er, up to sixteen players, not all players

not that there are seventeen answers yet ...

This is true. I thought I'd leave the finer details to the spec wording though...

I just pressed F5 to refresh the page I was reading in Firefox, not realising that focus was on the leaderboard tournament running in Chromium. I therefore lost the leaderboard and have to start running a new one. It just so happens that a new answer had just arrived (welcome @fireflame241!) so it would need to be restarted anyway, but I've added "confirmation before refresh" to my list of things to improve...

@trichoplax Do you think it is necessary to show the number of games played by each entry in tournaments once we have over 16 players in total?

hey...

what if ants used their amount of food as a determinant for changing strategy

you said that the trail makers are weaker towards the end, but if they abandon that strategy, they can negate that problem

again, I would participate, but js

darn js

I hate it

wild fire pwns ziggurat

probably shouldn't use red

idea:

combine the two ant carriage and trails

that way, you will be less vulnreable to trail eraser hopefully

also you will know which direction you are going

Hey, know your documentation :-) developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/…

another idea

sideways centipede

use 3 different types of workers

and worker types can't be broken by trail eraser

the main issue then would be encountering enemy ants

as in not being able to move around them

also ants in the middle getting food might be hard

@FrenzyLi That's a good point. I was hoping that the confidence column would be enough, so that a player who has had fewer games will have a confidence that reflects needing to run the tournament for longer, but maybe number of games they were in would be a good double check

@DestructibleLemon You may be interested to read the new answer, Pierce, by fireflame241. It has a 3 ant line that sounds similar

nah, my idea was more getting a massive marching line of ants

like, the three different types is so you can tell which way is to the queen

wait, i have another idea that will make it so ants don't get super messed up order

you get the queen to fetch the food

by communicating along the line

ftl communication

the only issue is some of the ants will start moving

but it can be mitigated

another idea

when you want to separate ants and then bring them back together

make the ants trail erase'

they will gravitate towards large patterns

might not work

kind of weird trail eraser works best when not trail erasing

@DestructibleLemon That's about what I am trying to do, but communicating along the line to deal with collision is hard™

Image dump incoming because postimg sux.

Looking forward to it :)

@Draco18s Are these the Wildfire frozen workers? I couldn't see these until uploaded in here

Yep

I've also ocassionally seen Wildfire laden workers in a white region completely enclosed in red, just running around and around the perimeter

That's unavoidable. Happens to my workers too

And the now-misleading scorecard :p I think that's all of the postimg images

Yerg, my bot has become so complex, with so many bizarre edge cases...

It seems that's what it takes to beat a lone queen strategy...

I increased the number of moves from 10,000 to 30,000 during testing in order to make lone queen strategies less competitive, but it now seems that the long running makes worker strategies lose effectiveness nearer the end game. Maybe we'll see strategies that use workers for the first half of the game and then switch to lone queen foraging nearer the end.

Its not that I lose effectiveness during the endgame, its that I have code for "what happens if I encounter a diagonal line of black/blue/yellow?" "what happens if two workers meet each other on a lone white space?" "what happens if a worker meets an enemy worker?" "what happens if a queen with 12 food acquires a 13th piece?"

running into someone else's blue trails still messes with my queen, and its super hard to find a fixed seed to test behaviors against

Well, you may want to just single out your entry and another blue trail entry and let them play for a fixed seed?

There's the risk of another player reaching your area of the arena and disrupting all your hard work though (like the blue trails you mention, or Trail-eraser), so I was wondering about a complex strategy that falls back on speeding off on a trail-less alternative strategy if disaster strikes

Like an insurance policy

There isn't an existent "blue trail" entry :P Artistic Ants is the closest, but it does Random Things

I'm fairly sure that the Artist is deterministic...

I actually tried to accuse it of an obfuscated Math.random() call, when it was due to very minor tweaks of my own code that was leading to a different number of ants on the field, which corrupted the random seed in the controller which lead to wildly different behavior from Artistic Ants

It's the view that gets rotated by the controller. As long as the rotation of the view is the same for the same seed, it should be fine.

Otherwise, it would be a bug in the controller.

If it doesn't explicitly call Math.random then I'd be surprised if it accidentally managed non-deterministic behaviour. If it turns out it does though, it'll be disqualified until it can be fixed

It's deterministic

The seed in the Artist is simply a bitwise OR of the color of the surrounding 8 cells shifted by its index.

But if my bot produces a 14th ant 3 turns sooner, then Artistic's result varies by a wide margin.

