« first day (7 days earlier)    last day (15 days later) » 
00:00 - 17:0017:00 - 19:00

5:00 PM
I'm going to run 4 20k round games simultaneously
times up
@RedwolfPrograms Good luck :P
Deadline over; games will be run once one timing game finishes
Tried out a few different tweaks just now but couldn't get the Turtle to first against this field. Maybe it'll get lucky in the Monte Carlo though :) Thanks for a fun game!
83 seconds for 1000 rounds; shouldn't take too long
@histocrat You're welcome!
5:02 PM
@histocrat Still performs very nice, I did 2 1k tests and it got 2nd place constantly.
bleh, didn't get my changes ready by the deadline. was trying to do something about being nearly perma-stunned by blue shell
@Draco18s Doesn't blue shell die pretty quickly?
it lasts to the end usually
I guess because none of the bots care about a no-worth bot
idk how it doesn't get murdered by serialkiller before that
5:04 PM
Neither do I
I guess because nothing puts it at a low enough health to be a target
low enough to be a target
but it sits at around 100 hp for most of the game
it gets down to about 40, but has a worth of 0 so no one kills it
SerialKiller would
It doesn't care about worth
serial killer probably died already
5:05 PM
Though doesn't serial killer die pretty fast?
i made a change so it should shield more effectively
I'm almost willing to complain that Blue Shell is non-competitive ( "EmoWolf" strategy of "gold? why should I collect gold?")
Usually it's not one that kills my Barbarian Man bot, so it must be gone by then
@Draco18s I had the exact same similar issue and I also had issue with the Teacher bot, had to stop upgrading my attacks and slow down a little to fix that issue.
@RedwolfPrograms it shields near the start, making it an unlikely target as there are many others that wont
5:07 PM
I think serial killer gets stun locked by anti war(my own bot...) b/c it has high attack
KS shields at turn 1 because of that
@Night2 yeah, but Blue Shell goes after worth which is harder to control well
Stun was badly implemented in my opinion; it should have either costed HP or gold
@RedwolfPrograms yeah, i thought about that too
5:08 PM
@Draco18s If you go low profile, the chance of it being dead before targets you grows too.
@Draco18s I still get targeted by it, but less.
Yeah, it just wasn't a fix I could implement quickly
I was planning on making another KoTH similar to this one in the next few weeks, but I think I've changed my mind
There's a lot of cool changes that could be made, but I'd rather make another KoTH than redo one I liked
@Draco18s I literally changed like 5 lines in my bot to go from 6th place to 2nd or 1st by just having a lower profile to not get targeted.
I think @pixma140 wanted another similar challenge, but I've got some ideas for auction-based challenges
5:11 PM
blue shell stunlocks you until either you die or it dies
@Night2 worth is harder to control than attack level
having a low attack level would help avoid getting stun locked by anti war
attack is easy, you just don't upgrade. worth relies on a lot of factors, a single attack on a low-health bot occurs in like 4 places in Kill Assist, and whether or not it jumps KA's worth to the top depends on who else made an attack or not
@SuperStormer just having 1 less than everyone else should be enough (though the others might die)
Maybe that's why Barbarian Man didn't do as well with level 15 attack...
It's now limited to 5
5:13 PM
@Draco18s a higher attack makes it more certain whether you'll be able to kill or should you stun though
@dzaima I arrived at the conclusion that upgrading attack with the existing bots is useless and deleted it all from my bot.
@dzaima Lots of bots attack weak bots, so even with a +5 attack you still get the kill most of the times.
A bot with an SHP less than 40 usually dies immediately
@Night2 but if the bot doesn't die as predicted (guesses about who'll attack were wrong), i've wasted a chance to stun
@dzaima Kill Assist never upgrades attack
@dzaima I also arrived at the conclusion that stun is useless, you slow yourself and a competitor, but make all other competitors stronger.
5:16 PM
made stunning take 2 hp
In teams it is a different story though.
@Night2 the idea is that I stun the competitor making it unable to get the kill, but still get the kill myself
...and its precisely that logic that pushes Kill Assist's worth up to the top
2HP would just make stun like farm, where you can stun-stun-stun-heal. I think 2 gold would be better
@dzaima But other competitors will still get the kill and you could have farmed that turn instead of stun.
