Robo Home | WaveSurfing | Changes | Preferences | AllPages

The DevilFISH Challenge

Are you WaveSurfing? Trying to get it perfect? There are a lot of things a wave surfer needs to address. One major thing is dealing with simple targeters with good and/or provocative movement. Marcelo?'s DevilFISH 1.3 is a bot hfires using some kind of LinearTargeting. And it also tries to fight very close. Evading enemies that are to close can be as simple as just spiraling away from them. But that's not very smart against LinearTargeting, is it? Here's the results from my latest CassiusClay b... uhmmm meating DevilFISH 1.3 in a 1000 round battle:

1st: pez.rumble.CassiusClay
2nd: Marcelo.Alpha.DevilFISH 1.3595200595200001000

See if you can beat that. Let's make it a contest:


Bot Author Score share
Dookious 1.534NDSVoidious 99.680%
PulsarMax 0.6.33 Pulsar 99.079%
Phoenix 0.312b David Alves 98.701%
Engineer 0.2.5 wcsv 97.709%
SilverSurfer 2.42 Axe 96.251%
CassiusClay PEZ 96.078%
GrubbmGrb 1.2.1 StopAndGo GrubbmGait 95.578%
DarkHallow jim 95.232%
SilverSurfer 2.35 HOTAxe 94.614%
Cyanide 1.41.2 Alcatraz 90.721%
CassiusClay with HeadOnTargetingPEZ modified by David Alves90.004%


I don't think these challenges are very scientific... You're not doing anything to control the other variables. For example, a bot with the same movement as CassiusClay but a much worse gun wouldn't do as well, since it wouldn't kill DevilFISH as quickly and therefore would give DevilFISH more time to shoot at it. The same applies to the /BarracudaChallenge. If you really want to do a new wave surfing challenge, use the same rules as the existing MovementChallenge: have the bot that's shooting at you hold still, and don't fire at it. That's the only way you can guarantee that your gun isn't affecting the outcome. I'll make some bots for a more scientific wave surfing challenge after work today if you like. --David Alves

1st: pez.rumble.CassiusClay with HeadOnTargeting132626494009880612151213000988120
2nd: Marcelo.Alpha.DevilFISH 1.3147296001201386813460129880

If this were *just* a test of WaveSurfing, changing to HeadOnTargeting would produce the same score. --David Alves

Interesting. Though none of these latter challenges where supposed to be scientific. They are tools in order to help people improve their movements (and guns, if that's where the work is needed). There is already the MovementChallenge for the more scientific tests. Your "more scientific" bots belong there. When you have them ready it's as easy to run CC there as to set "isMC = true" in the CassiusClay class. 90% against DevilFISH using HeadOnTargeting huh? I'm impressed by myself. Beat that! =) PEZ

I knew a DevilFISH challenge would wake you up Axe. =) Welcome back! -- PEZ

Just for fun (and also trying a "scientific" approach), I made a HeadOnTargeting version of SS... The result of 94,614% is way over my best expectations, and probably shows how weak is my cannon (maybe not that weak, but definitely HEAVY, it makes me skip some turns)... IŽll try the same HOT version in BarracudaChallenge... -- Axe

That score is mad! Well done Close Combat Master! -- PEZ

Whoooooo! Pulsar. What have you fed PM? Congrats! -- PEZ

Fed him some Marcelo.Alpha.DevilFISH! I'm quite surprised as well, just ran one try and came up with this and it seems to be quite reproducible. The other challenges are the best recorded results for PM, this is just about ordinary. The now-in-the-works 0.5.21 is currently scoring 98.567%. I'm suprised I'm not scoring better against BotB? now instead. --Pulsar

Well, todays boring insight is that DevilFISH doesn't move unless it thinks you're firing at it... Which makes the DF Challenge II a bit too different from the general challenge... However, if you have a good movement you can obviously score better (from an RR@H point of view) against DF by not firing at it. Maybe there are many bots like that? -- PEZ

Hmm, CC scores better by not firing at DevilFISH? Maybe you should put some if (!DevilFISH){fire();} logic into CC anc collect a few rating points. :-) --David Alves

I had noticed this some time ago, actually, lots of bots have this kind of behaviour, Barracuda is other. -- Axe

1st: davidalves.Phoenix 0.312b  	142893	50000	10000	69078	13815	0	0	1000	0	0
2nd: Marcelo.Alpha.DevilFISH 1.3	1881	0	0	1880	0	1	0	0	1000	0
98.7% --David Alves

Added the score for Dookious 1.09 (gun data saving turned off) over 30 seasons. -- Voidious

1st: gh.GrubbmGrb 1.2.1			145364	49950	9990	70641	13003	103	1676	999	1	0
2nd: Marcelo.Alpha.DevilFISH 1.3	6726	50	10	6630	15	20	0	1	999	0
Now against the right DevilFISH --GrubbmGait

Doh, I fought against wiki.nano ... I'll re-run later against the right one. -- Voidious

Ok, 96.94 instead of 96.91 against the "right" DevilFISH. -- Voidious

Wow, color me surprised - first place by a margin with my recent movement tweaks. That's averaged over 3 seasons. -- Voidious

Robo Home | WaveSurfing | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited April 1, 2007 4:52 EST by Voidious (diff)