Gold Collectors KoTH(s)

Redwolf Programs

00:08

Only two more bots!

Going to dinner, though

Redwolf Programs

00:52

1373 lines of bots and controller functions...let's test it!

Redwolf Programs

01:25

Okay, Execution.js in github has all bots included. Time to run one execution...

Night2

05:09

I tried to run you latest controller. But there seems to be many issues. Many of the bots are buggy. I removed them, but still controller seems to be buggy as well.

The botvariable in botNotes.storeData and botNotes.getData tends to be undefined and cause errors.

After I fixed that and ran the game for a few hundred rounds, it looks to me that the controller is not giving any points for coins and doesn't change coin locations

For example, looking at botData, my bot is sitting on [ 18, 5 ] for a long time and its coins are always at 0, looking at coinLocations, it is stuck on this for a long time: [[ 3, 11 ],[ 6, 19 ],[ 3, 21 ],[ 18, 5 ],[ 22, 9 ]]

sundar

10:02

@Night2 Could you link to your version of the controller with these fixes?

Night2

10:20

My fixes were just ignoring errors in botNotes with try/catch and removing bots that caused errors so I could run the controller. When I saw controller is buggy in other places, I just discarded it.

sundar

Alright, so the problem with botNotes was that getData and storeData should be using caller.index instead of just caller (and getData should use its own name to get the caller)

But that still leaves an issue when the bots use an inner function to call storeData or getData

because the inner function is the caller in that case, so the code can't figure out what the bot is

(it could try calling the caller if caller.index returns undefined, but there's hopefully a better less hacky solution than that)

sundar

11:06

@RedwolfPrograms I think the coins being stuck at 0 problem, that @Night2 mentioned, is solved by moving the ` while (arrayIndex(coinLocations, b.loc) !== -1)` loop to go inside the for loop that comes after it

Also, maybe the for loop for (n = 0; n < dat.length; n++) {(that comes after "//if three+ bots in same space") should use dat[n] and push[n], instead of i?

(Both of these are based on a cursory look and trial & error though, I haven't tried to figure out what all of it does.)

@OMᗺ Thanks for letting me know! I've deleted the original Katamari bot anyway (since it was pretty much a reimplementation of yours), but I've updated the two other bots that used the same code (Katamari with Values and Damacy)

@RedwolfPrograms Could you take into account the updates I've mentioned above ^ when running the bots?

Redwolf Programs

12:38

I'm aware of a few problems, but thanks for letting me know of the rest. I'll fix these when I get back in around an hour.

@Sundar They should be easy to fix, so I'll fix them and debug the bots once I get back

N. P.

13:20

One option for the data storage is to modify each bot to send three arguments: its own name, the key, and (when necessary) the value

Idk if you've already solved it though

Redwolf Programs

13:35

I don't think that's necessary.

Plus, it adds the "I still accessed (Someone else's) botData" loophole

N. P.

How'd you get it working then?

I'm not very familiar with javascript

Redwolf Programs

Well, if Sundar is right, it's a few simple errors I made

I put a code for the bot in a property of the function, but forgot to access it.

@Sundar yup, they were pretty simple issues. Of course, the 2 parts of the controller I didn't get around to testing have the most errors...

I'm going to debug the bots now. Wish me luck...

Wait...deleted Katamari?

Beta Decay

14:01

@RedwolfPrograms It was the same bot as Greedy, just in a different language and with a different name

Redwolf Programs

Oh

Potentially Victorious just potentially jumped off the board. And it did.

Ramillies

What?

That's interesting.

Redwolf Programs

(0 <= x < me.arenaLength) returns true always

Ramillies

Why is it so?

Redwolf Programs

I'm not even sure. But all I know is it does

Ramillies

14:08

x is not the position of the bot but the position where the potential is calculated. So in theory it should work OK.

I'll do some debugging in the TIO. Is the finalized controller published?

Redwolf Programs

But any numbers in (n <= n < n) always returns true

The finalized controller?

Oh wait, let me go publish that

Ramillies

The version without bugs.

By the way, thank you for that expression. I was so used to it.

But JS has obviously problems with this kind of comparison.

If any x <= y < z returns true.