THAT's what drove me nuts

I'm not CS savvy so I came up with that seed... Turns out it's quite okay at controlling the spawn rate of workers.

I may want to reduce the rate a bit, too.

It's super clever, Frenzy, I don't fully understand it. It was just making me pull my hair out when I would adjust the "how to handle seeing blue" code that would cause Artist Ant's blue trails to stop hitting my blob on turn 14000 and instead hit at 8000

Or 22000

or never

I'm sure other entries are using the surrounding information to make a seed variable. I just opted to use only the color information.

And it was all because the "saw blue" code would make a worker pause an extra frame here or there, which would mean it would see food a few turns sooner (or later) which meant that my queen produced a new worker at a different time, which would cause more or fewer calls to Math.random() from the controller, which meant that Artistic would hash its surroundings slightly differently due to the altered rotation.

And, if the seed happens to be equal to 666 mod 2333, and if the food is exactly 3k + 2, it'll spawn a worker.

(And if other entries do it, I haven't messed with them, I just had updated Black Hole and Artistic running in a sandbox by themselves)

It's deterministic. In the same sense, it's quite pseudorandom.

Yes, I know

I'm saying that the appearance of random was due to side effects in my own ant

Through a convoluted chain of events

I see...

I was hoping to see players that took advantage of the random orientation, and there are a few now that do to different extents. Your Antdom Walking Artist, Lone Wolf, Medusa, and the new Claustrophobic Queen

You're right on that one.

Ok, time to dig into my "what to do with blue" code again. I think the "how often to produce more workers after switching to hoard mode" code is as good as its going to get.

I'm still very happy with how I handled neutralizing Trail Eraser ^..^

@Draco18s I really admire your effort :-)

And ingenuity, too.

"That's an enemy worker." <Change cell under it to cyan>

If you want to include your latest code in an edit I can include it in the next leaderboard tournament to see if the disqualification problem is fixed in a long term run

No need to wait until you've finished before seeing how it does...

Give me two minutes

I'm likely to be posting the latest leaderboard in the next hour or so (there are just two players still in joint place)

Then I'll start a new tournament going

ok

To be frank, the claustrophobic queen by @DLosc is what I initially had in mind. Somehow I immersed myself in implementing the painting routine.

HEh

It's fun :p

One of my test strategies was much the same apart from the extra colours to escape being boxed in. The black escape route seems to make a big difference. I'm running another tournament in parallel with Claustrophobic Queen included, and it's in the top 6

@trichoplax post updated

Great. The current leaderboard looks not far from ready - up to 90% confidence

i.stack.imgur.com/prUMk.png this is a trail destroyers job right?

@dzaima yep

Although it's gotten weirdly conservative lately. I used to have incursions where the queen would hit the black hole and dump upwards of 12 workers. Now it switches to food hoard mode at 6 food (very early!) and be harmless

@Draco18s That may be because I suggested that a player that purely sabotages without also trying to win might be disqualified

There's nothing wrong with that

I just think "6 food" was a very very conservative cutoff

It seems to make it a split strategy - some games it doesn't reach the cutoff and just sabotages all game, others it reaches the cutoff and hardly causes any damage. It means it ends up with either high scores or zero, and ends up roughly in the middle of the leaderboard

Sure

I haven't found much time to work on my trailless bot recently. But I think I can mitigate the fact that the bots will have limited range (108x108 squares) with some kind of outpost strategy.

Unrelated, @trichoplax Have you seen some of the other adjustments I've made to the controller? I gave food in the zoom view a white outline so it can be more readily seen on dark tiles (cough, black)

Both types of workers are frozen there and refuse to move

@PhiNotPi If you post there will then be over 16 players, so we'll see how the controller copes with not including everyone in each game. I'm interested to see how it affects the speed of convergence.

@Draco18s I've been meaning to have a look. I have a long list of changes to make, and some of your suggestions may reach the top soon, at which point I'll need to decide whether to include your solutions, if you'd be happy with that?

@FrenzyLi Intentional, if you're running my 7/26 update. :D

@trichoplax Any tweaks I've made are fair game. :)

@FrenzyLi I take it only the non-Black Hole workers are frozen?

@Draco18s Thanks!

No, it's the anti-sabotage code: my workers are constantly replacing the cell color underneath the other workers with cyan. Those workers then update and go "this cell is not my color" and replace it again.

@Draco18s Oh I see. So you sacrifice one of your workers to freeze one of theirs?

Yup!

@trichoplax The auto-pause feature doesn't...completely work on Chrome (if the display is hidden it own't show and pause (works on firefox though). Still, very useful for testing.

@Draco18s It seems like there are about 5 or more black hole workers stuck there, and also 5 or more wildfire ones.

About equal.

Interesting choice

If my workers can't see each other (e.g. BXB) then they both think they're the only one on lockdown duty

If two BH workers can see each other they attempt to get one of them out of lockdown duty

Seems like you have also updated the coloring strategy

preferring to keep an orthogonal worker

@Draco18s So combining those two facts, a third worker arriving can release one of the two?

You mean the swirling effect?

@trichoplax yep, although it takes time to reach equilibrium again. It's an imperfect system, due to the nature of communication.

I don't know if it's swirling, but your queen ant seems to erase a few blue colors if it's possible that someone was tampering with it.

Claustrophobic ant seems to reflect upon seeing a vertical or straight line. If it's confined in a box, I wonder what will happen.

@FrenzyLi I guess at some point a sandbox feature would be nice, so you could draw on the arena and then release ants into it to see how they react

I may tweak the food threshold on my trail-eraser once I see the leaderboard come out

in my games, it was rare to find so much food and so few trails at all

I'll also make my trail-eraser code more resistant to tricks like draco18s's fill-with-cyan by decoloring more randomly if there is more than one adjacent color

which also means it can't get lost by wider trails

@FrenzyLi Yep, my ants will (attempt to) erase blue that looks like it was left by another ant.

I never saw that phenomonon earlier of ever hitting twelve food, so I didn't think it could happen

The leaderboard is taking a while to stabilise - there are just two players with a joint place now. Trail-eraser is going to be in either 7th or 8th place (probably 7th), and once that's settled the leaderboard will be posted

ok. good enough to beat like six other ants, while still unleashing it sabotage ...

who it trail-eraser tied with?

Romanesco Road, although Trail-eraser is higher scoring. It's just not quite higher scoring enough to be statistically significant yet. It's currently at 85% confident, and they will separate into 7th and 8th place when that reaches 95% confident

Since it's only two players, I may go ahead and post with a joint place if it doesn't resolve soon, as there are new players to include and a new tournament to be run

It's 101 games (12 hours) in so far, so it might not be realistic to avoid joint places completely

ok

and I have some tweaks to the code to make, too

@FrenzyLi Here's what I meant by Artistic behaving erratically. Same seed, same move count, the only thing that changed was an alteration to the blue-cell detection code of my own ants.

The first image is the test-case. The second image is after implementation of the changes.

... which will only get worse when more players come

@Draco18s It's a meta-strategy. @FrenzyLi has built a chaos engine to thwart your attempts at debugging...

Haha

still, this shows that something isn't using the seed that should

One thing I wondered about when designing the game was whether people would just end up copying each other's strategies and all improving in parallel. I see I had nothing to worry about. It's been great to see all the different approaches diverging and driving each other's development by competition rather than converging on one optimal strategy

@ppperry No this is correct behaviour. The seed can only give the same behaviour for the same code. A tweak to one player's code will change the behaviour of all players

oh

With the Black Hole code one way, the outcome for that seed is always as in the first image. With the Black Hole code tweaked, the outcome for that seed is always as in the second image.

I could reduce the effect by using the seed separately for each player (16 different instances of the pseudorandom number generator), but the behaviour would still diverge once they start affecting each other, due to the change in Black Hole's behaviour

@trichoplax It'd still be useful, maybe add that as an option

I guess if I add it I'll just change it rather than making it an option, as I can't imagine demand for the less useful version...

Added to my list...

The problem with the linked images @trichoplax is that prior to move 8500 or so, there was no interaction. Artistic ants never got close to the black hole until ~8550 (and my ants never went over to it), yet changes to my code affected what Artistic did

Oh, misread. You were suggesting a change to making the seed per-bot :D

@Draco18s yes - that would fix the problem you observe with deviations in the artists before you ever interact with them. Nothing can stop deviation after they meet, but I thought not deviating until they meet would make debugging a lot less stressful...

nod nod

It shouldn't be too difficult to implement, as the seeded random is already designed to be instantiated, and nothing else uses random apart from the input orientations once moves start (the food shuffling can be a separate instantiation as a one off before the game starts).

