Gold Collectors KoTH(s)

Redwolf Programs

00:05

Apparently, the motion of other bots confuses CloseCoin's algroithms

Redwolf Programs

00:18

But even without hunting capabilities, it ate Hunter, GUT, Third Gen Learning Algorithm, and Wild Goose Chase

By the way TBTPTGCBCBA eats Potentially Victorious quite often

Ramillies

Which version?

You're probably pitting the new TBTfqwjkenwhatever against the old Potentially Victorious?

Redwolf Programs

Oh, I'm running a pretty outdated everything

Ramillies

If you want, grab the botdata.js from the gist with my controller.

That should be pretty up-to-date.

Redwolf Programs

Nah, I'm just testing my bots at the moment. Hunter and Seeker are getting a new Indev defense system and teaming mechanism

#Constant6thPlacer

Hunter had a bug where enemies on his eastern side were hidden

Redwolf Programs

03:14

CloseCoin is becoming a deadly opponent

Aaand 1st place goes to CloseCoin by 300 coins

)Well, 104..same difference)

[1919] Bot CloseCoin ate Potentially Victorius and gained 272 coins

That's what I call BOOYAH

I got my comeuppance

[819] Bot Potentially Victorius ate CloseCoin and gained 127 coins

Redwolf Programs

03:52

Any strategies to beat TBTPTGCBCBA?

Night2

05:44

About running games, right now with 26 bots (updated last night) I can run 10K in about 1 to 1.5 second inside Google Chrome (no graphics). But the funny thing is, if I set a loop to do 50K, it will get stuck for a long time. I'm not sure what causes that, but doing 5 separate 10K runs is faster than a single 50K run!

And my PC isn't powerful at all, I have a "Intel® Core™ i7-4700HQ" CPU which still is good, but since only 1 core is used while running controller loops, at best an old 3.40 GHz core is running all those operations!

But anyways I think we should stop with non-serious bots, they all die fast and just leave the game open with a very large arena and a few bots in it at the end. Usually when I run 1K rounds, I see that just usual top bots and some survival bots (with very few coins) are around in a very large arena which reduces the chance of confrontation between top bots.

@Ramillies your controller looks quite good now

Beta Decay

10:55

Converted to JS:

1

A: Gold Collector KoTH

X Marks The Spot | JavaScript function(me, others, coins){ if (me.locationY != 0) { // If not on X axis if (others.every(other => other[1]==me.locationY-1)) { // If any in my way if (!others.every(other => other[0]==me.locationX-1)) { ...

Ramillies

@Night2: Thanks, glad you like it.

Redwolf Programs

14:15

@Night2 Intel Celeron N2830 Dual Core 2.16 GHz Processor. 2 GB DDR3L SDRAM. 16 GB Internal Storage. 11.6-Inch Screen 1366x768 Active Matrix TFT Color LCD, Intel HD Graphics.

Acer Chromebook 11

Night2

@RedwolfPrograms OK, that one is weak

Redwolf Programs

Can you link to your botData?

Night2

What do you mean?

Redwolf Programs

Mine isn't updated

Just your botData variable where the bots go

Night2

Oh, I'm using the one from @Ramillies controller

gist.github.com/ramillies/c2ca193f0d98039a5b2800fb39f5d62b

Redwolf Programs

14:17

Okay. By the way, I'm about to update Hunter

Night2

I'm also working on my bot

Seems other bots are way better than it now

Redwolf Programs

All of my bots have a few bugs with their survival systems

Night2

Yeah, they still die very early with few coins

Redwolf Programs

Hunter is lasting about 1k rounds now, rather than 10

With about 100 coins

Ramillies

@Night2, TBT...CGA regularly wins with my setup.

Redwolf Programs

14:19

Hunter couldn't see enemies on his Eastern side

Night2

TBTPTGCBCBA wins if you play on a few rounds or if it gains a good lucky lead

Redwolf Programs

Yeah

Night2

But in 100K runs, other bots have better end game strategy and they eventually gain more coins

Redwolf Programs

I think too much of the game is luck, honestly

But the extra rounds will take away the luck element

Ramillies

They won't.

Night2

14:21

No, luck is affecting it, but it is obvious to me that bots like "Polite Near-Sighted Drunk Bot" and "Big King Little Hill" are playing the end game better

fəˈnɛtɪk

How well does safety coin do now?

Night2

Even Safety coin sometimes performs very well

Ramillies

Luck has a good part here, but obviously some of the bots regularly winning can't be luck.

Night2

And the new Potentially Victorious which was on @Ramillies data behaves very good as well

@Ramillies btw I tried to use your post script and it didn't work, seems to have some errors, but the one with your controller is working well

Ramillies

@Night2, that's weird since what's in the post should be a direct copy. I'll have a look.

Night2

14:22

@fəˈnɛtɪk I found it many times dead, apart from that, when it makes to end game, it performs well and I have seen it win some times

Ramillies

Ouch, I already see the mistake.

Night2

huntingTimer

is not defined

Ramillies

Yes, the part that loads it from the botNotes is missing.

fəˈnɛtɪk

The bots which have a chance to not move while hunting enemies have a chance to kill safety coin. If I remove the ability for that to happen it makes it easier for it to get cornered

Ramillies

Hm, I'm stupid, I just copied the function for potential I think...

Yes I did.

Redwolf Programs

14:25

I'm running a test with the fully updated Hunter.Armageddon and CloseCoin

Ramillies

Should be now fixed, thank you.

Redwolf Programs

Hunter.Armageddon has a new feature that should make it stop getting in endless chases

Night2

As number of bots is growing and the more complicated the bots are getting, running 10Ks is getting slower and slower :D

Redwolf Programs

In my tests, CloseCoin does remarkably well

Night2

I used to be 20K in less than a second, now 10K take up to 1.5/2 seconds

Ramillies

14:29

I'm running only 2ks and they are slow enough.

Night2

@Ramillies your controller deosn't have the no graphics run feature

I do that manually in console with default controller

Ramillies

Yeah. I'm seriously contemplating adding it.

Night2

for(var xyzusnasd = 0; xyzusnasd < 10000; xyzusnasd++) runBots();

I just do this

And then

updateLeaderboard();updateCanvas();

Ramillies

I see.

Night2

If I update the graphics in each loop cycle, it will take forever :P

@RedwolfPrograms I can't find a bot named "CloseCoin", where is it?

Ramillies

14:32

I update them each fps/50th cycle if fps is large enough.

Redwolf Programs

In development

Ramillies

But I'll ditch that and add a separate textbox.

Night2

@Ramillies I tried with very high FPSes like 2000, but it doesn't go faster than a specific speed on my PC

Redwolf Programs

Mine goes up to about 150

But that's with graphics and leaderboard every turn

CloseCoin is designed to chase for up to 4 turns, then stop to prevent long, boring wastes of time

Ramillies

@Night2, probably the setInterval function won't fire up the callback until the previous had finished.

(I guess. Otherwise there would be very rough concurrency troubles and I would've probably noticed.)

Night2

14:41

It probably is a hardware limit on my side

Ramillies

Yeah. It makes sense: otherwise you would be able to make the computation as fast as you would want.

There has to be a cap somewhere.

Anyway, I added the textbox to render the canvas and leaderboard only each n-th round: gist.github.com/ramillies/c2ca193f0d98039a5b2800fb39f5d62b .

Redwolf Programs

Seeker needs to be reprogrammed to avoid enemies

Before, she just considered avoiding them, but a coin on the other side would cause her to run right into the enemy

Night2

Gold makes her blind :D

Redwolf Programs

LOL

Well, that went great. Seeker died instantly

CloseCoin tied for 2nd with Potentially Victorious

After 1e3

Ramillies

YEAH! You finally spelled my bot's name correctly!

(Although I have probably no right to say such things.)

Redwolf Programs

14:57

Yeah, I kept thinking "Something's wrong with my spelling, but I can't put my finger on it"

I seriously annoy people with my Grammar-Correcting

No! CloseCoin got eaten!

Night2

@Ramillies 🤨

Redwolf Programs

Firebird is getting upgraded EXTREMELY

Now, it moves in Clockwise or Counterclockwise circles?

After 4e3, the winner looks pretty obvious

By the way, Firebird was a joke. The new Firebird is like Wild Goose Chase, but with a few extra abilities

let firebird = new WildGooseChase("Okay").makeAwesome("cool features") * 1000000;

Ramillies

15:24

Potentially Victorious got an Extra Aggressiveness update.

Redwolf Programs

Yaaaay, just what we all needed

Ramillies

Now it even beats TBT...

(by ~ 1% of games...)

(Results of 100 games, 2k rounds each.)

Night2

I doubt if in +50K round games any of us can win like that.

@RedwolfPrograms is your decision on 100K rounds per game final?

Redwolf Programs

No, maybe less

Even 10k may work, but I was thinking 20k

Ramillies

@Night2, who does the best in those long runs?

Night2

15:36

Can you make a final decision on that? I want to program my bot behavior depending on how close we are to the end

Ramillies

I would go at most 10k, maybe only 5k. After that long the winner is usually crystal clear.

Redwolf Programs

If anybody has a problem with 20k, tell me in the next 0.85ms

Okay, we've reached a unanimous vote

Nobody contacted me in the last 0.85ms

Night2

@Ramillies Big King Little Hill and Polite Near-Sighted Drunk Bot and sometimes Safety coin if stays alive

Ramillies

I see.

My endgame logic is definitely a bit lacking (maybe a sizable bit, I'll have a look for myself).

Night2

@RedwolfPrograms so is 20K final?

I am thinking to try a risky approach if we are close to end of the game and I'm not in lead to try and corner a weaker bot somehow. Not sure if I can do it or not, must think on it, but for that I need to know finalized amount of rounds per game

Ramillies

15:41

I've been trying to do something like that with Potentially Victorious.

