The Nineteenth Byte

the solutions are built up by starting with just the inputs (not quite true but doesn't matter) and then for i=2;i++<inf we build all the functions of i size

@MartinBüttner if my maths is correct, yeah

overactor

some solutions with 2 inputs are equivqlent though, right?

VisualMelon

@overactor indeed

so I have binary tree which stores all the current soltuions

Martin Büttner

2:07 PM

I'm just gonna throw "Karnaugh map" in the room and leave for now.

VisualMelon

so I can quickly see if that result has an existing solution and ignore it

Nathan Merrill

@overactor I updated the post

overactor

can you build every 3 input solution out of 2 two input solutions?

@NathanMerrill, I'll check it out

VisualMelon

@overactor by definition yes

because the functions a and b only take 2 inputs and provide a single output

(sorry, I forgot to say that!)

overactor

So In stead of using functions a and b to solve a three input problem

why not use combinations of 2 possible 2 input solutions?

VisualMelon

2:10 PM

well, eventually the solver will find all 2input functions, there are only 2^(2^2) of them, afterall

overactor

Oh, I see the problem

I wasn't thinking right

google whatever @MartinBüttner said

That'll be more helpful than what I'm saying

VisualMelon

my 4-way paity checker is currently on iteration 13

@overactor it sounds like the name of some clever and generalizable technique, and I suffer from NIH syndrome badly

I'd be unhappy if I use a technique I hadn't come up with as the bulk of my solution :P

Sparr

for the domino circuit challenge, it seems like coming up with the logic circuit based on the truth table is a wholly separate (and solved elsewhere on the internet) problem from implementing the circuit with the minimal number of dominoes

Martin Büttner

National Institutes of Health syndrome?

VisualMelon

not-invented here syndrome

Martin Büttner

2:14 PM

@Sparr yes there are definitely two distinct stages of the problem, and using existing approaches for the first stages is totally fair game

(especially since that can't be solved optimally anyway)

Nathan Merrill

@Sparr I made it so the entire layout of the pacman maze is given at the beginning, and the contents are given via line-of-sight

Does that seem better?

Martin Büttner

@Sparr however, I'm suspecting that in general it will even be possible to come up with setups which don't even have any discernable gates

so there might be a way around that

Sparr

@NathanMerrill it seems like a less contrived problem :)

VisualMelon

unless I have some inspired idea my results aren't going to look much different to that gif

Sparr

@MartinBüttner that's very very likely. using discrete "gate" blocks is definitely not the ideal approach.

Martin Büttner

2:16 PM

@Sparr it's just the simpler one I think

I think for M = 2 you can probably even brute-force the smaller grid sizes to find a setup that works

overactor

@NathanMerrill Looks good to me, maybe be a bit more clear about the initial input

Do you have a controller written?

Nathan Merrill

overactor, debugging it now

overactor, the initial input seems clear to me. Can you expound?

VisualMelon

I could always ship a 10Exabyte RAID array with the code, that might be cheaper than the RAM....

overactor

w**2 characters, representing the wall configuration (starting from top left, and heading right).

VisualMelon

(and I could pre-compute everything, so it wouldn't be CPU intensive)

overactor

2:23 PM

And after it meets the wall, it goes to second row, leftmost square?

0

Q: When was the last time the date was divisible by n?

A date can be represented by an unsigned integer as such: YYYYMMDD. What you need to do, is write the shortest program that figures out the most recent date whose number was divisible by a given number n (including today's date) and then returns that date in the format showed above. If there has ...

For those interested

Nathan Merrill

ok, will update overactor

overactor

other than that, I have see no real problems

@NathanMerrill maybe mention time limits?

both every turn and when they first get the map.

@Martin, you darned sniper

Martin Büttner

;)

did it during my lunch break, so I just had to post it now :D

Nathan Merrill

2:40 PM

@overactor that's mentioned at the end...would it be better with the post?

