[Home]WaveSurfingChallenge

Robo Home | Changes | Preferences | AllPages

Difference (from prior author revision) (major diff, minor diff)

Changed: 39c39,40
|Freya 0.62ws2 |Loki |MR/ WaveSurfing | 75.36 | 33.38 | 54.37
|Freya 0.62ws2 |Loki |MR/ WaveSurfing| 75.36 | 33.3 | 54.37
|Evader 1.1 |Starrynte |S&G | 71.581 | 35.139 | 53.36

Added: 129a131,133

Evader is failing miserably...due to it's bad WallAvoidance...I'll upload new results once that's fixed --Starrynte
* Lol, in one of my testing, it 'accidentally' (not) WallSmoothed right into the BotA? (when it was disabled) and killed it - free points --Starrynte

Introduction

The WaveSurfingChallenge gives you a way to measure your bot's ability to avoid getting hit by very simple targeting systems in long battles. The WaveSurfingChallenge/FastLearning pits you against the same bots in 35 round battles like the ones in the RoboRumble@Home. You may also want to check out these other movement challenges:

Directions

Here's how you run a WaveSurfingChallenge:

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

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

Results
Bot Name Author Type Score vs. Bot A Score vs. Bot B Overall Score
Phoenix 0.30.1devDavid Alves WaveSurfing 99.952 99.548 99.75
Ascendant 0.9 Mue WaveSurfing 99.664 98.92 99.292
CassiusClay 1.9.9.77 PEZ WaveSurfing 99.948 98.304 99.126
PowerHouse 1.5e wcsv WaveSurfing 99.68 98.16 98.92
Pear 0.61.4dev iiley WaveSurfing 99.115 98.285 98.7
PulsarMax 0.5.20 Pulsar WaveSurfing 99.76 97.2 98.48
DarkHallow jim WaveSurfing 99.637 96.288 97.963
Crusader 0.1.2 trab WaveSurfing 99.904 94.56 97.232
Dookious 0.65 Voidious WaveSurfing 98.660 95.128 96.894
SilverSurfer 2.44 Axe WaveSurfing 99.972 91.765 95.8685
Cyanide 1.65 Alcatraz WaveSurfing 99.32 91.644 95.482
GrubbmGrb 1.1.3 GrubbmGaitStop&Go 85.536 94.396 89.966
Squirrel prerelease Bayen Oscillating 79.33 96.528 87.929
Chalk 1.01 Corbos WaveSurfing 98.684 53.234 75.959
Freya 0.62ws2 Loki MR/ WaveSurfing 75.36 33.3 54.37
Evader 1.1 Starrynte S&G 71.581 35.139 53.36
SandboxDT 1.91 Paul EvansRandomMovement 41.929 37.738 39.834

Results for SandboxDT 1.91 are included for reference, it has one of the best non-surfing movements.


Chat

Great score David! -- PEZ

Cant believe that Shiva only scores around 89% overall. Hope i did something wrong when testing :((( --deathcon

Just about everyone was scoring lower than that at first. What are your individual scores vs. A and B? --David Alves

Not only that. It doesn't seem to matter much for the real life performance of a bot. Some of the bots scoring sub 90% before and 97+% today have about the same RR@H score anyway. Shiva is probably alright man. =) -- PEZ

One problem that axe had was that his anti-ramming code was kicking in on rounds where the WaveSurfingChallenge bot was facing SilverSurfer. Something similar perhaps? --David Alves

Yeah, the anti-ramming is important in WaveSurfingChallenge, Pear did this not good, sometimes round begin, he is very close to enemy, but he cannot escape soon(I cannot find a good way to improve this all through), so it get low score here. -- iiley

But... I dont think SilverSurfer has this problem, if he/she has, he/she cannot get as high as 99.9+ score vs BotA?. -- iiley

Not any more, Axe fixed it. I was saying that deathcon might have some problem like that in Shiva. --David Alves

Yeah..., maybe he has the same problem with me, now i'm tring to fix it. But it's hard for my movement. Any way maybe can fix out a little better way. -- iiley

Details:

 1st: davidalves.Phoenix 0.27                 60001  50000  10000     0  0  1  0  1000     0  0
 2nd: wiki.challenge.WaveSurfingChallengeBotB  1144      0      0  1144  0  0  0     0  1000  0

 1st: davidalves.Phoenix 0.27                 60000  50000  10000    0  0  0  0  1000    0  0
 2nd: wiki.challenge.WaveSurfingChallengeBotA   164      0      0  164  0  0  0     0 1000  0
--David Alves

