@Rusher FYI my wolves are not supposed to move (only move is HOLD), they are just waiting for people to attack them. In the original script they were moving but i decided not to in the end. Thanks to both you and @ProgrammerDan for all your efforts in making my thing work! I promise my next submission will be in python (I'll use some snippets from PorgrammerDan's code for the connections).
I think I just broke my personal record for "slowest algorithm for a pretty simple task": a prime checker that takes about a second per prime at around 6.5 k
Python
Arrays are efficient. Let's use one of those.
def isprime(x):
l=[False]
for i in range(x):
i+=1
prime=True
for j in range(2,i):
if i%j==0 and j!=1:
prime=False
l.append(prime)
return l[x]
def ishalfprime(x):
ret...
@ProgrammerDan When you said "hooray, you implemented my idea!" I thought you meant "I had the idea of a wolf that just ran away from stuff". I just searched "sheep" in chat and apparently I thought of the same name as you, too. I didn't mean to steal your creativity, I swear I just came up with it on my own! :P
@undergroundmonorail the best ideas are often thought of independently and nearly simultaneously. No worries at all, I figured your submission was independent :-)
@Geobits It's possible to drive some screws by gentle application of a hammer, or to drive a nail into soft materials by hitting it with a screwdriver, but that doesn't make it the right tool.
True, if you have both at hand. Sometimes the right tool is the one you have.
@Rusher Exactly which wolves do you have in the arena? I'm asking because on my machine, Lazy is still winning 9/10 times even with Camper in the mix. Camper's averaging about 30-40 survival here.
@plannapus ProgrammerDan walked me through every bit of it too, so +rep him if you get a chance.
@Geobits I have Lion, Bear, Stone, EmoWolf, GamblerWolf, Sheep, GatheringWolf, GamblerWolf, WolvesCollectiveMemory, Wion, CamperWolf, and WolfWithoutFear.
There is a Scala and a Javascript one along with a few more Java submissions that came in yesterday evening that also aren't in yet that should be.
I also only ran 5 tests, and Camper won 4 of them. You still won one. I've got it running so fast now that I could just run 100 tests and leave no room for error.
Probaby. Those are the easy ones to include, but I'm forcing myself to get Gareth's scala one in first. I figure if there are more submissions than I can keep up with, then it isn't time to post results anyway.
There are some other class challenges from my college days that I want to post too. They were all pretty fun
@undergroundmonorail The other thing with KOTH is that you don't want more than one running at a time. I have one in mind for a game which is more complicated, but I'm waiting a bit before I post it to the sandbox.
@Rusher I keep waiting for OP to come back and fix, the changes seem easy enough. I was happy to help out plannipus b/c (1) I don't know R (2) None of the problems were lazy coding, they were all legitimate "tricky" bugs
@ProgrammerDan When you said genetic algorithm, did you mean something like this? Choose a sequence of moves and attacks and record my survival rate. Randomly mutate part of the sequence. Did it improve my survival? Great. If not, revert and try another mutation.
I watched the first few lectures on Machine Learning from Dr. Andrew Ng at Stanford and he brought it up in a discussion about chess engines that learn by playing thousands of games against each other.
That is a valid approach, but that's going to slowly. The best is this -- choose N sequences of moves (or approaches, or X), run them through the mill, and evaluate the fitness function for each sequence/approach/X. Pick some M of the best (Top 5, perhaps), and some random set from among the rest (to help avoid local maximums).
Now, make your subsequent generation by sexual or asexual genetic assignment. If asexual, randomly mutate (some/a) move(s). If sexual, apply a scheme to "cross" the sequences/approaches/X so that the result is a combination of two randomly chosen members of your "fittest" selections. In sexual as well, randomly mutate some move(s), if you want accelerated searching.
@Rusher The English questions which I currently see are about the meaning of the verb "to mush" in the context of huskies and the meaning of "15-strong armed gang". Should I check that my last will and testament is up-to-date?
I think the main problem with a GA Wolf is there's not a good way to do it without a lot of testing. 1000 iterations is too short to do it on-the-fly, but if you're doing it ahead of time, you're going to run it after every submission, because each one might change the approach significantly.
@Geobits If I figure out how to do it, I will run the tests on my own for many iterations. Once it has a good sequence, I can just submit the hardcoded sequence. But yea, it changes after each submission.
@Rusher Right, that's what I mean. But then when another submission gets added, it could completely invalidate the approach(larger board, different wolves, etc)
b/c origin placement is unique per execution of the game due to PRNG (unless GamblerWolf takes over ...) so the solver will be searching in a space that has no comparability.
GA solvers in chess, etc. work well because the gameboard starts the same, the search space is confined, fitness is well defined, etc ...
Here, you need something that in general solves the base problem well, e.g. survival in a crowded room full of wolves with very different natures that you can't tell apart
My favourite example of a genetic algorithm is some guy on youtube who set up an NES emulator to play Mario until the "you died" flag got set and "selected for" longest time before dying. The guy who set up the script didn't realize that dying to an enemy and dying via bottomless pit sets a different flag, so the script's "lose condition" was /only/ dying to enemies. Within a few generations they were jumping into the first possible pit and idling on the "game over" screen.