Martin Büttner

whoa github has a new interface for issues

overactor

I don't know how people feel about redundant information

but I feel liek repeating important information can't harm.

undergroundmonorail

3:26 PM

Hey, speaking of redundant information, would it be kosher to repost an answer to a KOTH challenge with one variable changed? I want to post this again, but with MODE set to 0. I'm curious which one would do better.

Martin Büttner

@undergroundmonorail you could just edit the second version into the same post

at least then no one could accuse you of farming for rep with it ;)

as for the game itself it seems fair to me

undergroundmonorail

Oh, true. I hadn't thought of that. :P

Thanks :)

Martin Büttner

loads of people did that on good-vs-evil because no one was sure if they were good enough to stick with the minority

DigitalTrauma

@MartinBüttner which rule am I specifically breaking?

Martin Büttner

the one in bold :D

"The program must output a definition of your word in 36 characters or more."

Sparr

3:36 PM

anyone here doing the dogfight challenge and not in the dogfight chat room?

DigitalTrauma

oh

I totally missed that

my parser is a bit dodgy sometimes ;-)

undergroundmonorail

@EoinCampbell I added another bot to my answer with the Gambler. I know you don't get a notification for edits, so I'm pinging you :)

I've never had to ping someone with a space in their name before. Will that work?

Sparr

@undergroundmonorail most contest runners will view the answer list sorted by activity, so they see edits first

undergroundmonorail

True, but I want to make sure he doesn't miss it, you feel? Probably not a big deal :P

Sparr

:)

Eoin Campbell

4:01 PM

@undergroundmonorail No problem man... will include it tonight.

undergroundmonorail

Thanks :)

undergroundmonorail

4:19 PM

Can someone help me with some C++? I'm just trying to print out a random number, but...

$ cat random.cpp
#include <iostream>
#include <random>

using namespace std;

int main()
{
        cout << rand() << '\n';
}
$ g++ -std=c++11 random.cpp
$ ./a.out
1804289383
$ ./a.out
1804289383
$ ./a.out
1804289383
$ ./a.out
1804289383
$ ./a.out
1804289383

Kyle Kanos

Don't you have to seed it?

undergroundmonorail

How do I do that? I'm very very new.

Kyle Kanos

I'm not a C++ programmer

So no clue

Geobits

8

A: C++ - Random seed at runtime

srand() As others have mentioned. srand() seeds the random number generator. This basically means it sets the start point for the sequence of random numbers. Therefore in a real application you want to call it once (usually the first thing you do in main (just after setting the locale)). int ma...

Kyle Kanos

But I am 150% positive you can Google search "C++ seed random number" and bam, it works

undergroundmonorail

4:21 PM

@Geobits @KyleKanos Thanks :)

Geobits

@KyleKanos The above is proof that your 150% is not too high :D

Kyle Kanos

@Geobits I was thinking that right as I read it

^My thoughts on pets

Martin Büttner

:D that's amazing

undergroundmonorail

Oh! I was looking for this‌. I watched it because it was interesting even though I'd never touched C++, but I was looking for it today to help me with random numbers.

I can't believe that out of all the things I searched, "rand() Considered Harmful" wasn't one of them.

I never bothered to make this computer understand most unicode, so I'm always surprised when the "I don't know what this is" rectangle shows up in the weirdest places...

Kyle Kanos

@undergroundmonorail I feel your pain. Especially on the posts that convert to unicode.

Martin Büttner

5:02 PM

yay, beaten PHP again

2

A: When was the last time the date was divisible by n?

Mathematica, 81 bytes For[d=Date[],!n∣(r=FromDigits@DateString[d,{"Year","Month","Day"}]),--d[[3]]];r Expects the input to be stored in n. Oh Mathematica, you and your lovely function and option names. Note that the vertical line ∣ is the unicode character for "divides", which I've counted as...