Cool down, David, did you did your surfing a especial to simple enemies as CC did?(if been hit times lower than x, then did especially). -- iiley

My surfing code doesn't have any special cases. --David Alves

Which means you don't have any visit flattener I guess? CC only activates its flattener when the enemies hit rate goes beyond a certain level. Other than that CC prepopulates its unsegmented hit stat buffer with one GF0 hit before the fight. -- PEZ

Exactly. I have a flattener but it makes Phoenix a horrible SlowBot so it's turned off for now. I'm not sure it helps anyway... it's probably buggy. --David Alves

Details:

1st: pez.rumble.CassiusClay 1.9.9.77              60001	50000	10000	0	0	1	0	1000	0
2nd: wiki.challenge.WaveSurfingChallengeBotA 1.0  52	0	0	52	0	0	0	0	1000

1st: pez.rumble.CassiusClay 1.9.9.77              60000	50000	10000	0	0	0	0	1000	0
2nd: wiki.challenge.WaveSurfingChallengeBotB 1.1  1696	0	0	1696	0	0	0	0	1000
-- PEZ

Although GrubbmGrb is in noway a WaveSurfer, the sole purpose of it's Stop&Go movement is to avoid simple targeters. Therefor I entered it here, and it doesn't do it that bad. --GrubbmGait

I know Dookious is still a ways from being the ideal WaveSurfer, but I figured it couldn't hurt to add his scores here. -- Voidious

Details:

1st: cjm.Chalk						60010	50000	10000	0	0	10	0	1000	0	0
2nd: wiki.challenge.WaveSurfingChallengeBotA 1.0	1316	0	0	1316	0	0	0	0	1000	0

1st: wiki.challenge.WaveSurfingChallengeBotB 1.1	58763	7500	1500	46766	2996	0	0	150	850	0
2nd: cjm.Chalk						51019	42500	8500	0	0	19	0	850	150	0

If a good score doesn't guarantee real-life performance, does a bad score come with guarantees? :) Yuck. --Corbos

Well, it is always nice knowing that your tank is doing well *despite* disappointing challenge scores :) One thing that comes to mind off the bat is that you might not have your EnemyWave segmentation actually in sync with what they are seeing before they fire... There's some good stuff on that EnemyWave page that's worth checking out about which tick's data to segment. (Basically, 2 ticks before the energy drop.) Just a thought, I gotta get back to work... -- Voidious

Also, this is just a tool to perfect your movement against the two simplest (of many) targeting methods. It just means you will score say 95% against a lineartargetter instead of the possible 99.x%. On the other hand, I've read somewhere that you rating will be influenced noticably if you could pump up you score from 95% to 97%. No calculations at hand to prove this though. -- GrubbmGait

I started work on a combined MinimumRisk and WaveSurfing/ShrapnelSurf type of movement for my melee bot Freya. My first attempt obviously works reasonably well for HOT, but fails miserably for LT... I still have a looooong way to go. But that is part of the fun :) --Loki

1st: davidalves.Phoenix 0.30.1	                60000	50000	10000	0	0	0	0	1000	0	0
2nd: wiki.challenge.WaveSurfingChallengeBotA	48	0	0	48	0	0	0	0

1st: davidalves.Phoenix 0.30.1	                        60000	50000	10000	0	0	0	0	1000	0	0
2nd: wiki.challenge.WaveSurfingChallengeBotB 1.1	452	0	0	452	0	0	0	0	1000	0

Beat that. ;-) --David Alves

Loki, is that bvh.mini.Freya or bvh.fry.Freya? --David Alves

It's bvh.fry.Freya. I can't get WS working in a megabot, let alone in a mini :) But i am still working on it. I have slightly better results against BotB?... But i have broken something against BotA?... B.t.w. i showed the wrong version in the results above, the Freya entered in Melee at this moment doesnot use WS at all. --Loki

Well I was going to offer to take a look at the code and see if I could spot any problems... but then I opened up the java file and saw ++beweging.aantalBotsingenMetWand;... Might take a while for me to understand what's going on in your bot... =P --David Alves

thanks for the offer! I will struggle some more, keep coding my bots using a combination of Java and Dutch and if i get really stuck, you can always enter the code in Altavista's Babelfish :) It gives 'comprehensible' translations like "++ movement.number collision with partition;". At least 4 out of 5 words are correct! --Loki

Evader is failing miserably...due to it's bad WallAvoidance...I'll upload new results once that's fixed --Starrynte


Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited August 19, 2007 2:44 EST by Starrynte (diff)
Search: