"Multiplayer" AI challenge on StackExchange right now: Create Your Wolf

General Discussions
#21

I must admit i took his results on face value… my own limited testing of the deep wolf has similar results to you i.e. 70’s

#22

Finally posted SpyWolf

I’m happy that I could come up with a 100% original algorithm that is as competitive as it is; here’s two random 5-run averages (the same as the real competition, albeit missing some breeds):


HerjanWolf               85.0
SpyWolf                  85.0 * 
HybridWolf               82.8
DeepWolf                 82.4
HonorWolf                82.0
OmegaWolf                81.4
ProAlpha                 79.6
ShadowWolf               73.6
AlphaWolf                72.6
StoneEatingWolf          63.4
LazyWolf                 48.0
BlindWolf                44.0
Sheep                    40.8
LionHunterWolf           28.8
CamperWolf               27.4
GatheringWolf            23.6
EmoWolf                  20.0
WolfWithoutFear          16.0
GamblerWolf              15.0
MOSHPITFRENZYWolf        8.4
Stone                    2.2
MigratingWolf            2.0
Wion                     0.2
MimicWolf                0.2
WhenIGrowUp              0.0
WolfRunningWithScissors  0.0
Bear                     0.0
Lion                     0.0

SpyWolf                  85.2 * 
OmegaWolf                83.4
DeepWolf                 83.0
HerjanWolf               81.8
ProAlpha                 80.6
HybridWolf               79.8
HonorWolf                79.0
AlphaWolf                73.6
ShadowWolf               70.6
StoneEatingWolf          66.0
LazyWolf                 51.0
BlindWolf                44.2
Sheep                    40.8
GatheringWolf            26.4
LionHunterWolf           25.6
EmoWolf                  25.4
CamperWolf               24.6
WolfWithoutFear          16.8
GamblerWolf              14.4
MOSHPITFRENZYWolf        10.4
MigratingWolf            1.2
Stone                    1.0
MimicWolf                0.8
Wion                     0.4
WhenIGrowUp              0.0
WolfRunningWithScissors  0.0
Bear                     0.0
Lion                     0.0

It’s not hard data, a 5-round run is still pretty loose, but I think SpyWolf does well.
Also Deep is still being strange, I’ve witnessed it pull an 84+ and then a 58. (and the 58 was a 200-run average!)

#23

Congratz @Herjan for being top of the current official leaderboard!

Alas, SpyWolf had a couple poor runs, ruined it’s average.

#24

Is it too late to give it a go?

I’m thinking of having a go at using neural networks to work out a decent AI for me.

#25

i am not even sure that there is a end date… so more the merrier

#26

I thought of that as well, would have to do a lot of training though, and re-train whenever new wolves are added.
Wonder if there’s a secure way to have it download test data from a sever to kick-start the AI, and upload new data whenever someone runs tests with it. Outsourced data collection!

#27

Yeah, finally people respect me for my work! (joke) I’m so pathetic :slight_smile:
But it feels good, nonetheless. And moogie is second! (One wolf difference, :-*)

Pizza, I see you have uploaded your ‘Spywolf’, he has a nice score as well (top 5)!

And BTW:

[quote]You should really fear Hybrid, it’s nearly unstoppable.
[/quote]
I just thought of a funny way to learn him a lesson, see my comment on his wolf :stuck_out_tongue:

#28 

private static int wolfCount = 0, startCount = 0;
private static boolean startedCounting = false;
private boolean sabotage = false;

public ReallyGoodWolfYouShouldUseMrHybrid() {
     super('W');
     if (isHybridLoaded()) {
          if (!startedCounting) {
               startCount = wolfCount;
               startedCounting = true;
          }
          if (wolfCount < startCount + 100) sabotage = true;
     }
     wolfCount++;
}

Using reflection to detect whether or not HybridWolf is loaded, and knowing how many of this breed of wolf have been instantiated, we can sabotage Hybrid while the other wolves function normally.
Of course heavy obfuscation is in order.