5:18 PM
@Night2 you stun the guy with the most worth
you don't CARE abut anyone else
But it doesn't gain you anything
@Night2 well losing ~30 worth is worth making the top bot lose 5-10k
@Draco18s When there are 5-6 bots with very close worth, the most worth doesn't mean much
@RedwolfPrograms it gains me less competition
5:18 PM
Oh, true
The point is that after 5-6 kills, all of the other 5-6 lose one kill and I lose none
Another way to tweak stun would be to make whatever the stunned bot tried to do that turn apply to you instead. So you can't stunlock someone forever because they'll detect that they're being stunned and farm you to death, but stunning becomes a more viable strategy to actually win.
I lose out of 30*5 gold, they each lose 10000
My logic is simple, you stun X, you and X lose a turn, all other bots close to you 2 don't lose that turn.
@histocrat Interesting
5:19 PM
@histocrat that requires a huge overhaul on the controller, because stunn'd bots logic doesn't run
@Night2 the time I and X lose turns are 1 turn apart though
@dzaima You still lose 1 turn which could be used to farm with level +30
@Draco18s Not a huge overhaul, just a quick check to see if it's stunned, then the return value is routed elsewhere
@RedwolfPrograms huge as in "effect" not as in "lines of code"
@Night2 but I can also use that turn to drag the average score of some other player down -10000
5:21 PM
@dzaima Yes, but 6 other players are benefiting from that
@Night2 Would you rather have 30 gold, or prevent someone else from gaining 10,000?
@Draco18s Yeah, it would make the strategy completely different, and the controller would need to be a lot different
@Night2 if I farm those 6 others benefit anyway
@Night2 noones benefiting from it, I still get the kill
@dzaima No, you have lost a turn with at least +60 worth
@Draco18s But you farmed, so you didn't lose anything
5:22 PM
@Night2 but i got another bot to lose 10000 worth....
@dzaima Again, yes, you and X lost something, other all have benefited.
@Night2 benefited with what?
     I stun       Killing blow     total
me:   +0              +10,000    10k
bot1:   +30           <stunned>    0  <-- definitely > 30
bot2:   +30           +10,000    10k
bot3:   +30           +10,000    10k
Benefited from you and X not getting some points. X didn't get a kill and you didn't farm
@Night2 You have lost 30 gold, they have lost 10k
5:23 PM
@RedwolfPrograms So I still lost something, right? And 6 other player didn't lose that.
@Night2 but my loss is worth the loss of the other
No, but they're not as big of threats
@Night2 but averaging out, if each of those 6 is targeted equally, it's still positive average gain for me
@dzaima You don't get my point. My point is if 2 people fight in a group of 10 people, the other 8 will benefit.
@Night2 9* benefit
And that's a poor comparison
5:25 PM
@Night2 From a theoretical perspective you are correct, but practically it makes more sense to remove the second best competitor
its more like 10 people splitting a cake at a party and I decide to pop one guy's tires before we all leave for the party
@Draco18s XD
I don't have to give up any cake
@RedwolfPrograms You are only seeing second best in this game in current turn, you don't know about the overall status.
the other 8 people don't give up any cake
5:26 PM
@Night2 right, I can stun one person, but if I prevent each of the 6 to get a 5k kill, everyone else has lost 5k and I've lost 30*6, making a 4.9k difference bonus for me
the other people and I all get more (relative) cake
do this enough times (popping a different person's tires each time) and relative to everyone else I get more cake
in practice there's only ~2-3 players that are serious contenders to be stunned by me
In a game with 40 bots that 30 of them die, you might be wasting at least 30 turns in each game to stun someone.
fact is, adding that stunning logic originally did make me by a big margin
@Night2 The one stunned is the one with the highest worth
5:28 PM
...not a random bot
@Night2 most die very quickly, and there's calculation happening to decide which will attack, and whether its worth it
@dzaima Yes, because you had less competitors in that time, when I updated my bot, it was winning from all others without a single stun.
@RedwolfPrograms Highest worth only at this moment in this game.
you know what, heck this. Let me run a sim
@Night2 but on average highest worth is gonna be the winner
@dzaima In a 1v1 my bot would lose to yours
Agree? :P
5:29 PM
@Night2 yeah, but this game is not a 1v1..