@KyleKanos your submission doesn't work

(also yay Unicode! :D)

you're decrementing the number, but you should be decrementing dates

Kyle Kanos

@MartinBüttner Oh poop, you are right

Sparr

5:16 PM

I'm writing a wrapper so a java-only contest can accept other language entrants. It will talk to those entries via stdin/out. Do you guys think JSON or a flat text format would be more appropriate?

Geobits

If the object of a wrapper is to be as inclusive as possible, flat text seems better IMO.

Otherwise if a language requires a JSON library, the OP would have to install that along with the language itself.

VisualMelon

base64 encoded binary ;)

Geobits

Assuming a library is even available. If not, then JSON has no advantage over flat text for them.

Sparr

ok, forget json

so, flat text

this is for dogfight, btw

each round you get info about four planes

would you want that on one line or four lines?

Geobits

Doesn't matter much to me, I know Java ;)

But four lines is probably better in general.

Kyle Kanos

5:37 PM

@MartinBüttner Fixed :D

Martin Büttner

aaaaand Mathematica takes the lead... :)

have an upvote though

I don't actually like 1:1 mappings for golfing, but I like Lua, so you're lucky :P

Kyle Kanos

@MartinBüttner Only by 9 chars

Martin Büttner

;)

Kyle Kanos

I kinda like 1:1 because I can write it in Lua & then replace keywords

Then delete all whitespace

(so glad Geany allows regex for Find+Replace)

Martin Büttner

I just prefer languages competing the way they are. If someone did a single-unicode-character mapping of all Mathematica functions all other languages could go home and cry.

2

I realise Lua as it is is at a bit of a disadvantage when it comes to golfing but so is Java, and people don't seem to have a problem with that

Kyle Kanos

5:41 PM

@MartinBüttner Do it & call it Martinica

Geobits

I'd just call it M

Martin Büttner

If I used that I'd get so many downvotes by butthurt GolfScript users :D

2

@Geobits I'd prefer fileformat.info/info/unicode/char/2133/index.htm

(which would also be a nice alias for MandelbrotSetPlot)

Kyle Kanos

@Geobits That's a waaaaaaay too short for a particular Trope

Geobits

@MartinBüttner I like

@KyleKanos I have a problem with any golf-specific language that has a name longer than one character. It just seems wrong.

Yes, that includes most (all?) of them.

Martin Büttner

J? that wasn't intended to be golfspecific though, was it?

I guess most of the single-letter ones were already taken

Geobits

5:47 PM

Yea, I don't think it counts, it's just good at it.

Martin Büttner

these are left: A, H, I, M, N, O, P, U, V, W, X, Y, Z

if you create one named "A" or "I" I'll give you the googlability award

"I" might even still work when searching for "I language", but "A language" is just a lost case

Geobits

For ultimate googlability, name a language "nude models".

Martin Büttner

as soon as you add "programming language" that would still work

but I'm pretty sure google just ignores "a" in english language search queries

Geobits

maybe.... but I still wouldn't try it at work

Or do the whole mapping with various curse/terrible keywords and name it NSFW

Sparr

anyone avoiding Dogfight because it's java only and want to test a wrapper I'm writing?

Martin Büttner

5:54 PM

the code would look like written by a voice-to-text programmer with Tourette's

@Sparr yes on the former, but I gotta get to work on Revival attempt number 3 :D

also I'd still have to get the controller running ^^

I'll use your wrapper when it's working though ;)

VisualMelon

ah, the parity checker has been solved

had to explore 65418 of the 65536 possible solutions, isn't that nice

Martin Büttner

nice! how long did it take?

haha, that's kinda sad

VisualMelon

too long ;)

Eoin Campbell

Gah! parallelisation sux

VisualMelon

I think 4hours or something, I can't remember

let's see what it looks like...

Martin Büttner

5:59 PM

@EoinCampbell There are three hard things concurrency, in computer science: cache invalidation, naming things, off-by-one errors and .

2

Eoin Campbell

yup.

though I think in this case it's MS specific...