Redwolf Programs

I tried (x < 0 || x >= me.arenaLength || y < 0 || y >= me.arenaLength), but it still seems to head north at [0,0]

Ramillies

Hm... I will really need to have a look at it in the TIO. Maybe JS doesn't know that Infinity is larger than every other number? :—)

Redwolf Programs

It's getting NaN as x and y vars

Ramillies

14:14

Well, that baffles me completely.

I haven't the slightest idea where any NaN could come from in there.

Redwolf Programs

I'm checking something

But this is an interesting test!

Whoops!

About half of it was me testing it with the wrong controller, the other half was the bot.

It works well now that I switched (n <= n < n) and used the most updated testing controller

@Ramillies Fixed

Okay, got a South/

Ramillies

That's a good thing. (At least if it gave South anywhere else than on the southernmost boundary of the arena :—))

Redwolf Programs

:-)

Ramillies

I tested it in the TIO and it luckily won't get off the arena too.

Redwolf Programs

Good

I'm making progress debugging. I

am at The GUT

Ramillies

14:23

I wish you luck with that. I think it should have been called the Euler-Lagrange bot.

Redwolf Programs

It seems to always return none, even when a way more powerful bot is right next to it

Ramillies

:—D

Redwolf Programs

The GUT is thinking that the first parameter is an array.

It's not.

I have my work cut out for me, fixing this bot

x_acceleration = NaN
y_acceleration = NaN
You guys hate numbers

Mike Nichols

Don't you think bug fixing every bot might be a little above and beyond the call of duty?

Ramillies

I wanted to ask the same thing.

Redwolf Programs

14:33

Well, I need practice, and it's important that I run a bunch of functional bots

It's no fun if they don't work

Ramillies

Normally, anybody organizing KotH will just tell the buggy bot that it has this and that bug, and the people fix that for themselves.

Redwolf Programs

Problem is, by the time they see the message and fix it, I could have already done it, right?

Ramillies

Right, but you could be doing different things in the meantime.

Redwolf Programs

Unfortunately, I have nothing else to do

and it's kind fun

kind of fun

Ramillies

You could tell everybody which bugs they have and come to collect the fixed bots the next day. But of course if you want...

then the people will surely be glad that you fixed it for them.

Redwolf Programs

14:35

The GUT seems to have no idea what to do when it has no coins

Ramillies

That is because all of its decisions are made according to the value k = coins of something to grab / its own coins.

Apparently spreading NaNs through the whole system like mad.

Redwolf Programs

Yeah

Once it gets it's first coin, it functions

fəˈnɛtɪk

Ramillies to fix this problem in my proton bot I used Math.max(1,own coins)

Redwolf Programs

I'm going to let the owner fix that

Ramillies

I notified him in the comments.

Redwolf Programs

14:39

Yeah, I just fix bugs (Not complete mistakes in the formula)

Ramillies

Well, I just use -k / (r + 1/5) in order to circumvent that.

(in the other place where there is a potential division by 0)

Redwolf Programs

Math.max(0.1, me.coins) works fine for this

but I won't fix it until @BetaDecay does

Ramillies

Yeah. Everybody has their own view on that. I, for instance, don't think that the division is the right thing to do here, so I would abandon it completely.

BetaDecay acknowledged the comment, so he will be probably fixing it soon-ish.

Redwolf Programs

Damacy is casuing errors already

Night2

You need to let everyone know about their bugs

Redwolf Programs

14:44

Oops, it's because botNotes wasn't updated in time

Night2

Otherwise, when they update their bots, you will have to fix same bugs again!

Redwolf Programs

Hmm

Beta Decay

Done

Redwolf Programs

@BetaDecay While your at it, me[0] should be replaced by me.locationX and so on

me[1] should be me.locationY

me[2] should be me.coins

and finally, me[3] should be me.arenaLength

Beta Decay

How about others

And coins?

Redwolf Programs

14:46

They're fine

Beta Decay

There you go

Redwolf Programs

The GUT looks like it's doing fine, I'm now testing @Sundar's Damacy

Beta Decay

How's X marks the spot?

Do you want me to convert it to JS

Redwolf Programs

Already did. I haven't tested it yet, though.

But you might want to fix the me[3] type errors on that one, too

Damacy seems to love south

Night2

Some bots even have debug codes to output to console

Beta Decay

14:54

@RedwolfPrograms It's in Python though

Don't you have the translation?

Redwolf Programs

Oh, yeah. NvM

Sorry

Beta Decay

It's fine

Night2

Your arrayIndex function seems to be not working

arrayIndex([1,1], [1,1]) returns -1

Actually it seems it always returns -1

Ramillies

That should be OK, no? Try arrayIndex([1, 1], [[1, 1]]). That should give 0.

(from a quick skim of the code yesterday)

Night2

But he is using it like this: arrayIndex(coinLocations, b.loc)

And b.loc = [x,x]

It is not [[x,x]]

Anyways, arrayIndex([1, 1], [[1, 1]]) still returns -1

Ramillies

15:06

Well, then it should be arrayIndex([[1, 1]], [1, 1]).

I just tried in TIO and it gives correctly 0.

It tries to find an array in an array of arrays.

Redwolf Programs

And -1 if does not exist

Night2

Alright, this one works

But why isn't it giving any bots on top of a coin the points?

Redwolf Programs

Damacy is fine

You all have such different styles of formatting...it's cool, because you can almost tell whose bot is whose by just looking at the code

Ramillies

I just wonder whether Damacy doesn't try to eat more powerful bots too.

Redwolf Programs

Hmm, I guess we'll have to see

Ramillies

15:11

I did a couple of 1vs1's with it and Damacy always gets eaten by me. (In one case, we have eaten each other.)

Redwolf Programs

Huh

Well, that's an actual issue (Not a bug)

Ramillies

But now I tried again and it has eaten me!

I wonder how that happened.

Redwolf Programs

LOL

How many ms are your rounds?

Night2

@Ramillies, how are you testing it when the controller is not working?

Ramillies

That shouldn't have been ever possible :D.

2e4 steps.

Night2

15:13

Do I have the wrong controller?

Redwolf Programs

I updated the Github

Ramillies

Well, I just grabbed one from the link in the OP.

Night2

gist.github.com/RedwolfPrograms/…

Ramillies

So if that one is broken, then I'm doing the testing on a broken one too.

Night2

Execution.js

Redwolf Programs

15:14

That one is broken

Controller.js isn't

Ramillies

I use the Controller.js and fill in the bots as I want.

Redwolf Programs

I'm debugging Execution now

Ramillies

(I don't need free for all anyway.)

Night2

OK, I saw the bot codes in Execution.js so I though that is the complete one

I'm going to test with Controller.js then

My complaints were all about the wrong one then! Sorry!

Ramillies

That's the intent I guess. I'm just copying the bots from the thread as needed.

Redwolf Programs

15:15

@Night2 It's fine

I've been getting mad at people for having lots of bugs

then relaizing that I'm

using the wrong controller!

Night2

Well this one still has the issue with botNotes: TypeError: bot is undefined line:1041

Redwolf Programs

Execution.js still has the error, yes

Ramillies

I'm not getting that one...

Redwolf Programs

Controller.js is not 1,000 lines long anyway

Execution.js is, though

KatamariWithValues gets my approval for actually avoiding an enemy, while not having any bugs

Night2

OK, now that I got a working controller, I can confirm that my bot is the best ^_^

After a few hundreds of runs, alive ones:

0: Object { name: "Goldilocks", coins: 20, alive: true, … }
1: Object { name: "Proton", coins: 12, alive: true, … }
2: Object { name: "TBTPTGCBCBA", coins: 47, alive: true, … }
3: Object { name: "Wild Goose Chase", coins: 0, alive: true, … }

Redwolf Programs

15:20

Which one is yours, again?

Night2

The Bot That Plays The Game Cautiously But Can Be Aggressive

Redwolf Programs

Oh, I liked that one

Haven't tested it yet

Night2

Well it was the first bot with a score/value/weight or whatever you call it, decision system :D

Redwolf Programs

Yeah

Uh-oh, Proton forgot to add Math. in front of sin, cos, etc.

Ramillies

Yeah. I actually didn't notice that you were using the same system until much later than I posted mine...

Night2

15:24

@RedwolfPrograms Yeah, and soon you will find someone else who forgot to puth Math. before abs :D

Redwolf Programs

Also, I shortned the name to TBTPTGCBCBA

:D

Night2

@RedwolfPrograms I see, that is OK the purpose of the long name was to be unique, TBTPTGCBCBA is still unique!

Ramillies

@RedwolfPrograms; your controller shouldn't be passing the dead bots to all the living ones.

Redwolf Programs

It doesn't

Ramillies

That's weird.

Redwolf Programs

15:25

It .filters them

Ramillies

In my testing, it does.

And Potentially Victorious gets always stuck in the NW corner because of that.

Redwolf Programs

botPos stores dead bots at [-2,-2]

Really?

Ramillies

Yes, really.

Redwolf Programs

botPos.filter(function(a){return a!=[-2,-2]})

Ramillies

I see that. I'm baffled once again.

I'll have to check what code am I running.

Redwolf Programs

15:27

You're right, I checked and it doesn't filter anything!

Ramillies

It does.

Redwolf Programs

OH!

Ramillies

But it filters the bots to which you're sending the messages.

Redwolf Programs

It's because no two arrays are ever equal

They are objects acoording to JS

Ramillies

Aaand that doesn't help as well.

(but I didn't mean this)

Redwolf Programs

15:28

I'll fix it

I already have an arrayEqual() function

Ramillies

Hm, actually the bug I found is nonexistent, I can't read.

Redwolf Programs

I just need to Ctrl + C and Ctrl + V

Ramillies

So if you correct that filter, it should be OK.

Redwolf Programs

Yes

Added it to Controller.js

Night2

My prediction about the non-dying bots seems to be correct

After thousands of runs, some bots keep avoiding stronger bots

And the game doesn't seem to get any end by bot terminations

Redwolf Programs

15:33

Hmm, how long does 20k runs take?

Ramillies

In 1:1, on TIO, probably a second.

Night2

Well it depends on your hardware and software :D

Ramillies

Only .243 sec.

Redwolf Programs

So 50 bots is probably around 25 times that

Night2

But I was doing it in steps of 100 rounds manually and then checking bot statuses

Ramillies

15:34

By the way, game won't end by itself.

Redwolf Programs

If it is run 20k times, then the status checked, then 80k times, then tiebreaker, we should be good

N. P.

Is the git updated?

Ramillies

In the 1:1's I'm running, usually Potentially Victorious eats the other bot around the round 2k and runs around amassing tons of coins ever after :—).

Redwolf Programs

Only Controller.js

Ramillies

(the only question is why does the bot ever eat the other bot when it shouldn't be doing so)

Redwolf Programs

15:35

???

Ramillies

(or it should, but I didn't think it would actually manage to eat anything)

Night2

21 bots, 20,000 rounds on Google Chrome took less than 1 second

But I have removed debug codes that output to console from the bots

Redwolf Programs

That's insane. It won;t take 5 mins, it'll take 5 secs!

Night2

You should put a rounds limits on each game instead of time limit

Redwolf Programs

Yes. 100k.

Ramillies

15:37

Hooray! Finally!

Redwolf Programs

10,000 rounds, 20 at a time, that's [worst case scenario] 1 hour

Night2

After 20K, TBTPTGCBCBA with 59 coins is playing while "Polite Near-Sighted Drunk Bot" with 0 coins is running :D

Redwolf Programs

This'll be fun

N. P.

lol, I think my bot might be broken then

Redwolf Programs

Maybe I should move the due date closer?

N. P.

15:39

Polite Near-Sighted Drunk Bot shouldn't have 0

At least it refuses to die :P

Ramillies

(Hm, after 20K, TBTPTGCBCBA with 1257 has been long dead while Potentially Victorious with 26164 runs in circles.)

N. P.

Is it telling you to bugger off?

Redwolf Programs

Console.log is off

N. P.

Aw

Night2

My Google Chrome with a 100K run is broken

Not sure what is the issue

It is eating 1 core of my CPU and is not finishing at all!

My an infinite loop is happening when only 1 bot stays alive, hmm ....

Redwolf Programs

15:41

BUGGER OFF

lol

Night2

@Ramillies Is that a prediction or did you actually got this? :D

N. P.

Polite NS Drunk only has so much patience

Chase it too long and you'll upset it :P

Ramillies

I got that.

However, I get some weird things too.

Night2

Yes, something look weird

Ramillies

I'm afraid that if a coin appears at a spot where a bot had died, then the dead bot collects that coin!

