MeleeSurfingChallenge

Think you're pretty good at dodging bullets? How many can you dodge at once?

Directions

Here's how you run a MeleeSurfingChallenge:

• Get the reference bots here: http://davidalves.net/robocode/wiki.challenge.MeleeSurfingChallengeTeams_1.0.jar
• Team A fires with HeadOnTargeting, Team B fires with LinearTargeting. Source code for both bots is included.
• Run 1000 rounds against each of them
• Arena size 1000 x 1000
• Everything else set to Robocode defaults
• You are not allowed to fire any bullets

Your bot's score is average amount of life you had at the end of each round. 100 means that the reference teams never hit you. You calculate your score like this:

 100 - (reference team's bullet damage score / 1000)

Results

Bot Name	Author	Type	Score vs. Team A	Score vs. Team B	Overall Score	Comment
Slartibartfass 0.4 + dev	Krabb	MR/MR	48.989	0.656	24.8225	score vs team B is broken.
Phoenix 0.41	David Alves	MR/WS	0.488	3.607	2.0475
Griezel 0.3	GrubbmGait	MR/MR	0.451	3.543	1.9970
Sparrow 2.5TC	Dummy	RM/RM	0.408	0.991	0.6995

Sparrow's Score is included for reference, it has a decent random melee movement.

For the "Type" column, show your melee and 1-v-1 movement like this: RM/WS for RandomMovement in melee, WaveSurfing in 1-v-1.

Key:
RM - RandomMovement
WS - WaveSurfing
SG - Stop 'n' Go
MR - MinimumRiskMovement

Chat

---

Let's see some high scores! 20 or more should be possible. =) --David Alves

I was on the way to build exactly this challenge bots, to test my team bots. Thanks for doing it :) But i think the reference bots should stop shooting, if they would hit a team mate. Or they must avoid friendly bullets. Otherwise a position in one line with two reference bots would give you an advantage. I´m not a melee specialist, but i think it´s not an advantage to do this in a normal melee match?! --Krabb

I'm not a melee expert, but it seems like it might make sense to run this challenge on 1000x1000, since that's what the melee rumble runs in. Some MiniBots (and smaller) might have the size of the field hardcoded, I would think. Anyway, cool challenge, and nice scores Krabb. - Voidious

Ah sorry, field size was supposed to be 1000x1000, fixed now. Phoenix and Sparrow were run on 1000x1000.

I think there is to much free space in this challenge, its not realistic. May be we should raise the reference bots to 9 and let only 3 of them fire at the chellanger. But the actual chellange is still good to mesure the movement against HOT and LT, which is i think the goal of this challenge.--Krabb

I think without needing to avoid being targeted the test won't accurately reflect the robot's melee dodging ability. Although it would do terribly in a real battle, FemtoWalls? sets a new high MeleeSurfingChallenge score (62.239 vs. team A). Krabb's idea might work to stop wall riders like this because the challenge bots will be getting disabled more evenly, making it much harder for robots that move a lot to win. Another (pretty drastic) solution would be making the challenge something like 6 challenge bots, the challenger, and 3 other movement-only bots with the sample bots shooting at the closest target.

public class FemtoWalls extends Robot{ //this could implement droid too
	public void run(){
		turnLeft(getHeading() % 90);
		do{
			ahead(Double.POSITIVE_INFINITY);
			turnRight(90);
		}while(true);
	}
}

1st: kc.tests.FemtoWalls?	142861	119600	22950	0	0	309	1	765	235	0
2nd: Team: wiki.challenge.MeleeSurfingChallengeTeamA?	78774	30400	6080	37614	4679	0	0	235	1228	0

-- Kev

This is not a challenge to mesure your meleeruble ranking! Its a good one to test your bullet avoidance against HOT and LT, and this is sure not the only factor in a melee battle. We must create more challenges to get an accurate reflection of the rumble and i think this is a good start. --Krabb

I see what you mean. Some challenges like the SpinBotChallenge are more of challenges in themselves rather than say the TargetingChallenge2K6 which is kind of an extension of the rumble. I think it's good to see more melee-based challenges, as it is a very underlooked aspect of Robocode. -- Kev

This is not really meant as a good measure of bot performance, just a fun (and hard!) challenge, like SpinBotChallenge. =) --David Alves

Griezel survives 114 times against teamB and 0 (=zero!) times against teamA. I think Dummy's BlindDroid? could score very well in this challenge. -- GrubbmGait

@Krabb: Why do you say "score vs. Bot B broken?" on your score comment? Was there some problem running the challenge? @Kev, you're right, we need more melee challenges... Let's revive the SampleBotMeleeChallenge. =) --David Alves

I first run the chellange on 800x600 and then tried it on 1000x1000. But while the 1000x1000 battle vs. Team A was running i worked on my gun and accidentally reduced the movement lateral velocity segmentation to one segment. May be "broken" is the wrong vocable :) --Krabb

Lol, I was just working on reference teams for exactly this! I didn't realize it was the same till just now. Do your bots avoid shooting each other?