Mhm

semi-related, I still haven't figured out how 3 ants can change the cell color 3 times, yet from move to move, the color doesn't change. e.g. 2 black hole ants and 1 trail eraser.

if the cell is black, trail eraser sets it to white. then BOTH black hole ants observe the cell being white and change it to cyan

trail-eraser must be after black hole in the move-ordering

what happens is that both black hole workers do their thing, then trail eraser makes the cell white

or maybe one black hole ant chnages it to cyan, then the trail eraser sets it to white, then the other black hole sets it to cyan, then then the next global move, shouldn't it be cyan?

(end of move, controlle shows white)

next move, both black holers cyan it, and then trail-eraser whitens it

What I mean, @ppperry is that if I use "step 1 move" the game view always shows the cell as the same color, yet there are 3 changes

rinse and repeat

But I always observe between moves that the cell is white

step one move means that all ants get one turn

yes

so trail-eraser always goes last by chance, and thus wins with the coloration

also, why are you wasting two workers on one trail-eraser

but 3 changes. white -> cyan -> third ant moves away. Or cyan -> white -> cyan, view displays cyan. or cyan -> second ant moves away -> white

none of those happen

I don't waste 2 workers on one trail eraser. not intentionally

that's the problem

this whole thing will be moot soon enough -- I'll change trail-eraser to be less predicatble in large colored areas

some how BOTH ants see the trail eraser's cell as being white, so both go into lockdown mode. yet, if there are two ants coloring the cell cyan, it doesn't matter what order they execute in: One of the two black hole ants should see the cell as already having been changed at some point

also, if you find a seed where you can reliably test trail eraser, let me know :)

I don't know then

I'm not doing seeded games

curses

time for me to add a seed display to the controller!

yah!

except I'm not using your controller version either

@trichoplax

You...you heathen! jk

no, you're the bad one for trying to defeat my trail-eraser

even if I were to find such a seed, why should I tell you?

all that would help you do would be make your black hole stronger against my tricks!

... end of rant

/me figured you would decline anyway :p

also poop, randomseeds doesn't use the same random system as the seeded random. There's no way to convert

<sad dragon>

IT HAS BEEN DONE

I AM IMMUNE TO TRAILS

Only took me like three days

@trichoplax You started a new leaderboard yet? o..o

THE TRAIL-ERASER WILL BE BACK!

I'm sure it will

I look forward to neutralizing it again

@Draco18s Try "Step ant" (I don't think the keyboard shortcut is working at present, but you can lock the zoom by clicking on the arena, then you can click Step ant to move one ant at a time, rather than the whole population once).

I've done that too, but still never figured out why I was seeing what I was seeing

... ... ...

But you did see exactly what happened for each individual ant?

gg, wildfire it trapped my repairer with 5 workers and blocked the rail for regular food finders with 10..

@Draco18s Note that all Black Hole ants move before any other player's ants (or after, but it's cyclical so it doesn't matter which way it's phrased)

So then if BH1 moves and changes a cell color, BH2 should see that color.

BH1 / BH2 being ants #1 and #2 of the Black Hole faction

The turn order is: Player A's workers in creation order, Player A's queen, Player B's workers in creation order, Player B's queen, ...

@Draco18s This will apply if BH1 was created first

(which should be true as they always move in the same order)

I figured they were in creation order (queen last), what I meant was: if I have only 2 ants (plus queen, ignore her), the first ant goes and changes the cell color. Shouldn't the second ant see that new cell color?

Yes

As far as the controller is concerned, each move only involves moving one ant. The presentation to the viewer of the entire population moving each turn is not how the controller works internally

This is what makes the faster than light communication I mentioned at some point possible

That's what I thought.

which makes the "two black hole, one trail eraser" puzzle confusing

I'm curious about this now though. Unless it's a controller bug I can't think how it could happen

But I can't imagine how there could be a controller bug with the turn ordering that doesn't show up in more obvious ways elsewhere

shrug

If you paste your code into the new challenger area and tick debugging, then pause it when this happens, you should be able to see the exact inputs and responses for the Black Hole ants (although not for the Trail-erasers).

Does cyan under the enemy worker definitely trigger a Black Hole worker to move directly away?

(in general)

Black Hole ants treat enemy workers as "my color" for the purposes of trying to move

OK. So you'd expect BH2 to move around the enemy worker until it finds more of its own colour and continues building?

I would expect BH2 to move away, yeah.

Also. This is weird.

...-->

I put console.log("========") at the top of my function.

I clicked "step ant" once

I got this:

That was six calls to my function

not to mention trail-eraser now erases all but one of the colors around it randomly before moving, so it will no longer get fully stopped by placing colors under it

@Draco18s Did you have fewer than 6 ants? Also, does your console always display things in order? Sometimes I've seen messages that are identical combined into a single entry with a number after it

I uncollapsed the messages. And started with an empty console.

