to improve Shadow's movement to the point of beating it. ;) -- ABC
I'd rather that DT and shadow stayed closed-source. I'd rather be in awe of them then dissect them and see exactly what makes them so good. Besides, if nobody can see the source, maybe people will come up with many different solutions to the same problem (e.g. melee movement) rather than stick to the ideas that are already proven to work. Just my two cents. :-) --David Alves
I don't mind if some bots are closed sourced. But the problem is than the closed sourced bot has the advantage to update some ideas based on the open sourced bots, with out sharing any code or ideas, I find this unfair. -- SSO?
Here's another thought, if ABC feels like making another experiment - how would the movement in Shiz fare with a Shadow-like gun? I think that it may have some strengths over HawkOnFire in some areas, but may be overall more predictable than HawkOnFire, which is its failure. I say that because against all micros, Shiz is nearly even with HawkOnFire, but with the best bots, it's somewhat lagging. -- Kawigi
I'm with David here, I spent a very long time with the firm belief that I would never make a 1on1 movement as good as DT's. But then one day I "freed my mind" ;), and tried something I always thought impossible. The result was a movemet way better than DT's... Of course Paul imediatly realised the possibility of what I had done, and came up with his own version of the same (or similar) idea, and he probably did it quite different from my implementation (mine is still slightly better ;)). I will eventually release source code or an explaination of Shadow's movement, but first I would like to see more people breaking the 1950 barrier by themselves. Here is the basic secret: you will never perfectly flatten a movement profile by hand, at least not as well as a powerfull processor can on the fly. There is absolutely no Math.random() involved in Shadow's movement! -- ABC
Yes, I guessed that one already. :-) What puzzles me is how you managed to make it work so quickly and accurately, rather than what you're doing. My best efforts at the same thing have worked reasonably well, but run at about 40 fps, minimized, and even then it's a balance between skipped turns and good movement. But working it out is definitely a lot of fun, and I'm convinced my *next* idea will be the one... ;-) -- Jamougha
"I would like to see more people breaking the 1950 barrier by themselves"... 1950? I'm gunning for 2000! (someday) --David Alves
![[Home]](/images/RoboWiki.png)