@Night2 But it tends to scale; higher worth after 100 turns is likely going to be long enough for a pattern to be established
@dzaima Exactly, and there are many competitors to waste your turn for focusing on a single one.
and it won't stun with only one other player left at all anyway
If there were 10 bots and only 3-4 competitors, then that would work, but when there are more, it will be a waste, you don't farm, and stop top worth, someone else will be top worth, not you. You cannot fight with all.
But that's one less bot to have to pull ahead of
5:32 PM
@Night2 whatβ€½ I do fight, I just stun when I know someone else would go for a kill and fight after, meaning top attacker loses 1-10k and I lose 40 or so worth
@dzaima So again that is what I say, you and X lose, others don't.
No no no no nno
Only X loses
@Night2 I lose practically nothing......... others on average lose a couple thousand worth
@dzaima 30 turns each worth 60 at least in 10,000 games is a lot.
30*60 is 1800
killing a bot late game is worth 20,000
5:34 PM
Less than one big kill
When teaming was allowed, my second bot was designed to target all other top worth bots that weren't my main bot. It wasn't designed to only stun top worth, it would be useless because someone else could become. I would select a random top worth which wasn't my main bot and stun it. Only 40-50 stuns per top bot was enough to make sure my main bot gets 1st place with large margin.
Even in mid-game, 30 is hardly anything
@Night2 It doesn't matter the order
@Draco18s In late game, bots rarely die, that is why scores are around 40k for most of top bots now.
jesus christ, pop open dz's controller and scroll down to the bottom and look at the last 3-5 kills
Well, it must work if KA and KS do so much better with than without
5:37 PM
@RedwolfPrograms Yes it works, but the less the other bots are, the more it works, the more the other bots are, the less it works.
@Night2 the less attacking and high-worth bots there are
@Draco18s So why are all the top bots now limited to 35-40k worth at end game? If they were to kill 2 20k each they should have had a lot more.
@Night2 KA & KS get most of their score from killing others
@Night2 because the field has become flooded with defensive bots. There aren't enough attackers to actually start a killing chain
@Draco18s So you agree that late game kills are rare now?
5:40 PM
@Night2 average score. I've seen single-round scores over 130,000
At least any kills with a lot of worth
Over 170k I've seen
anyway, I'm done here. This conversation is stupid
@Draco18s Yes but with a whole different situation.
We can agree to disagree
@Night2 fact is, it did work, and imma look into if it still works (which i bet it does)
5:42 PM
@dzaima It just works less and less as the number of bots grow
@Night2 less and less as the number of attacking and high-worth bots grow
even if the average worth of a bot upon getting killed is 1000 and there is a 10% chance of attacking the bot that I want, it's still an average of 100 worth loss for that bot vs a 40 loss for me, and that's very unlikely
here's a list of what and how many times my bot attempted to stun a bot
@dzaima In how many games?
that's a 30% chance of targeting Kill Assist, and if the average kill worth is 1000, it loses 333 on average for my loss of 40
So you used around 8k to 9k of your turns to stun others.
Which averages 8-9 turns per game...so nothing
5:51 PM
It also shows that "Kill Assist" is probably individually very strong, and you have made him weak, but that has not given you the first place, instead the first place has gone to "The Bot That Farms ...".
@Night2 right, I didn't get to perfecting the strategy for the current bot set
my stun or attack deciding is also probably pretty flawed and could be made much better
If you would have targeted "The Bot That Farms" instead, then the first place could be back to "Kill Assist" or someone else. That is my idea.
Kill Assist is based on KS and it does well, too
in total, stunning loses me 307 worth average per-game
Which is a drop in the bucket
5:54 PM
calculating the success rate of stunning would be a bit more difficult though
@dzaima Yeap
Too cool things down, here is a photo of my cat and her 2 weeks old kitten: i.imgur.com/KhFvGoy.jpg
Games complete
@RedwolfPrograms 😲
Dang. Missed the deadline. Caught up in work. GG
@RedwolfPrograms Already?!
6:01 PM
@Alion 4 20k games simultaneously
With a few modifications to the controller
Who won?
