@Rusher you don't have time to make more units. you have six bases with 5 units each. but enemy bots could attack all 6 bases with, say 60 units in total. you'll simply die.
Question: If we play two-player Risk, there is no collusion between players. Can you tell after two turns who will win simply because they're in the lead?
it doesn't matter if I'm losing as well, in the end you're left without any bases, and I have yours, so I can recover with an enemy less to worry about
Logical bots would attack a base at a time. Once that base is gone, you don't produce there any more. They still have all theirs. They move to next base. Now they're gaining on you.
your enemies collectively have more units than you. if they act as one (because they all decide to attack the leader) how is that not equivalent to you not having the lead?
Let me put it this way since you guys are both being stubborn. If everyone gangs up on one player, then he can lose the lead. If not, then the leading player cascades. Is that better?
I don't really think it takes much. Two decent-strength players could gang up on the leader. In a multiplayer game like this, I don't see that as unlikely.
I'd be very surprised if less than 20% of the bots submitted didn't have "attack the leader" as their base strategy in at least some phase of the fight.
in any case I really can't tell how much of an effect the dice will have, and I won't be able to tell until I've written the controller and two or three simple bots to test it with
I never claimed that "attack the leader" is a perfect strategy, and I wouldn't build a bot that way. It might be part of an overall strategy, but I don't know yet since I haven't sat down and played with the control program. My argument is that "don't attack the leader" is flawed, also.
@Rusher simplified: my bots are idle, I want to attack a base (because all planets are taken), I look at all bases of the leading player, and attack the one with the least units. as long as the fight is going well, I stick to that fight until that base is mine.
What I keep repeating is, "I can't tell", and I don't know how you can. The strategic space is way too vast to make any predictions on how the interaction between 5 or 10 different bots may turn out.
Yes. As Geobits said, such a simplified strategy will not be the best one. But I don't know how you can be so sure that there will be no good strategy that will make it harder on leading players.
And even if a bad strategy is the only way to prevent leading players from securing the win early on, that doesn't mean that such a bot won't be written and neutralise the initial problem which was that the game could be decided too early on.
Assume that attacking the leader is part of the best strategy. To attack the leader, you must not be in the lead. Therefore, attacking the leader is not a part of the best strategy.
Are we purely arguing the best strategy for the game now? Because I don't think that's relevant to the challenge design any more. The original problem that affected the challenge was "can the game be decided by an early lead". And I don't think that's the case, because it depends on what kinds of bots will participate.
The strategy sucks because I can see that "limiting your options to attacking the leader", "limiting your option in some cases to attacking the leader", "limiting your option after some threshold to attacking the leader", etc. are all poor strategies.
In order to dethrone him, something would have to incentivise two other bots to attack him. If not handled quickly (because it will cascade), then it will require three other bots to attack him, and four, and so on. And it won't rise linearly, because each additional lead builds on itself. I don't see what incentive other bots have to react to this threat quickly.
Again, if someone writes a bot specifically designed to target leaders, then that bot will surely lose and then after he loses the remainder of the game will behave as I said it would.
and if you're bot is strong enough to take on all other players, then he has won the game, simple as that... it's the same in risk... except there it's a bit more likely for all other players to work together
The only time attacking the leader is a good idea in risk is if you think you can find a way to come out on top or at least improve your standing in the world.
Destroying ships or taking bases owned by the leader gives you a significant bonus that ever so slightly offsets the defender bonus gained from defending.
@m.buettner Qualify that with "failing to be able to identify the leader, either accept that an early lead due to RNG could outweigh the best strategy or scrap the challenge"
@m.buettner I'm done discussing this at least for now, but one last thing: Don't scrap the challenge. Let him enter his obviously-better-bot. I remember I had an "obviously-good" strategy for a KotH once, and it did win for quite a while. But then LazyWolf got beaten after people did a bit more analysis.
@Geobits yeah, I wasn't going to scrap it without at least trying it out with 3 plants and seeing what happens, but even then I can't anticipate what will happen with others' bots
I would use bases. A player with tons of bases will quickly have tons of units. A player with tons of units may have 1 base and be guarding it all game.
I just suggested bases. It doesn't really matter. If you put a bounty on the people with the most bases, then they'll think twice about capturing more than they can chew.
And since the goal of the game is ultimately to have the most bases, then they have to leave their shell at some point
Then bots have to find a balance between growing wide and growing tall.
also "from a current leader" is still not objectively decidable... you'd always have to get the bonus but it would have to increase depending on the number of bases that player still has
well, it's almost half 2 over here, so I'll go to bed now
I'll think about this a bit more, maybe such an incentive is a good idea anyway, but I'm still not certain early wins are bound to happen otherwise. (or if they do, if they are actually so early that they are a problem)
@Trimsty progress! mbuettner.github.io/midgard/public you might want to crank up the number of polygons to 1024 or similar. and if your screen's resolution is "only" full HD, you'll need to view it on full screen.
@m.buettner I don't know what a relaxation pass is, but I told it to do 9999 of them and it just about killed my whole browser session. Even after I had managed to kill the tab, things weren't responding. :P
If I write a code-golf challenge and give a -5 bonus for not using the plus operator, I would still want to tag it code-golf unless there is an overwhelming community problem with it.
Writing challenges is the one place I will conform whether I like it or not. I only argue here :-/
the question is where do you draw the line. what are the bonuses significant. so IMO it makes sense to make it strict and say "bonuses -> code-challenge".
The meta guy had a point. Basically everything on the site is a code-challenge. Just use the tag when your post isn't covered already by other well-known tags.
I think it makes more sense to have code-golf-with-bonuses fall under code-golf because someone who wants to write short code isn't going to look under code-challenge for somewhere they can do that.
@undergroundmonorail sure but if your bonuses massively outweigh the code size and the code size becomes negligible?
we had this case this morning in the sandbox: meta.codegolf.stackexchange.com/a/1634/8478 ... it's fixed now after I pointed it out, but as it was originally posed, it was certainly not code-golf
@m.buettner We run into the same issue as before. Define "negligible". Until we have a way to do that, it makes more sense to make the rule work with the more common case (bonuses are relatively small).
@undergroundmonorail in the case from the sandbox, is what obvious that they were negligible, because code-size was only able to break ties of scores which were determined only by bonuses
And if you said yourself "If the community wants your challenge to be searchable by code-golf tag, then they'll tell you". If we're leaving tags up to the community, would you say that you can recognize what the community will decide for any possible case?
The difference between tags and votes is that votes are meant to be subjective. They're defined that way in the help center. I don't even understand what your point is, but you have seem to a pretty nasty habit of getting upset with people who try to decipher your point so if that's the direction we're going we might as well drop it here.
Fuck, I came off like an asshole there, that's not what I meant to do. I'm sorry.
I get defensive when people try to get more out of me than I said.
If the question was "why do you do things like that" I would happily answer. But the question is always "Why do you do things like that? It's as if you believe blah blah blah."
It's as if, you seem to mean, you implied that, etc.
And who is the one trying to put a number on the word negligible?
Not me.
If you want to put a number on it and you get community support I'll follow the rules but don't challenge my opinion and then call my response pointless.
my point was you can't objectively determine negligible
all I said was, I don't care how you tag a code-golf with bonuses. I said you can't draw an objective line, so I find it understandable why there would be hard and fast rules in either direction.
@m.buettner I made a risk dice simulator. The math is interesting. Turns out that for equal-sized armies, the defender has a large advantage when the armies are small, and the attacker has a large advantage when the armies are big. The turning point is at about 11 each.
anyway, I don't want to spend time on this discussion, because it really doesn't matter at all to me. someone make a new meta-post if it's important. I've gotta go.
@Geobits could this be due to the attacker running out of chances to attack with 3 units too early?
@Rusher I was attempting to say that if we decide whether something is code-golf or not by whether the code size is negligible or not, we need to know what negligible means. For that reason, I don't feel that it's a good way to decide how to tag something. I'm sorry I didn't make that as clear as I had thought.
