[Home]SpinBotChallenge

Robo Home | Changes | Preferences | AllPages

The SpinBot Challenge

Beat SpinBot better than anyone else. Setup:

  1. The results must come from 1000 round OneOnOne battles - Challenger vs SpinBot
  2. Battlefield size is 800x600
  3. Best score share (percent of total score) wins.
    1. If two bots get the same score share - the Challenger with the highest score wins
  4. Your bot must extend Robot

Results

Challenger Author Score share (%)Total score
AntiSpinBot? with firing David Alves 99.539 162997
AntiSpinBot? without firing David Alves 97.560 60025
SpinBotHunter Goofy 95.737 166077
sample.Walls Mat Nelson 94.492 124737
St00pid Dummy 93.777 163352
Krazy Alcatraz 86.358 153699
WallsPoetAS? PEZ 73.631 158761

Chat

And if you write a bot to extend Robot, consider entering it in the ExtendsRobotCompetition! -- Alcatraz

Does anyone know why SpinBot sometimes stops spinning and drives straight until it hits a wall? I've looked at the code and I can't figure out why it does that... --David Alves

No, but it can help me beat it, as it loses energy when it does so :) So far my main effort is beating SpinBot, but the best % I've got is 63.46, total of 126500. Actually I think Walls extends AdvancedRobot, making it invalid -- Alphax

Good point on walls, I removed it and posted my score so far. AntiSpinBot?? only gets hit in the beginning of rounds where it starts next to SpinBot and occaisionally when the "go straight bug" kicks in. Maybe I'll try again later, I'm going over to my girlfriend's house to watch the presidential debates.

1st: davidalves.robot.AntiSpinBot 59973 49900 9980 0 0 93 0 998 2 0
2nd: sample.SpinBot 6757 100 20 6479 33 124 0 2 998 0
AntiSpinBot? does not fire. :-)

--David Alves

My entry: dummy.St00pid. I bet I can get a better score if I implement some wallsmoothing and a better prediction of incoming enemyfire. But I'll do that some other time; it's 3am here right now, so I'm off to bed :-)

1st: dummy.St00pid 165318 49100 9820 88676 17380 340 0 982 18 0
2nd: sample.SpinBot 13585 900 180 11921 426 157 0 19 982 0
St00pid does fire. :-p

--Dummy

Only difference between Krazy 1.45 and the version I used for this is that this version doesn't change directions. With some tweaking it can be better. -- Alcatraz

 73.631%
1st: pez.femto.WallsPoetAS      158761	47800	9560	84425	16209	723	41	957	 43
2nd: sample.SpinBot	         30007	 2150	 430	26369	  746	310	 0	 46	956

-- PEZ

SpinBot goes straight when its turning (setTurnRight?(10000)) is complete. At that moment its moving (ahead(10000) is not yet complete. Therefor it goes in a straight line until its movement is complete. To correct this, just increase the number of turns to take. However, tampering with the samples is not allowed I think. ;-) -- GrubbmGait

The solution is of course to not get hit by SpinBot. Or hit it with bullets. That way it will die before it runs out of turns. -- PEZ

Actually, sample.Walls extends Robot, so it's results are valid... --Dummy

The reason sample.SpinBot sometimes just drives straight is when his "setTurnRight?(10000)" runs out and his "Ahead(10000)" still has some distance remaining. --Goofy

1st: dummy.St00pid 163352 49700 9940 86437 17261 13 0 997 6 0
2nd: sample.SpinBot 10840 300 60 10356 37 86 0 6 994 0

Nice scores everyone. By the way Dummy, this is PEZ's challenge not mine. --David Alves

My challenge? Hardly. This is the RoboWiki. So it's everyones challenge. But, yes, I founded it. I didn't find a reasonable way to sign the foundation though. =) -- PEZ

I mean you proposed the challenge. He credits me in his bot's source. :-P --David Alves

With firing:
1st: davidalves.robot.AntiSpinBot 156666 49950 9990 80642 16083 0 0 999 0 0
2nd: sample.SpinBot 3061 50 10 2984 6 10 0 1 999 0
score = 98.083%

Without firing:

1st: davidalves.robot.AntiSpinBot 59872 49800 9960 0 0 112 0 996 4 0
2nd: sample.SpinBot 5760 200 40 5304 69 146 0 4 996 0
score = 91.224%

--David Alves

David, grab my SpinBotHunter from the RobocodeRepository and stick my gun on your movement. --Goofy

Thanks for the offer, but you're a bit too late, I just fixed the bug in the gun. I forgot to convert from radians to degrees. Also added more intelligent dodging at the beginning and end of the round.