Night2

15:43

Yes

Safety Coin has been dead very soon

But on my run I have this:

{name: "Safety Coin", run: ƒ, coins: 5929584, ...}

I think I should wait for the final controller

Ramillies

In my last run, the TBTawjnkwawhatever took me out very soon (I have no idea how) and Potentially Victorious amassed 700 coins while dead!

Redwolf Programs

Oops!

Dead bots collecting coins?

Sounds like...oh no.

Ramillies

Yeah. I think the problem is that you store the location in 2 different places.

botData[...].loc and botLocations.

Redwolf Programs

No, it's the .filter array problem

where 2 arrays can never be equal

No, it's because I put the while loop in the loop of every bot

Not the other loop of living bots

Ramillies

I see.

N. P.

15:46

We okay with 8x8 pixels for every square?

Redwolf Programs

Maybe 10x10

Ramillies

Certainly.

Redwolf Programs

If it's going on youTube, it needs to be visible

N. P.

I was hoping to keep it standard so people can use sprites

Ramillies

:—).

N. P.

15:47

8x8 or 16x16 seems to be common for sprite sizes

Redwolf Programs

This will obviously not run at 20k/second, so I will not use this graphic output for the competition

Ramillies

Actually, the whole thing will be maybe 100×100 at the very most, so 8×8 should be OK.

That's obvious.

N. P.

Yeah, the visual is just for observing behavior "in the wild"

Ramillies

But I would love to see a graphic version soon. Now there are some cases when Potentially Victorious gets eaten, and I have no idea why.

N. P.

Actual simulations should <i>never</i> use graphics when possible

I'm working on it

Should get it done before lunch

Ramillies

15:48

Great!

Redwolf Programs

Added b.alive part of the while loop, so dead bots don't collect coins

Uodated the github

Almost 3k views! Thanks so much!

Ben

Hello!

Redwolf Programs

Perhaps 10fps? That would make the games action filled, but not too fast

Hello!

Ramillies

The most action happens in the first 1k rounds anyway, so 10 fps should be OK.

Redwolf Programs

So around 10 minutes should give a very nice look at how the game works

Or, maybe, speed up as the game goes on

N. P.

15:53

I'm planning to tweak the FPS until it looks "good"

Redwolf Programs

Hmm

N. P.

where good is arbitrarily determined by me, lol

Redwolf Programs

Maybe double the speed when you hit 1k moves

N. P.

You're free to adjust it once I publish

Beta Decay

So how long d'you reckon it'll be before we get a proper run?

Redwolf Programs

15:54

We can already get a proper run, I believe

Beta Decay

ooh neat

Ben

I'd love to see that

Night2

There are some things very dynamic, my previous 20k run was done less than 1 second, but my 1k run now is 10 seconds!

Ben

Because I don't know what this map looks like

Redwolf Programs

But not all the bots are ready

The arena is a big square

Ben

15:56

Made of ASCII or pixels?

Redwolf Programs

Pixels

Ben

Ok cool

I'm learning so please excuse my amateur-ness

Redwolf Programs

Have you thought about building a bot?

It's easy

N. P.

I'm also learning as I go, don't worry about it!

Ramillies

Proper runs are being done even now. // @Ben: definitely go ahead and build a bot, that's the best way to learn!

N. P.

15:57

I think Redwolf is too :P

Ben

No. This is like a top tier golf in my opinion

Redwolf Programs

First KoTH

N. P.

Hahahaha, if this is top tier, we can definitely raise the bar then!

Ramillies

No way. There are a lot of simple bots out there.

Redwolf Programs

Except Polite Near-SIghted Drunk Bot, who drinks at the bar instead of raising it

N. P.

15:58

Hey Red, what's the structure of coins again?

[x, y, size]?

Ben

Yeah. I'm just like super-way beginner because I don't know how I'd start my bot. I've been glancing at the other bots to try and rap my head around them

N. P.

[[x, y, size], [x, y, size]]?

Ramillies

[ position of gold, position of silver, ... position of silver]

Redwolf Programs

[[x,y],[x,y]]

Ramillies

where position = [x, y].

N. P.

15:58

OH, that's why my bot is broken

I thought it was [x, y, size of coin] for each sub-array

Redwolf Programs

Found the problem?

Oh, that would explain it!

Ramillies

Perhaps that will be the reason :—).

N. P.

It's also why my display is broken, loo

Redwolf Programs

Is there any reason to use let rather than var?

Ramillies

Yes.

Transcript for