I'm adding the results right now
For those who can't wait:
1. The Bot That Farms, Attacks, Shields, And Even Heals But Never Stuns: 38803.7456
2. reallyCommittedTurtle: 36241.1837
3. Kill Assist: 35206.8535
4. Switzerland: 34830.458
5. Kill Stealer: 32517.2357
6. ScaredBot: 29148.7197
7. <-------------------The Accountant------------------->: 27790.6789
8. LettuceBot: 24808.5283
9. Unkillable: 24590.975
10. CampBot: 21529.9598
11. Friendly bot: 17025.6304
12. Barbarian Man: 12276.0505
13. Bandit: 12241.592
14. Captain Farmer: 11790.019
15. Reformed TeamBot: 11048.7122
As you can see, JustFarm won
For suspense, I'll do them one by one, backwards
or you could be reasonab;le
6:09 PM
Ok, fine
I need to email them to myself to copy/paste
Can we get a unicode drumroll?
6:11 PM
1. The Bot That Farms, Attacks, Shields, And Even Heals But Never Stuns: 39051.055775
2. reallyCommittedTurtle: 35982.39805
3. Kill Assist: 35099.4138
4. Switzerland: 34947.2974375
5. Kill Stealer: 32485.679775
6. ScaredBot: 29223.2232875
7. The Accountant: 27730.03395
8. LettuceBot: 24900.8237375
9. Unkillable: 24631.6802625
10. CampBot: 21691.1124875
11. Friendly bot: 17074.6225
12. Barbarian Man: 12453.3356
13. Bandit: 12221.1209125
14. Captain Farmer: 11930.23085
15. Reformed TeamBot: 10941.71495
@RedwolfPrograms Thanks for hosting this KoTH, was fun like the last one!
You're welcome!
Ayyy, didn't get last somehow
Now it's time for the awarding of the bounty...congratulations to Night2's bot with the ridiculously long name for winning!
I fell pretty far since yesterday
6:14 PM
I should've built a real entry but eh
Yeah, I used to have 5th place with Capt. Farmer
Living for the meme dream
in testing, i had to put <----------------------------------------------______-------------------------------------> next to my name so it was easy to spot
next to that insane name
@Veskah What was your bot, again?
@B0RDERS What insane name? :P
6:15 PM
@Night2, The Bot That Farms, Attacks, Shields, And Even Heals But Never Stuns
way to long and not at all catchy
@B0RDERS What kind of idiot would name a bot like that, I wonder!
@Veskah He died pretty quick since he was actually worth something...Persistent Blue Shell almost never died
at least I tried making good names: The equalizer, the accountant, sniperbot, smartfarm
Eh? Should've been worth 3 coins ever
6:16 PM
@B0RDERS The "and never, ever stuns" part is kinda catchy
@B0RDERS I always make a long name for my main bot, I'm sorry about it!
Which I guess is more than 0, u rite
@B0RDERS In my last one, it was The Bot That Plays The Game Cautiously But Can Be Aggressive
it would be catchy if there was a never, ever. There is just a never, though
6:18 PM
@Night2, I see
is it for comedic effect?
@B0RDERS Just a tradition!
@Night2, now that I know that, it is more respectable
@B0RDERS It's his trick to winning
He won both of my KoTHs with long-named bots
I used to do that 10 years ago when competing in university ACM competitions with very long names for my team, and from that time I always select long names!
If anyone wants the PHP for my AutoKoTH system, or ever needs it for one of their own KoTHs, feel free to ping me!
I'll post the source code here, anyway
<!DOCTYPE html>
$answers = json_decode(gzdecode(file_get_contents("http://api.stackexchange.com/2.2/questions/189110/answers?order=desc&sort=activity&site=codegolf&filter=!.mQhx_Oju_oX&pagesize=100")), true);
foreach ($answers["items"] as $item) {
preg_match("/<h1>(.*)<\/h1>|<h2>(.*)<\/h2>/", $item["body"], $headings);
preg_match("/<pre.*><code>([\s\S]*?)<\/code><\/pre>/", $item["body"], $code);
$source .= strip_tags($code[0]) . "\n\n";
6:23 PM
After reviewing all the major bots, I personally would vote for "Kill Assist", a very strong individual, and the most important part is that it was updated 2 days ago and was on top for most of the time. Everyone had 2 days of time to counter its strategy and still is better than all bots individually.
00:00 - 17:0017:00 - 19:00

« first day (7 days earlier)    last day (15 days later) »