VisualMelon

love it

Martin Büttner

wow my brain must have been so wrecked last night... this is what I did in Ruby to figure out if there are no zeroes in an array:

array.reduce(:*) > 0

VisualMelon

everyone loves even parity

Geobits

Oh, no, I get that. I was trying to figure out what it was doing. Like I said, I don't know ruby.

Martin Büttner

6:02 PM

oh they were definitely all non-negative

but a simple array.count(0) == 0 would have sufficed ^^

lol

Geobits

wtf...

I must have pressed up accidentally.

Kyle Kanos

I've got a decent idea for a KOTH, but it's going to be all in Fortran :(

Martin Büttner

concurrency, man

@KyleKanos as long as you don't require people to write their bots in Fortran ^^

@Geobits but yes, that did multiply them all together

Kyle Kanos

@MartinBüttner But I'd give the subroutine template......

No, Fortran has a system subroutine that allows you to run terminal commands

Geobits

Why not Malbolge? I hear that's a fun language.

Kyle Kanos

6:05 PM

But I don't have every language, nor bleeding edge compilers/interpreters

Martin Büttner

@KyleKanos btw I started on that project I mentioned... the first week was fun... I found some old code that did parts of what I was trying to do... problem is: the code was in F77 and for Clawpack 4.6 (current version is 5.2, I wasn't familiar with either)... also I haven't even used Python productively...

so with that background I ported F77 code written for Clawpack 4.6 in parts to F90 and in parts to Python such that it works with Clawpack 5.2 ...

Kyle Kanos

Yeah. Good old Randy LeVeque. Programming Fortran 77 in the 2000's. So sad :(

Martin Büttner

well they switched to F90 with Clawpack 5 now

at last

Kyle Kanos

@MartinBüttner Not much worse than Castro

Made up of C, C++, and Fortran 77 and Fortran 90

Martin Büttner

lovely

well now I can Python

Kyle Kanos

6:08 PM

Similarly with Enzo.

Martin Büttner

and I also can a bit of Fortran

Kyle Kanos

Fortran is the best language out there for Array-based programming.

(Waiting for some "hey Matlab....")

Martin Büttner

god no

although Python with numpy and co is pretty damn awesome

of course in Python you have to vectorise everything to get any decent speed at all

Kyle Kanos

@MartinBüttner Might look awesome, but it's horrendously slow...

Martin Büttner

@KyleKanos well yes, for most high performance stuff none of those are an option. but I could imagine that a nice vectorised Python routine could actually be faster than something very naive in compiled languages... because the vectorised stuff does happen in compiled and optimised C code after all

(it does, doesn't it?)

also Julia :) ... but they've still got a lot of work to do

Kyle Kanos

6:18 PM

I'm pretty sure vectorization is standard at optimization level >O0 for Fortran, not at all sure about C

Something like the n-body simulation, Fortran & C dominate: benchmarksgame.alioth.debian.org/u32q/…

While Python sucks

16 minutes for Python to solve the same thing it took less than 10 seconds on C/C++/Fortran

VisualMelon

no no no, don't get me thinking about re-writing this abomination

Kyle Kanos

8

Q: How did they take photos of Jupiter?

How did they take photos of Jupiter - I mean Jupiter is illuminated and that's a lot of light to produce. Am I missing something, and there was some sort of dark photo technology used, or was there simply enough light from Sun to begin with? Or is this photo a fake?

astrophotography jupiter

Great question with 8 upvotes. Two answers (mine included) have 20+ each

That's sad

Martin Büttner

@KyleKanos Well that's obviously well optimised code in all languages. In that case Python has no chance (but that's "just" an order of magnitude difference, that's not even too bad). I'm just saying if I wrote bad (cache inefficient, unoptimised) C code, that can easily be a factor of 10 as well.

Geobits

@KyleKanos The linked youtube video showing an example of over-photoshopping made the entire thing worth it, imo.

Kyle Kanos

@MartinBüttner That brings us to one of the criticisms: CLBG forces people to use the same algorithm. Something could make Python3 move a lot faster than 16 minutes, but they don't allow it.

