Angsaichmophobia

An advanced robot that specializes in melees but performs better in duels by Kinsen.

Current melee rating/ranking: 1579.02
Current duel rating/ranking: 1627.47

Version History

• Version 1.8c
  • September 21, 2006
  • 1579.02 (melee)/ 1627.47 (duel)
  • Fixed the pattern matcher and also fixed a bug that caused it to fire at the previous predicted location.
• Version 1.8b
  • September 20, 2006
  • Pulled due to a bug in the pattern matcher.
  • Refined the pattern matcher and made a few minor tweaks.
• Version 1.8
  • September 20, 2006
  • Pulled because it had a bad if-condition that caused it to not fire.
  • Added a simple pattern matcher and fixed some bugs with the guns
• Version 1.7
  • September 17, 2006
  • 1577.29 (melee) / 1609.67 (duel)
  • Fixed a bug in the oscillator gun and tweaked the virtual bullets.
• Version 1.6b
  • September 15, 2006
  • 1556.35 (melee) / 1608.28 (duel)
  • Fixed a major bug that caused it to crash.
• Version 1.6
  • September 15, 2006
  • Pulled from rumble because of a major bug.
  • Added average circular and guess factor gun.
• Version 1.5
  • September 13, 2006
  • 1569.09 (melee) / 1600.6 (duel)
  • Added virtual bullets.
• Version 1.4
  • September 11, 2006
  • 1592.15 (melee) / 1575.24 (duel)
  • Fixed bug that was not actually fixed in version 1.3.
• Version 1.3
  • September 10, 2006
  • 1579.56 (melee) / 1578.32 (duel)
  • Fixed bug in bullet avoidance that affected its melee performance only.
• Version 1.2
  • September 8, 2006
  • 1599.49 (melee) / 1576.25 (duel)
  • Added oscillator gun.
• Version 1.1
  • September 7, 2006
  • 1599.66 (melee) / 1572.04 (duel)
  • Added stop and go gun and factored in robot path for movement.
• Version 1.0
  • September 6, 2006
  • 1583.48 (melee) / 1513.56 (duel)
  • Initial release.

Definition of Robot's Name

The name was made from a combination of angstrom, one ten-billionth of a meter, and aichmophobia, fear of sharp and pointed objects.

Movement

It uses a simple relative grid system to determine the best location to be in that is close to its own position. I assume that it is similar to an anti-gravity robot because it tries to go to the place with the fewest bullets although it does not actually bounce away from bullets. It also only dodges very simple bullets such as head-on and circular bullets so it is vulnerable to more advanced robots.

Targeting

Its gun is also simple and consists of five guns: Head-on, circular, linear, half-linear, and stop and go. It uses a real time hit-percentage to determine the best gun although it has no good gun against certain robots and does not use virtual bullets.

Future Versions

Since this is my only released melee robot, I will probably experiment a lot with melee battles using this robot. However, I (do not) expect astounding results and I plan to add more advanced targeting and movement to it in later versions.

• Advanced targeting (Slightly done) and better bullet dodging
• Multimode movement

Current Bugs

Feel free to add to this list if you wish to help the robot do better.

• It occasionally gets stuck next to a corner because it is heading towards a lower location which is removed once it starts heading towards that direction because it is too close to the walls.
• It sometimes hits the wall due to a large turn while using the grid movement.

Code Usage

The robot is open source under the terms of RWPCL and includes graphical debugging including its own on-screen debugging console.

Credits

This robot does not contain any code directly taken from a single robot but the RoboWiki has been particularly helpful. Open source robots were also a great resource which is why I am releasing the source code.

Downloading the Robot

It can be downloaded from https://geocities.com/seakinsen1/robocode/kinsen.melee.Angsaichmophobia_1.8c.jar

Questions and Comments Below

---

Does it have a fear of trigonometry? -- Martin

No, it is a combination of angstrom and aichmophobia; one ten-billionth of a meter and fear of sharp and pointed objects, so fear of sharp and pointed objects that are very small--bullets. I will probably release it this week but I can not yet because it is for a school competition. In fact, I wrote it in a week so it is not too good especially since I do not specialize in melees. -- Kinsen

You're lucky you grabbed this name, I was just about to make a bot named Angsaichmophobia. --David Alves Just kidding. =)

---

I was very surprised by a 60 point jump in its ranking due to a simple stop and go gun and also the robot path factoring addition. -- Kinsen

What do you mean by "robot path factoring addition"?

I couldn't think of how to describe it in a few words but when choosing its next best location it factors in the path it will take to the next location so that it will not be going to a bullet free spot but have to pass through a lot of bullets. -- Kinsen

---

See, this guy brought this bot into a competition. In AP Computer Science. In the first week of a class. Against bots that would probably lose to MyFirstRobot.

I hate you, Kinsen. You've scarred my soul forever. -- Flare

You're lucky ABC isn't in your class. ;-) --David Alves

• delicate shudder* He's the guy with Tron and Shadow, right? Actually, my current in-development bot can take Kinsen's, but not all of the code is mine so I have no right to gloat. --Flare

I have a newer version that I will upload once I am done testing. Have fun. -- Kinsen

That's the one I saw in class yesterday, right? 'Cuz if it's not, I'm going to cry. v.v --Flare

Yes and no. The one you saw was my pattern matching gun only. And there is a bug somewhere so I have to pull the new version for now. -- Kinsen

...uh...umm...

• half-cries* --Flare
• It would be funny if Flare were just playing coy, and is going to bring his WaveSurfing, GuessFactorTargeting MegaBot to the next competition. :P -- Voidious
• Curses! Back to the drawing board. ...well, the way my bot is going now, it may well be the first MegaBot to feature heads-on targeting. x.X --Flare
• My robot has head-on targeting, but it also has about seven others... -- Kinsen

v1.8b is running very slow. -- Martin

Yeah, prohibitively so... It'd be great if you could speed that up :-\ It definitely gets slower as the match goes on. -- Voidious

It could be because I am not dumping any of the old pattern matching data. Also, I almost never try it in 35 round battles. -- Kinsen

Okay, that's IT. My next bot is going to be named Angsaichmophobiaphobia. >.> --Flare

Please try a 35-round battle. It's about 1/5th the speed of Shadow. =( -- Voidious

I think more to the point is: Please improve its performance drastically before returning it to the rumble. Or pull a card from my deck and enter it in the rumble but don't make it available, instead dropping it directly into your own roborumble/robots directory. I think that's a good way to test frequent or slow versions without causing tension, like Florent v. Corbos a few months ago. -- Martin

Hey Kinsen, I appreciate you pulling your tank the way you did, though I'm sorry it came to that. You would probably save time on the whole - just in your own testing / development - if you took the time to make a faster solution, though. For PatternMatchers?, Corbos' design for a FoldedPatternMatcher could improve speed dramatically (note that Che uses it, and stays pretty fast over 500 rounds in the MovementChallenge2K6). Another point to note is that GuessFactor guns tend to outperform PatternMatchers? while also executing quite a bit faster. Good luck in any case! Cheers, -- Voidious

No. Somehow, these recent versions have been frustrating for me because I have always found a bug that caused it to fail (and I even tested the last one a lot). I tried running the version for a few rounds and it took a long time because it was in an infinite loop or something like that. -- Kinsen

Well, if you release the source, or parts of it, I'd be happy to take a look at it for you. Also, an easy way to find an infinite loop is to sprinkle your code with out.println() statements. The infinite loop happens after the last one that's shown in the log. =) For general performance debugging, you can use System.nanoTime() to measure how long your code is taking to execute. --David Alves