haha, now summer is here and i don't have any exams i can actually finish the beast. As we speak i am tuning the movement and i am fairly pleased with how it is looking (i'm only writing this as i'm waiting for another 1000 battle to finish). Give me a couple of weeks to finish tuning the movement and stick my old set of guns on and i'll unleash him onto the world (please note i am not hyping him up kawigi style, he won't be very good).|
Well, today i stripped out all the complicated and un-necessary code from Roosevelt and found it works much better. This means no melee strategy, no dynamic distancing and far fewer guns. It isn't properly tested yet, but i can't really be bothered so i'll just let him run wild and see how he does.|
Oh, and while i think about it, current white whale is StatistRobot, it beats me 60%-70% of the time seemingly regardless of what i do. Could it be because of my current lacklustre guns?? -- Brainfade|
Great, I want to try it. Where can I download it?
You can't quite yet, I botched in some ideas and code from a few test bots and i still have a couple of bugs i need to get rid of. Also it doesnt save data in between rounds, so i'll add that tomorrow morning then upload him.
My current bot, Tax, beats StatistRobot quite regularly, and the only guns he is currently using are HeadOnTargeting and CircularTargeting. I think perhaps losing to StatistRobot points to a weakness in movement. Although, of course, a better gun can't hurt. -- nano|
How competitive is it?
That's an interesting question. Currently in its v0.1 guise it's not very. It can push a few of the top bots to low energies quite consistently before being killed, hopwfully i can work some of this out either by tuning or actually putting some thought into my endgame. I'm hoping he'll place about 50th.
How does it move?
Well, it is basically random movement that i added some APM to. This really helps it against pattern matchers but means it get crucified by anything else. I'll either tweak or remove this before i release.
How does it fire?
It has 2 guns, a guess factor gun and a pattern matcher, and i'm hoping that one will pick up the slack from the other. The guess factor gun is segmented into 6 dimensions, and the VGun array is loosely segmented into 4.
How does it dodge bullets?
Well at the moment i pattern match myself, then when i'm getting my new position, if this is close to the predict position i'll pick another spot. I found taht actual bullet dodging was incredibly difficult to do well.
How does the melee strategy differ from one-on-one strategy?
It doesn't - I ripped out all the melee strategy, the new version has never been in a match with more than 1 opponent.
How does it select a target to attack/avoid in melee?
It doesn't, although i might make it so it just tries to take out DT, i'll do badly overall but it'd be worth it... :)
What does it save between rounds and matches?
Between rounds it saves movement, and targetting stats (PM Data, guess factors, VirtualGun? stats), i haven't sorted out between matches yet but it'll be along the same lines.
Where did you get the name?
There's an american band called "Slow Roosevelt", and when i was trying to think of a name i was listening to some of their stuff, just a coincidence really.
Can I use your code?
Not at the moment and i doubt you'd want to either, it's pretty ugly and unorganised. I'll release it when i've managed to get the melee strategy sorted.
What's next for your robot?
Melee strategy, maybe another stat gun, tune some parameters, make it skip less turns, clean up the code. And depending on how it does i might rework the movement.
Does it have any WhiteWhales?
I haven't really tested enough to work this out yet, but no doubt fermat'll give me some grief.
What other robot(s) is it based on?
It's not directly decended from anyone, although somewhere on these pages is Aspid's radar code, that i nicked - cheers albert! (that's the only code in the bot that i creamed straight off someone else). Other than that i've taken ideas from all over the place.
Comments, questions, feedback: