Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[WanderingWinder]

This recent topic sparked a lot of debate, and so I figured, hey, why not let the simulators figure it out? Also, I've been meaning to run a simulation tournament anyway. So, here's the challenge: make a bot using whatever the heck you want*, starting in an average case scenario (you can make it do different things based on how you start, but no special start state!), to beat all the others. Now, a few caveats. You only get to reference up to 5 different kingdom cards (plus whatever basics, colony, platinum, and potion), so that in any given matchup, a 10 card set can actually be made out of the clash from any two of these bots. Furthermore, I'm going to go ahead and ban vault, Bishop, torturer, young witch, and possession. Vault, Bishop, and torturer because it's really annoying for the opponent not to be able to make the 'correct' decision in response, as you can't do in the simulator, young witch because the bane is such a big deal and the opponent can't know what it is, and possession because, well, how do you know what you're doing with their deck if you have no idea what their deck is? I don't actually think any of these cards would be used in a winning submission, but I do regret having to prohibit them. If you want to lobby me with some great reason to allow them, here in the thread or via PM, I'm listening. Also if there's some other card that I've missed that I really oughta ban.

Specifics: Each person can submit up to three, that's 3, yes 3 different bots. You can't have any more than one card of your five overlapping between two sets; i.e. I can have herbalist in set A, secret chamber in set B, and both in set C, but I can't have secret chamber AND herbalist in two of my three sets. Submissions should be sent to me via PM no later than midnight GMT on the night of April 21st-22nd, 2012. I'll try to post the results sometime later that week. If I get too many entries, I may do some quick pruning of what seem to be weak bots before pitting them against each other in the finals. Hopefully it won't come to that, but I will at least look at and test every bot that comes in. The finals will be round-robin, starting with 10,000 game trial runs. Anything within 5% will get re-run at 100,000, and if it's within 1%, I'll rerun it to probably 300,000. Things that are exceptionally close (i.e. like a tenth of a percent after 300,000 runs) will just get called ties. All bots should be in Geronimoo's simulator.


I think that's everything, though I may update this post later with more info.

[WanderingWinder]

I won't be participating this time, since I'm running it, though I do have a few home-brew 3- and 4-card combos that are pretty strong. But as a word of note to all you guys, there's one built-in strategy in the sim that's about the toughest thing to beat that I've seen.....

[Tables]

I was thinking about this, actually. More specifically, about if everything is available, does every strategy have a suitable counter-strategy (something that beats it at least a majority of the time)? My intuition says, yes, probably. But I can't know for sure.

I might try and whip something up for this.

[blueblimp]

I feel like it will be very hard to make a lot of interesting engines work without custom play rules. For example, in the original thread, it was suggested that maybe KC/Goons/Masquerade would do well, but I'm pretty sure this is impossible to code in Geronimoo's simulator. (Not that I'm claiming this would actually do well. It's just an example.)

So I'd suggest maybe having a parallel tournament in Dominate, so that entrants can write for whichever simulator suits best. Dominate is missing some cards, and its default play rules are worse than Geronimoo's, but it is very customizable and should support all sorts of strategies.

(I'm biased here because I suspect that rush strategies will come out on top if custom play rules aren't possible, and I'm pretty bad at rushes.)

[Young Nick]

I will not be competing, but I must admit I am confused by Possession being prohibited. Sure, it is probably not useful in most solutions, but if someone can play it reliably (I'm thinking Golem, Scheme, and/or KC as enablers) then what's the big deal? If it slows down simulations a ton, I understand, but otherwise, there seems no reason to exclude it.

[blueblimp]

On second thought I'm not sure whether I'm up to tweaking Dominate play rules enough to actually do really crazy stuff there. So maybe ignore my suggestion of a parallel Dominate tournament.

[DStu]

Submissions should be sent to me via PM no later than midnight GMT on the night of 21st-22nd, 2012.

Nice you specify the year, but wouldn't the month be more important first
/scnr

Serious question: May/Should we assume Colonies?

[DG]

The bishop might be worth prohibiting as well. It again derails many strategies when it offers free trashing to the opponent.

[WanderingWinder]

blueblimp: I'm aware that custom play rules will make for much better results. Though a rush strategy, I assure you, will be required to win - you're just going to be rushing different things
Young Nick: the problem is not that that can't be strong or somehow isn't fair, it's that... great, so now I've possessed your deck. The problem is, I don't know what kind of deck you're running. So I'll do more or less exactly what you would do, which sometimes is great, but other times is terrible. For example, you're playing some wharf-based deck. This is pretty good for me, I buy lots of the colonies that you're going for. On the other hand, you're playing a goons-based deck. This is terrible, I'm basically giving you an extra turn of attacking me and accumulating VP chips. So, it does really wacky things based on matchups, having very little bearing in lots of cases on what would actually happen in a game.
DStu: Thanks for catching that, whoops. Secondly, yes, colonies always in play (part of every card) as implied here:

You only get to reference up to 5 different kingdom cards (plus whatever basics, colony, platinum, and potion)

DG: Bishop is gone, thanks.

[DStu]

as implied here:

Upps, sorry.

[Geronimoo]

This is cool... But please ban Ironworks as well...

[DStu]

This is cool... But please ban Ironworks as well...

?

[Geronimoo]

Ironworks is too much work and it's not defined as a terminal action, so Crossroads gets played after the Ironworks

[theory]

I would think Ironworks is a pretty important card here, though ...

[blueblimp]

Gotta keep Ironworks I'd think, but I empathize with difficulty getting turns to play correctly.

[DG]

Embargo might be worth prohibiting since it will produce unpredictable results. The effects will be purely determined by the opposition script : "will look through buy rules of opponents until it finds a card that is not yet Embargoed and isn’t present in its own buy rules, then puts an Embargo token on that card’s pile."

[blueblimp]

Embargo might be worth prohibiting since it will produce unpredictable results. The effects will be purely determined by the opposition script : "will look through buy rules of opponents until it finds a card that is not yet Embargoed and isn’t present in its own buy rules, then puts an Embargo token on that card’s pile."

Why is that a reason to ban Embargo? That's a reasonable approximation of how a human would play it.

[DG]

For one thing, it gets very difficult to make purchasing conditions when you don't know how many + and - points you get for buying a card. It also weighs very heavily in favour of card gaining and alternative vps.

[Geronimoo]

If you have Embargo in your buy rules, that means you'll only have 4 cards to build your deck, so I doubt the Embargo is going to be worth it very often.

[WanderingWinder]

On the other hand, in a real game, you can switch up your strategy to not buy those embargoed cards as often. Or not, if the curses aren't so big a deal. Also, if you're opening embargo, which is the thing that would make sense here, you may not know what your opponent is going for/be able to embargo the right thing. So I'm inclined to get rid of it. But as you say, Geronimoo, I doubt it's going to be important. So, I guess we'll go with banning it, just so people don't have to worry about it. But if somebody tries to tell me that embargo is critical to their strategy... by all means.
The ironworks thing... seriously Geronimoo? Can't get rid of ironworks. It's too important in too many different strategies. If you don't like the play priorities, you have to work to get around it. That it's in the simulator is part of the challenge, and the part that means this won't really answer the question in the thread I reference in the OP. Wait a a minute... Geronimoo? If YOU don't like the play priorities, change them!

[blueblimp]

My argument for retaining Embargo is that if you remove every card that an opponent might play sub-optimally against, you're left with almost nothing. (For example, in an engine, you have to decide on which type of village you want. If your opponent is doing some kind of Vineyards strategy and decides to mass the same type of village you chose, you're doomed. Does this mean villages should be banned?)

And furthermore, if it's usually bad, well then all the better if someone finds a way to use it well. I'm not currently using it, but earlier I found it useful for one of my bots.

Also, the worst possible damage that could be caused by Embargo is that you get some curses you weren't planning on. The same thing could happen if your opponent buys an IGG.

[Geronimoo]

Does it matter if I beat a bot by 2% or by 20%?

[Geronimoo]

While testing the mirror match of one of my strongest bots, the game would very often end up with a pin where one player gets his hand reduced to 0 and 5 cards of his deck gone each turn. Totally not meant to do it, but still cool to see this happen.

[WanderingWinder]

Does it matter if I beat a bot by 2% or by 20%?

Nope. Just win, baby.

[DG]

I think my scripts will end up more complicated than WW's entries for the pairs, and that's even after making some very broad generalizations. There are just too many conditions to cover. Some suggestions for improving condition clauses will be going Geronimoo's way!