A few details about T500

[Atico]

12 dim shields... that's 36 turns... that's not fair...

Why would you not bring at least a little perm control in to gold or platinum?

Believe or not - I saw in Arena few decks with 12 Dim Shields and 12 Protect Artifact It doesn't work very good with AI, but it shows that in Arena we have a lot of annoying decks.

What is more - in my opinion it isn't so good that we have a lot of SoG, Dim Shields etc in Arena games. Only few elements has got PC, so we must play Fire/Dark in gold/platinum...?  It's not fair in compare with other elements... And remember, that You can have 6 Explosions - Your Opponent can have 12 each permament.

Idea of Arena is very very good, but something goes wrong (we can see less plays every day...)

[Falcon4415]

Arena decks are chaining BHs on me. Not only that, they also use multiple BHs per turn. I WANTS NURRRF ON BH.

[Essence]

The nerf is already in place.  The way it works is this: every time past the first that a BH is cast, the AI gains a 'guilt' point.  Each guilt point makes the AI less likely to cast BH again.  But they still might.  It's just less likely.

If an entire turn goes by with no BH being cast, all guilt points are erased.

Same rules apply to Earthquake, Silence, and I think one other spell, but I don't remember what.

[BluePriest]

The nerf is already in place.  The way it works is this: every time past the first that a BH is cast, the AI gains a 'guilt' point.  Each guilt point makes the AI less likely to cast BH again.  But they still might.  It's just less likely.

If an entire turn goes by with no BH being cast, all guilt points are erased.

Same rules apply to Earthquake, Silence, and I think one other spell, but I don't remember what.

Nightmare?

This sounds like a good system to me. Im glad Zanz didnt outright stop them from chaining them.

[petersenk]

yeah right.

First: So our AI sucks, so let the AI cheat like there is no tomorrow. A fair game? Not here.
Second: let the AI cheat even more. More Bonuses for the AI!
Third: uhm, dang. Did we cheat too much? NO WAY! Just let's make the AI even MORE STUPID. Let's make the AI show some mercy, right? And then add some more unfair bonuses for the AI.
...

oh boy. The arena shows, what a trainwreck this game has become by now.
Just stop it already.

[Essence]

Insert obsenity-laden-epithet filled rant here.

[YoungSot]

Insert obsenity-laden-epithet filled rant here.

This sort of thing is not acceptable here in the Elements community. Please keep your posts polite and remember that we have younger players on these forums.

[petersenk]

Insert obsenity-laden-epithet filled rant here.

Yeah, sorry for that.


Nevertheless I really think it's a problem when AI is only a challenge, when it's granted a tons of bonuses and other cheap cheats and tricks. Sure. Programming AI is hard. But the latest development with the arena clearly shows the problems with this "cheap" approach.

[Hodari]

Nevertheless I really think it's a problem when AI is only a challenge, when it's granted a tons of bonuses and other cheap cheats and tricks. Sure. Programming AI is hard. But the latest development with the arena clearly shows the problems with this "cheap" approach.

I imagine that if you are willing to write the code to make the AI play flawlessly and to donate it for free and maintain it etc, Zanz would consider it...

  Right now, the AI seems to work by applying a few very simple rules to determine whether or  not to play a card at any time and if so, what to target with it.  These are usually as simple as (If hp <60 OR expected_hp after opponents next turn < 10, play Miracle if able to) or choosing the best creature to target (attack + ability_modifier).  I'm sure a few more simple rules like this could be added to improve things slightly but to get the level of AI you seem to want would take far more than this.

  Basically you'd have to have something that worked in the same way as a chess engine where it evaluates EVERY possible move and what the results would be and then every possible next move after that, evaluating out this tree of possibilities until either the game would end or out to some arbitrary depth and then calculating the probability of winning and what the best move to make would be based on that.  Except in the case of Elements, this would be FAR more complicated than it would for chess. 

  First of all, there are far more possible moves at every step to consider..in chess there are likely to be only a couple dozen moves available on any turn, many of which could quickly be discarded as obviously bad.  For elements, you'd have to look at each card in your hand one by one, considering every possible target for it(just for one spell, this could easily give as many possible moves as an entire turn in chess would).  Then there's the fact that in chess, each turn consists of only one move per player, while in Elements, there could be a dozen or more moves(cards played, creature abilities used, more cards drawn via hourglasses etc) in a single turn.  Chess only has half a dozen different types of pieces to consider while Elements would have hundreds..and the interactions between them is far more complicated.  Last but certainly not least, in chess the entire position and in theory at least, it is always possible to calculate every move and response out to the end of the game and get an exact answer to which move is best.  In elements, the AI won't(or at least shouldn't, unless you want to substitute one "unfair" advantage for another) even know what all the cards in your deck are, let alone which cards either player will draw next or any other of the many random events in the game.  In fact, currently the AI doesn't even really seem to know what's in it's OWN deck.  Its decisions are limited to the cards currently in play and its hand.  However, changing it so it could understand its own deck and how to take full advantage of it would again require a fairly complicated analysis of how dozens of different cards interact.  Oh yeah, and finally the rules and pieces for chess have been relatively constant for hundreds of years, compared to Elements having significant new stuff added every few months(with corresponding need to adjust the AI accordingly).

  Yet even the comparatively "trivial" task of making a really good chess engine took decades of work by many of the top programmers in the world.  It would hardly be fair to expect Zanz alone to be able to write anything remotely comparable for Elements by himself, especially for free.  So obviously, there is no way the AI can be nearly as good or expected to play anything close to the level a human would.  At best, a few of the rules for what to target when could be improved slightly.  Personally I'd be happy if it was as simple as slightly better targeting on Fractal(knowing when to keep going for the small creatures and when it can afford to use a dragon instead, or Decay not targetting my RoLs and ruining its whole deck's strategy in the process) or Lionheart not drawing through his entire deck in 5 turns and decking himself out because he also made all his creatures untargetable and has nothing left to rewind.

Sorry this is a bit long, but before complaining about the AI too much, it seems only fair to think about what would actually be required to make it much better.  I don't see any practical way a truly "intelligent" AI could be made for a game this complicated, so odds are making it challenging will always require giving it some other "unfair" advantages.

[petersenk]

I imagine that if you are willing to write the code to make the AI play flawlessly and to donate it for free and maintain it etc, Zanz would consider it...

Last time I checked, Elements wasn't Open Source. Thus, I'm afraid, we're out of luck.

And sure, AI is hard. It's the hardest part of any game. But this doesn't mean that the situation is hopeless. Nobody needs perfect solutions, good approximations are fine too.

How about this for some input (at least): let there be an interface to express AI-rules, exactly like you described them, with simple checks of certain conditions. Then, let the user define such rules for his arena-decks.

I'm thinking about rules like: don't play my creature X, unless we can immediately put card A and card B on it.
If we have to discard a card, do it in this order.
Target creatures in that order.
Use my protect articfact for this and this only.
And so on.

I don't think it would be too hard to come up with a generic interface to express such rules for any cards and lots of circumstances. Then it's up to us, to write groovie rules for our personal arena decks.

And maybe, just maybe, the AI wouldn't need 12 explosions anymore to be some sort of opponent.

The point here is, that it's really hard to write a rule-set for all decks imaginable. But a good rule set for a distinct fine deck isnt that hard after all.

[Hodari]

How about this for some input (at least): let there be an interface to express AI-rules, exactly like you described them, with simple checks of certain conditions. Then, let the user define such rules for his arena-decks.
The point here is, that it's really hard to write a rule-set for all decks imaginable. But a good rule set for a distinct fine deck isnt that hard after all.

That certainly would be possible(though probably not nearly as easy as you think) and would be a definite improvement.  The number of rules would probably have to be fairly limited and with the default rules used if you don't have a custom one for a given situation but even one or two rules could make a difference with some decks.  I also imagine something like this would have to be done VERY carefully to avoid the chance of it being exploited in some way though.

[BluePriest]

People seem to forget elements only has 1 developer....