And...I think I have 12 ants

May be 10 at that point, I have 12 now

So six calls is too few

Uh.

Isn't "Step Ant (A)" supposed to move one ant

not an entire team?

Just now I cleared the log and clicked "step ant" again and got three calls

(none of which were the queen)

Then seven calls

Then 3 calls: the queen and two non-queens in that order

Oh "Step ant" - sorry I misread that. No it's more complicated, I should have explained: When not zoomed "Step ant" moves exactly one ant. When zoomed, "Step ant" moves exactly one zoomed ant, but will move as many other ants outside the zoom as necessary in order to achieve that. This is so you're not stuck pressing "Step ant" a hundred times just to see the 2 ants in the zoom move once

oh

that's super extra confusing

When viewing the zoomed area, it should hopefully be what you expect, but when viewing the console it will look weird...

@Draco18s I agree, but it's very useful

@trichoplax That's actually pretty convenient.

Definitely, just not what I expected :D

I.e. desired and intended, but undocumentented behavior

I'd like to make it less surprising, but I'm not sure how. If it only ever causes confusion when viewing the console, then perhaps an extra message there, which anyone else won't have to think about?

could there be an option to step trough the single ants of a team?

@trichoplax add a hover-text that says that

On the Step ant button?

@trichoplax yeah

"step zoom (Z)" vs. "step ant (A)"

I'm wary of depending on hover text though, as not all systems will show it. I've tried to make the controller usable on as wide a range of machines as reasonably practical

@Draco18s That would be unambiguous, but take more thought to use. I'm torn

Mmm

Actually it would need to be something like "Step zoomed ant", otherwise someone might expect all the zoomed ants to move at once

what browsers don't support hover text?

mobile

I can't see anyone being confused if they're not viewing the console, so I think just an extra line in the console should do it. I've avoided logging to keep speed up, but it will only happen when stepping so it will make no difference

@dzaima yes touchscreen devices are what I was thinking of. Not sure how likely mobile is, but tablets and even touchscreen laptops/desktops are fairly widespread now

Just wait until we have exovision displays.

cough, I mean mixed reality

Looks like the leaderboard should be stabilised in the next game or two

I'll want to update Black Hole one more time before you start on the next one

I'll just start the next one running and then restart it once you have an update - it may as well be running just in case

ok

There are other player updates too, so it helps with screening out any that will be disqualified to save time later

sure

I'm running two in parallel sometimes too, with and without a returning disqualified player, so whether they get disqualified again or not there will be correct results available

/me nod

-- and trail-eraser keeps chaning

it can no longer be stopped by coloring under it

and now waits until nine food before entering hoard mode

@PhiNotPi in one of my games, medusa was hit early by a trail-eraser and now gets into a state when the queen's square and all eight neighbors are white.

When that happens, none of the colors change

why doesn't the guard randomly pick a color and start rebuilding with it?

I think medusa might actually infinite loop in that state

(or at least some ant is, because my game is frozen(

@ppperry Sounds like I need to move that up my priority list if freezing is actually happening now

@ppperry Medusa's long-term support period ended yesterday. Any newly discovered security vulnerabilities will most likely not get patched.

oh

maybe it's not medusa after all, but something is freezing

(I'll look into it when I get a chance, after my next bot is done)

New leaderboard is up

it doesn't seem to be an infinite loop, just a crash actaully

still

unfortunately, I wan't using a seed when that happened

Ah OK. It would still be useful to make the controller freeze proof at some point

(the bad behavior of medusa when beheaded still happens, though

Does it freeze the controller?

it seemed to at the time

but I got it started again by typing "processAnts()"

in the console

That makes it sound like it wasn't frozen, just cancelled the timeout that is meant to call processAnts each turn

In which case everything would be responsive, just no gameplay, as if paused

oh

I didn't click pause, though

never mind, forget all of this. I must have lcicked on pause and forgot I did so

oops

Unless there's a bug that has the same effect. Pause just cancels the timeout, so if anything else did that it would look like it was paused, just it wouldn't grey out the pause button.

I think pause was greyed out, though

D'oh!

At least that's one less thing for me to investigate then :)

yah

trail-eraser can sometimes still meet its switch threshold of 9 food

it does so on seed 10

black hole is stull vulnerable to premature switches

no, it seems to draw wierd patterns of black and blue