I'm thinking of a boids-related KotH. A huge flock of prey, submissions as predators(with fatigue/speed based on hunger level). The prey would naturally try to avoid them and use flocking behavior. Hopefully I can work out some details that are bothering me.
@Rusher No, the only real "flock" is the prey(around 1000). Predators would be given a small "pack", maybe 5-10 individuals. If you get too hungry, you die. Last man standing wins.
I'm going to try to run I/O similar to what Gareth did with the BattleBots, so as long as I can run it from a command line on my box, it should be good.
I basically crafted a fully-realized socket-based API for your Wolf challenge, @Rusher, by specifying every interaction with a remote program as a request-response protocol.
The only difference is my interaction happens at a system PIPE level -- there's functionally no difference in many languages to extend that same approach to a Socket level. Just need a well-formed API.
@Geobits True. I guess if you want to encourage broader base of submissions ...
se circulates misc ads from other sites apparently.
its not so much that they "disapprove" its just they dont vote much, the metas are rather quiet usually except when theres an occasional "disagreement"...
suggested that to doorknob. he seems to be a leader type around here aka "alpha male" wink
you guys are quite amazing how much you upvote hot questions around here.
the whole se site is oriented around active voting. to active voters go the spoils so to speak.
so some "ninja type special operations" could be impressive, wink, wink
afaik you get +100 just for signing up to other sites & then can immed vote on meta threads...!!!
generally se inhabitants rarely coalesce into factions/special interests, there is a lot of contention even on individual sites, but do see some "collaboration". eg around here!
the community ad mechanism seems underutilized on various sites...
@vzn You remind me of how I used to be in Junior High. I used to form theories about the circles that JH kids form and how some people would slowly be shut out and others would be absorbed. And some of the circles would occasionally merge and then devolve in chaos. And about how speaking up would suddenly bring you back into the circle, or laughing at a joke that wasn't funny.
@Geobits If you do that, it might be worth looking into what Thomas Eding suggested on a comment to the BattleBots question. I suffered from speed issues that it might have solved.
@Gareth That's probably a good idea. Although I must say, doing nothing but removing Rifter sped up my test runs considerably. He liked to add an exponential amount of work to the process.
@ProgrammerDan I thought about that also. Especially considering the logistics of my idea... Running a 1000+ flock with another x entities, all running on a short timestep... The tricky part is judging how long it takes on my testbox vs your testbox.
@Geobits @ProgrammerDan While Rifter was slow, I still think that the fact I was starting each process for every round didn't help matters much at all.
@Gareth oh, yeah ... that's a big issue. That's why I went the more complicated route (on my end) of persisting the process invocation over all rounds for @Rusher's challenge, to avoid that overhead.
Remember what I said the other day about lots of smart people getting together and building something? I bet we could build a KotH "platform" whose sole purpose was to communicate with clients written in many languages over stdin/stdout, handle exceptions and timeouts, etc. It would allow for easy writing of KotH challenges.
Ever thought about selling? The game is becoming kind of popular (in other words, extending to non geeks) but it won't stay that way I believe. It has to die at some point.
I might have a few Ravnica, but most are somewhere between Revised and Fifth Dawn. Lots of 5,7,Tempest. So old, but not crazy old or very valuable. I'm torn between selling and waiting until my son grows into them, and so far he's winning.
Well, he's 7, so it's a bit early I think. He plays the Pokemon TCG, and seems fine at that, but MTG has some complex rules, and he's not great at the subtleties yet.
@PeterTaylor Actually, the largest Grand Prix ever held was Modern instead of Standard. Granted, there are less Modern GPs so more people showed, but 5000 people is nothing to sneeze at.
Well, I don't care about tournaments or sets, since I mainly just played with friends. Our rule was "whatever cards you have". Either way, a huge box of old cards is a good way to introduce someone to the game.
Plus I have 20-odd built decks sitting there I can walk him through to explain why they kick so much ass.
Since we were talking about machine learning not to long ago, someone on Board and Card games tried to argue with me that MTG could be broken down to a dominant strategy. en.wikipedia.org/wiki/Strategic_dominance
I looked through every single comment I have on B&C and it isn't there...
The question was like "Optimal Number of Lands"
The OP said something like "Every tool or guideline that helps suggest lands so far has been built on "experience" rather than cold hard math."
And then he started downvoting responses that contained any nugget of wisdom that wasn't based on math.
I suggested that MTG didn't have a finite set of possible moves, and so couldn't be solved, and he retorted that there should still be a best number of lands for a deck, given the cards.
I didn't really know where to go from there, and I have a rep for arguing on B&C so I stopped responding.
I suggest you guys stay away. They only like people who are there to "have fun". Serious rules questions are shot down the moment you appear to be proving someone else wrong. Proving someone else wrong is like our equivalent of homework questions.
It's probably true that there is an optimal number of lands for a given deck, but it's a difficult problem to define (unless you're also willing to reject the received wisdom that you shouldn't exceed 60 cards, or you define an order in which you'll exclude cards from your list). And it's a long way from that to the existence of a dominant strategy for the entire game.
Don't worry: I'm not in any hurry to regain my rules advisor qualification, so I have no real incentive to learn the corner cases of recent mechanics.
I can only see getting an optimal number of lands if you're given both the deck and the intended strategy. If your deck comes down to a combo of two mid-cost cards, you're going to need more land quicker than if they aren't intended for play on the same turn. etc, etc.
But to hand a computer a deck and say "how many lands" seems an exercise in futility if you're looking for a true optimum.
Talk of Nash equilibria is irrelevant anyway, because metagames are partly influenced by fashion and personal preference rather than being mixed strategies executed by purely rational actors.
I suppose it is difficult to get the tumbleweed badge here. One would have to ask an incredibly challenging (to avoid answers), perfectly clear (to avoid comments), somewhat boring (to avoid views) question. And even then, you might still get a vote.
So I need to ask for tips in an obscure language, but not soon because you guys will remember I said this. I'll bide my time and strike when the tumbleweeds are ripe.
Hmmm. I guess the opposite extreme would have to occur then. The site has so much traffic, that a new question is buried so fast that nobody sees it. Anyone that does see it should find the question too challenging or boring to bother.
And gah... I have to be moderator for a year to get a badge. Now I have two reasons to improve my English
Hiding information in Cats
code-golf image-processing
You are a secret agent trying to communicate with your fatherland. Of course the information needs to be hidden so no one eaves drops your message. What would be better suited than a cat? Everyone loves funny pictures of cats [citation neede...
@Geobits I Googled that word because I thought you mispelled it, and was promptly corrected by Wikipedia. "Stenography" redirects here. It is not to be confused with Steganography.
I had a 103 fever one time, and was watching The Land Before Time. When they thought Petri had died by falling off a cliff into the Big Water, I started bawling. Turns out he could fly and all was good.
It's one of the things I wish most strongly I had participated in during highschool. Or even college.
Instead I did some programming challenges, but robotics would have radically changed my career progression, and from where I sit now? Only for the better.
If I ever have a kid, we're going to move to a 5A school so he can specialize in something other than football. My school didn't even have a robotics club. We didn't even have a programming UIL.
@Rusher if my son shows any interest, I'll encourage the hell out of programming/robotics. I would much prefer that to sports, the returns are so much greater to formulate a formidable mind vs. a short-lived formidable body. In my opinion. :)
and as a disclaimer, if I should ever have a daughter, I'll do the same to her!
On the surface, robotics is a sport, with teams competing to win. At it's core, it is about working together to accomplish what you never thought you could, and making friendships too. And building a robot.
Nothing wrong with sports, exactly. The problem is that the job market for it is ridiculously limited. You can get a job in professional robotics without "making the draft". Of course if it weren't that way, it wouldn't pay shit.
That, and people thinking they don't need a fallback position.
@ProgrammerDan I'll be honest, I hated being forced into sports as a kid. I cried on the way to my first tennis court. Three months ago I got a real job and started feeling really tired and lazy every day. I started playing tennis again at the local club and running, and now I feel great. Being in shape really improves my focus.
@Geobits Yeah, that's exactly my thinking. I was involved in sports and loved it. So I'm not going to discourage that, unless the sport is extremely high-risk (e.g. football). But the returns on the average, or even in any standard deviation from norm, just aren't there.
You've gotta be incredible to get any kind of return, and even then your overall "lifetime" in the sport is short.
Yep, but people do it for the same reason they play the lottery. If you do hit it, you can hit it big. Then, most people don't know how probability works, either.
Hiding information in Cats
code-golf image-processing
You are a secret agent trying to communicate with your fatherland. Of course the information needs to be hidden so no one eaves drops your message. What would be better suited than a cat? Everyone loves funny pictures of cats [citation neede...
Our program should check that there are enough bits in the given image to encode the message, or should it rescale/resample the image until there are enough bits?
Yeah, for Doorknob's recent "Roll D&D Dice" challenge I wrote something I was pretty proud of, then realized that adding the required input validation would be a huge number of characters and scrapped the whole thing.
You are a secret agent trying to communicate with your fatherland. Of course the information needs to be hidden so no one eaves drops your message. What would be better suited than a cat? Everyone loves funny pictures of cats [citation needed], so they won't suspect secret information hiding in t...
Ruby, 30
require'prime';f=->n{n.prime?}
Declares a function that returns whether its argument is prime.
Alternative version (same length):
require'prime';f=proc &:prime?
Obviously bending the rules, but you never said built-in prime functions aren't allowed.
Python
Arrays are efficient. Let's use one of those.
def isprime(x):
l=[False]
for i in range(x):
i+=1
prime=True
for j in range(2,i):
if i%j==0 and j!=1:
prime=False
l.append(prime)
return l[x]
def ishalfprime(x):
ret...
Ruby, 30
require'prime';f=->n{n.prime?}
Declares a function that returns whether its argument is prime.
Alternative version (same length):
require'prime';f=proc &:prime?
Obviously bending the rules, but you never said built-in prime functions aren't allowed.