With the old version of TBT..., it was successful because I was often able to eat him and get a gazillion of coins.

But as soon as you updated him, it ceased to be worth the hassle.

Night2

@Ramillies Your controller was buggy, TBTPTGCBCBA is not eatable unless 3 stronger bots corner him :D

Ramillies

Is or was?

I mean the first version (probably).

Night2

was

Ramillies

Hm, OK, then the strategy was useless right from the start :—D

Night2

Even the first version had similar running technique

Only on the real corner of board 2 bots can kill 1 weaker bot

in middle, you will need 3 bots

Ramillies

15:44

I know.

I've been trying to make PV ram the weaker bot into a bunch of more powerful bots.

Night2

But I'm thinking to pick on the weakest bot on end game with hope of another random chase helping me suddenly change the winner

Beta Decay

It would be cool to use the Stack Exchange API to automatically update the bot data

Night2

But anyways I need to know how long each game is going to be to be able to select the right time for strategy change

Ramillies

In hope that I will be able to eat it. // It won't help you with anything if you don't eat it.

Night2

@Ramillies But when it is close to the end of the game and I'm not in lead, my only hope will be eating a weaker bot

Ramillies

15:46

Of course.

Night2

Can I put a copyright on this idea? :D

Ramillies

I just want to say that I had exactly the same thoughts before and I now think that you should rather hope for the gold coin spawning right next to you 100 times in a row.

If you want, do so :—D.

Night2

Just kidding

I might add another bot with task of helping to corner a weak bot at the end

^_^

The Bot That Plays The Game Sensibly But Suddenly Goes Crazy At The End

Night2

16:08

@Ramillies TBTPTGCBCBA also seems to be the most aggressive bot with most kills. I like that one :D

If I don't win the real competition, I will win the award for most kills :P

Redwolf Programs

16:20

20k is final

fəˈnɛtɪk

I had a round where safety coin somehow had 9 kills.

I have no clue how that happened

Redwolf Programs

3 days to final competitions! Would anyone like to volunteer for running a small percentage of the rounds?

The more people that volunteer, the faster it goes

Night2

@RedwolfPrograms I might be able to do that, but since I'm part of competition, I'm not sure if that is fair to do that

Redwolf Programs

I would give you a version of the controller that is designed by me

And ensure that the results are valid

Night2

Well, someone can always change the controller or change the results

Redwolf Programs

16:24

Although I doubt any of y'all would cheat

Maybe I could have it encrypt the results (:

Night2

Of course we won't, but it also doesn't feel fair

I think you should ask from more people about this and see what they think

Redwolf Programs

Well, obviously, if Goldilocks gets 80 first-place wins I'll know something's up

And I'll compare results and see if anything's odd about them

Night2

@RedwolfPrograms Well as long as the code/result are on same PC, anything can be cheated, your encryption can be decrypted as well

Redwolf Programs

True

I'll think about it, and ask around

Maybe there's something on Meta

Night2

@RedwolfPrograms how many rounds per seconds can you run now? And how many games do you plan to run?

Redwolf Programs

16:29

I don't know how many rounds per second (I'm going to check in a few minutes)

I plan on 8k games

Maybe I could ask a question on Meta about having competitors run some of the games?

Night2

That is a good idea

Fact: each time I benchmark a 20K run, I get a very different number, from 3-4 seconds to 15 seconds

It seems based on how each game goes and what bots stay alive longer, the times do varry

Fact 2: Google Chrome's JS engine is a lot faster than Mozilla Firefox's one

fəˈnɛtɪk

I learned that Chrome runs JS faster when writing JS code for a fastest code challenge :P

Redwolf Programs

Hmm

Beta Decay

16:52

That's stupid

fəˈnɛtɪk

This challenge and this challenge were the ones which I tested the code in Chrome and other browsers

Ramillies

17:23

@Night2: TBT... has most kills, but Potentially Victorious has usually a bit more coins in them.

Redwolf Programs

TBT... just died, same as BKLH

Seeker jumped!

fəˈnɛtɪk

With the bot data included in ramillies botdata, safety coin pretty much either dies early or ends up 2nd :P

Redwolf Programs

Nobody is at all running any of the same code as anyone else

Has BaitBot been fixed yet?

0

Q: Should competitors be able to run games for KoTHs?

In this question, every "game" takes about 5 seconds. Ten thousand runs, maybe eight thousand, are planned. Running these on a single computer would take forever, so I suggested letting competitors and volunteers run some of the games. However, this result in cheating. My question is: should this...

Don't you hate it when your two flagship bots eat each other?

Beta Decay

19:02

@RedwolfPrograms Should be fine as long as everyone uses the same controller

Ramillies

Exactly. I don't really suspect any cheating in here.

Ramillies

19:17

Potentially Victorious just got a kingly meal:
[16642] Potentially Victorious has eaten Feudal Noble for 1606 coins.

Transcript for

Gold Collectors KoTH(s)