@Geobits I <3 Pizza. It is beautiful :D

Sparr

6:32 PM

slightly off topic, but pertinent to a codegolf/koth project I'm trying to tackle. I'm having trouble with pauses on buffered I/O between processes in Java. Specifically:
moveString = reader.readLine(); <-- this is using a BufferedReader, using an InputStreamReader. Execution of my program appears to pause at this point. When I ctrl+c the process, execution resumes with the input read and runs a few more lines of code before actually stopping on/before this line: Pattern p = Pattern.compile("([NSWEDU]+) ([01]) ([01])");

Nathan Merrill

so, it doesn't read it all, it just stops?

what character does readLine() pause on?

Martin Büttner

@Sparr does the program you're reading from not flush the STDOUT?

Geobits

Or, does it actually contain a newline character?

Sparr

@MartinBüttner you're a genius :)

Martin Büttner

;)

been there

took me forever to figure out when doing the Vector Racing challenge, because I assumed all the time there was a problem with flushing in the controller, but I forgot to add the flush to the solver as well

Sparr

6:40 PM

how does this look for a flat text format for non-java dogfight planes?

NEW CONTEST 14 100
NEW OPPONENT 10
NEW FIGHT 0 0 0
ROUNDS LEFT 99
NEW TURN
alive 13 8 13 N 0
alive 13 5 13 N 0
alive 0 5 0 S 0
alive 0 8 0 S 0
[get 2 lines of input of a format similar to "N 0 0"]
ROUNDS LEFT 98
NEW TURN
alive 12 8 13 N 0
alive 12 5 13 N 0
alive 1 6 1 SEU 0
alive 1 7 1 SWU 0
[get 2 lines of input of a format similar to "N 0 0"]
...
[no explicit output for the end of the fight]
NEW FIGHT 1 0 1
ROUNDS LEFT 99

it's basically just a dump of the data that goes into the methods of the java planes

Martin Büttner

This gave me eye cancer codegolf.stackexchange.com/q/35418/8478

It's also a bad challenge, I think.

oh, I remember where I've seen him before codegolf.stackexchange.com/q/25329/8478 :-/

Kyle Kanos

6:57 PM

@MartinBüttner Too broad for sure

Sparr

@MartinBüttner ok, if you get time to write a plane for dogfight, give my wrapper a shot. github.com/sparr/Dogfight-KOTH/commit/…

Martin Büttner

7:18 PM

this dude is really good with criticism codegolf.stackexchange.com/questions/35418/…

and I tried really hard neither to be condescending nor telling him that is challenge is crap

@Lembik, do you still get pings in here?

@Doorknob, could you superping Lembik? I'd like to tell him that there was no need to delete his challenge in panic.

Geobits

7:50 PM

I'm thinking this would be a good capture-the-flag map

Sparr

you'll have to elaborate a bit on the ctf mechanics/rules before we can judge that

Geobits

That's coming, working out the mechanics now.

Sparr

be vague now? :)

Geobits

Three troop types, attacker/defender/scout. 500 total troops, team size varies by number of entries per team.

More entries for team A mean fewer troops for team A.

Each entry chooses a troop type and team.

Objective: grab the flag and bring it back "home".

Martin Büttner

I wonder where you got the idea to choose from one of three bots ;)

Sparr

7:54 PM

so I'm writing a bot that has to cooperate with other people's bots on my team?

Geobits

I was going to have 5 types, but figured it wasn't needed.

@Sparr Yes

Also, each troop gets one mine to place during the game.

Sparr

so if I'm on team A and you are both on team B, there are maybe 300 copies of my bot and 100 copies of each of yours?

(assuming a three entry match)

Geobits

It's team level = 500 * (other / (team + other))

Where team is your team entry count, other is the other's count.

Martin Büttner

ah, how much a bash solution beating you can inspire you to greatness :D codegolf.stackexchange.com/a/35394/8478

if that's not an improvement

Geobits

So for 1v2 it would end up being 266v133

Martin Büttner

7:56 PM

(I've rarely been so proud of finding a golfing trick)

Geobits

Obviously I'm hoping for more than 3 entries :)

Sparr

do I decide which team my entry goes on, or does it decide at runtime? every match?

Geobits

You decide, but it has to be constant. Same for troop type. You can't choose at runtime, you just tell the controller at runtime.

Also, you can have multiple entries, but only for the same team.

Kyle Kanos

@MartinBüttner: Wow, cutting off 30% of your byte count....very impressive

Martin Büttner

@KyleKanos actually 7 more characters

I don't even need ~Take~3

Sparr

8:02 PM

Having to choose a team permanently seems... not awesome. Sorry if that seems insulting.

how do you pick the winning entry?

Geobits

1) Flip-flopping is less awesome IMO 2) No "winning" entry exactly, just winning team. So, no green checkmark.

Or at least, no way to pick one that I can see.

Hopefully people don't feel that's such a big deal.

overactor

It's not exactly a KOTH then, is it?

since KOTH implies a victory condition

You could make it a KOTH popcon combination

highest voted from the winning team wins

Martin Büttner

@KyleKanos there's the fixed version and explanation

Geobits

Exactly, but I don't know of a better way to describe it, and no tag yet for something like this. The koth tag-wiki doesn't explicitly give a winning condition.

Sparr

if an entry could choose to switch teams (like good vs evil) that would make it much more KOTH-y

Kyle Kanos

8:07 PM

@MartinBüttner I was about to ask for an explanation

Martin Büttner

@overactor that would be horrible

Geobits

I do like the idea of highest voted from winning team.... maybe.

Martin Büttner

I'd say put it up on meta

Geobits

The problem with that is EmoWolf

Martin Büttner

what people think about a team-based KotH

Sparr

8:07 PM

alternately, you could arbitrarily team up entries, and give them cummulative scores.

overactor

@MartinBüttner I see your point

Geobits

@Sparr That would get quite messy with a lot of entries. Testing every combination of teams would take forever.

Martin Büttner

I had already suggested rerunning the trial once for each bot from the winning team without that bot, and give the green checkmark to the bot with the worst resulting trial

overactor

Maybe actions give you points?

Sparr

@Geobits don't test every combination, then. randomize it, if you have to.

overactor

8:09 PM

winner is one with most points

Martin Büttner

(since you could argue that that was the best bot from that team)... but even that would take too long

overactor

a bit like in world of tanks

Martin Büttner

@overactor it's going to be hard to balance the different troops against each other

Geobits

@overactor What gains a point, though?

overactor

depends on the nature of the game

Geobits

8:10 PM

With attackers, defenders, and scouts...

Sparr

@MartinBüttner why would that take too long? with 50 entries that means running 50 copies of the contest. an hour or two?

overactor

spotting an enemy and alerting your team of it's position

carrying the flag

killing an enemy with the flag

or taking it from the enemy

Geobits

@Sparr You're assuming that a whole contest only takes a minute or two with 50 entries?

Sparr

@overactor bad idea. no scoring metrics would be immune to game-ing

overactor

What about best score from winning team?

Sparr

8:12 PM

@Geobits maybe? you'd want to choose a format that optimizes runtime. no process-per-player-per-round like caveman duels. something like dogfight.

overactor

That way you have to balance helping your team and playing for yourself

Sparr

@overactor still very game-able. the winner would definitely not be someone who helped their team the most.

Geobits

@Sparr I was planning on doing it with pipes like I did with Hunger Gaming.

Martin Büttner

@Sparr still... with 50 entries I'd expect the game itself to take an hour at least

@Geobits how long did hunger gaming trials take?

Geobits

~10 entries, around 20 minutes, but it was highly dependent on exactly which bots played and how fast they ate.

The ones that ate quickly shrunk the time. The ones that ate only when needed made it last much longer.

The main drag on time for that one was the huge flocks involved.

Sparr

8:14 PM

ok, so you run your 25vs25 match for an hour. then you run a bunch of 24vs25 matches that are much shorter to figure out which entries are contributing the most. Then you only need to run maybe 4 more full-length matches. 5 hours, total?

Geobits

Okay, assuming I do, how do I define "contributing the most"?

Sparr

as mentioned above, remove them, see how much worse the team does

Geobits

Define "how much worse".

Sparr

fewer flags captured

Martin Büttner

@Geobits time the flag is captured

or number of flags if there are multiple

Sparr

8:16 PM

or the opposite of what @MartinBüttner said :)

Martin Büttner

;)

Geobits

How could a defender "contribute" then? It wouldn't affect the flag-capture time at all, unless removing them actually made the team lose.

Sparr

I've been assuming the game keeps going after the flag gets captured, for multiple flag captures, like a CTF video game

Geobits

The object is to get the flag back home. If someone is carrying the flag and dies, it sits there until picked up by another teammate.

Single flag per game.

Sparr

but there are multiple games?

Martin Büttner

8:18 PM

@Geobits enemy flag distance from home base?

Geobits

Possibly, not sure.

Sparr

no offense to futbol fans, but playing to a score of 1 isn't a very reliably comparable game :p

Martin Büttner

some mixed mixed score of these?

Geobits

@Sparr Chess fans would probably disagree.

Martin Büttner

:D

Sparr

8:19 PM

chess players play each other more than once, at competitive levels

Geobits

Yep, and I said "possibly" :D

Sparr

well, if you're willing to accept the suggestion, I think you'd have more scoring and strategy flexibility if you went with a video-game-like ctf system. game runs for X turns no matter what. number of flag captures during the game is your team's score.

like tagpro.gg :)

(or unreal tournament, or quake, or ...)

Geobits

Eh. I'm thinking more like actual CTF that you'd play. In a field/forest. In real life.

Sparr

last time I played airsoft ctf, we played first team to 5 captures

Geobits

That's.... not... game runs for x minutes, so I don't see how that would be different than playing single flag, 5-of-9

Sparr

8:23 PM

5 of 9 is better than 1 of 1 :)

if you're playing 5 of 9, then you identify MVP by seeing the point spread without a player

Geobits

Right. I'm trying to decide how/if to do multiple rounds.

Sparr

if we win 5:4 and you remove bot X, then we lose 4:5, then bot X was worth 1 point.

Geobits

If I'm reading you right, that would mean 9*n runs.

Sparr

problem with such small numbers is there's a lot of statistical noise to overcome

yeah, I hadn't considered that you might run an arbitrarily long round. that kills most options for repetition.

I didn't do the hunger games challenge. I've mostly done the ones with turn limits.

Geobits

I'm definitely putting a hard cap on number of rounds per game, but I don't know what that will be until I run a couple test bots.

It will be high, though, and only to eliminate stalemate situations.

Also, if an entire team is eliminated, the round is over without the need to go capture their flag.

Sparr

8:26 PM

thoughts on what languages you'll support?

Geobits

Anything that can read/write to STD I/O

Sparr

awesome

Geobits

I do want to do multiple rounds, I just don't know how many it will be at this point, since I don't know how long a match will take.

The turn cap will probably be something like "X turns with no deaths or flag movement" or similar.

If capped, the team that got the flag closest to home wins. If tied, by troop count.

Sparr

absolute troop count or relative to starting count?

Martin Büttner

here is a suggestion for quantifying who well the team did: obviously if it lost it did worse than if it won. between two runs where it lost or won in both, you can find a score based on distance between flag and base of the losing team as well as time taken by the winning team to get the flag back.

Sparr

8:41 PM

I think there will be a lot of situations of one bot interfering with another specific bot. not on purpose, just because their algorithms don't mesh well.

Martin Büttner

yeah, I think it's generally problematic

hence (@Geobits) make a meta post how people feel about a KotH without a winner

Transcript for