1st: davidalves.robot.AntiSpinBot 164500 50000 10000 87097 17403 0 0 1000 0 0
2nd: sample.SpinBot 1273 0 0 1264 0 9 0 0 1000 0
score: 99.232%

--David Alves

Wow Dave! Can't you publish the code now? -- PEZ

There's still a bug in movement. I'm still around 91% if I don't fire. I'll publish soon though. :-) --David Alves

Heh, well, since you seem to have gotten to the end of this challenge, I guess I can cease developing my SpinBotHunter. I was going to model what happens when he starts walking, so that I just have to scan him once, then I just have to fire exactly the right number of bullets and it's all over. But that seems useless. --Goofy

The end of the challenge is 100%. I may be ahead but I haven't won by any means. ;-) --David Alves

Is 100% the end? When writing "If two bots get the same score share - the Challenger with the highest score wins" I had 100% in mind. It's rather unlikeley any two bots would get the same score share otherwise. Anyway, I know I wouldn't be able to reach 100% with an extends Robot if I tried forever. When calls block, so do my brain. =) -- PEZ

What I meant was that this is a trivial question, now. Since Dave has some minor tweaks to prevent him from ever getting hit, and with some relatively minor tweaks, I could get a gun that not only never misses, but only needs to scan the enemy once, I don't think there's much more to look for. I very much believe that a Robot can get 100% against SpinBot, and I see that as the end. --Goofy

It has to be proven first I'd say. In 1000 rounds lots of start positions get tried. -- PEZ

I think the limit is roughly 99.7%. In other words about half of the damage I'm taking right now is preventable, the other half is just due to bad luck in start positions. --David Alves

with firing:

1st: davidalves.robot.AntiSpinBot 164143 50000 10000 86785 17357 1 0 1000 0 0
2nd: sample.SpinBot 890 0 0 880 0 10 0 0 1000 0
score: 99.461%

without firing:

1st: davidalves.robot.AntiSpinBot 60025 50000 10000 0 0 25 0 1000 0 0
2nd: sample.SpinBot 1501 0 0 1440 0 61 0 0 1000 0
score: 97.560% --David Alves

AntiSpinBot? 1.0 released. I don't think I'm going to work on this any more unless someone beats me. :-)

1st: davidalves.robot.AntiSpinBot 1.0 162997 50000 10000 85834 17157 6 0 1000 0 0
2nd: sample.SpinBot 755 0 0 727 0 28 0 0 1000 0
Code is here: David Alves/AntiSpinBot

Jar is here: https://davidalves.net/robocode/robots/davidalves.robot.AntiSpinBot_1.0.jar

--David Alves

Do I get any credit for inspiring the SpinBotChallenge :P ? --Alphax

I think you just claimed it. =) -- PEZ

While we're making extends-Robot challenges, how about samplebot melee? Your bot vs. all the sample.* bots that fire (everyone except Target and SittingDuck) on a 1000x1000 field. Melee seems like it would be very hard to do well if you can only use blocking calls, harder than 1-v-1 --David Alves

I'd like such a challenge if it's not constrained to extending Robot. But don't let me discourage such a challenge anyway. I am lousy at doing anything non-femto with Robot and I have never given Melee stuff attention either. Thus, my vote shouldn't count much. -- PEZ

If it's not restricted to non-advanced Robots then just stick Shadow or Aleph in there and it's all over. --David Alves

I wonder if you can sneak inside Spinbot's turn radius, in which case he'd never see you to target you. This is just idle speculation, but I think if I were to take up this challenge I'd sneak inside Spinbot's turn radius if I spawned close to him (and could make it before he started firing), or if I started far away I'd just snipe his predictable bullets out of the air. Meh.. I already spent too much time writing Ugluk.. -- Martin

I don't think it would be worth doing from a scoring standpoint. But personally, I think it would be a great trick to pull off. Clever! --David Alves

Walls could, at least most of the time, defeat all other sample bots. Walls extends robot... The robot should have something like this: All bots fire with HOT, so any movement should be fine MeleePatternMatching, when it is fixed, should do the trick for the targeting Target the closest bot! --Starrynte


Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited February 17, 2008 18:51 EST by 75-132-207-26.dhcp.stls.mo.charter.com (diff)
Search: