AI improvements

[PuppyChow]

Puppy did what I did before I edited my post. He treated it as a 6/6 creature and a 1/1 creature, and didn't add the 5/5 to the 1/1.

This. I thought it was just a skeleton or something, not a lycan waiting to grow .

Here was my logic:

-You snipe the 1/1 and end your turn.
-Opponent ends turn (total damage: 6)
-You snipe the lycan once and end your turn.
-Opponent ends turn (total damage: 6)
-You kill the lycan.
Total Damage: 12.

-You snipe the 6/6 and end your turn.
-Opponent ends turn (total damage: 1+6=7)
-You snipe & kill the 6/6 and end your turn.
-Opponent ends turn (total damage: 1)
-You kill the skeleton.
Total Damage: 8.

So yeah my logic was wrong since I didn't know the 1/1 was another lycan .

[alienorigin]

Yay for miscommunication where I forgot to mention that it was a lycan 1|1...oops. 

Sorry about that. 

[coinich]

Yeah, my numbers took that into account because thats what happened in most of the situations I found myself in.

[Bloodshadow]

This is not something that already happened, but rather something that could happen.

If you have multiple Floodings, they don't do anything more than one single Flooding, right? But if you have more than one Flooding per turn, do their quanta-draining effects stack?

If they do, the AI is probably not smart enough to not play multiple Floodings.

[killsdazombies]

its alot like night fall, the Ai will play about 6 of them and then have to wait another turn to use its dragon.

[Dreadshadow]

I am not sure if the things i am gonna say were posted before by others, so i am sorry if i repeat something somebody else said.

1) AI has Unstable Gas casted. Cool!
AI casts a phoenix and THEN detonates Unstable Gas. LOL!
Ash stays as it is for 5 rounds, while AI has 45 or more  . LMAO!


Gravity Nymph is down (i casted it). An Elite Otyugh is also casted.
AI has Owl's Eye, and targets Elite Otyugh twice, instead of the Nymph.

[unionruler]

I have many rays of light. Rainbow casts gravity pull on a ray of light. Rainbow congeals the same ray of light.
I put out an electrocutor. Rainbow does not snipe for the turn, but instead steals my electrocutor and replaces it instantaneously with another eagle's eye.
I put out a light dragon. Firefly Queen snipes it with eagle's eye. I fractal the light dragon and put out a few more. Firefly Queen turns its attention to another light dragon, ignoring the one already damaged.

[Kurohami]

I'd say we should at least have some intelligent advantage over the AI, because the half-gods and false gods have many other advantage over us. If Rainbow learns how to target wisely, our chance of winning is pretty nil, If obliterator's protected pulvy no longer targets bonewall over other stuff, then he is basically unbeatable from the point he got his pulvy protected. I would appreciate if the AI stays the way they are now.

[Xinef]

Kurohami, it's obvious that when AI gets improved, some of the advantages AI has will be removed, or some other solution will be found to give players more advantage, so that fighting FG still makes sense.

I think chess playing algorithms could be a good example.
Imagine the time when programmers were just beginning to write chess algorithms and computers were still slow and had little memory.
Any chess player with a bit of experience could easily beat AI, because of their 'stupidity'.
So, to make chess games against AI more fun, let's allow AI to make two moves every time you move once, or give him more figures, or allow him to 'respawn' figures you take down with 50% probability... or something like that.
Now playing against AI is fun once again... but does that solve the problem? No!

The better solution is to improve AI, so that it can play as good as a living person. This happened in chess, so that now AI can beat the best chess players even without 'fake advantages'. You can even adjust how 'intelligent' AI is through changing some variables in it's algorithm, to make easier AIs that are fun to play against by inexperienced players.

Obviously it's a bit different with a collectible card game. First of all making a really good AI is really hard and takes time. Second thing is that False Gods having some advantages makes sense due to story settings (they are not elementals, but more powerful creatures, something similar to bosses in computer games).

My conclusion is that there should be some AI level where you can play an equal match (AI playing like a real living person, with no advantages like 200 hp, but with a reasonable artificial intelligence), and some uber-AI-level where not only he plays with intelligence, but also some advantages, so that experienced players can have some kind of challange similar to fighting bosses, but there could be some way to improve your chances so that these experienced players can actually win... it could be anything - from knowing your opponent beforehand, so that you can construct a couterdeck, to ability to fight multiple players against a single uber-AI... I guess time will show what Zanz can think of, to solve this problem, but clearly leaving AI as dumb as it is right now is not the best solution. (IMHO )

[unionruler]

The better solution is to improve AI, so that it can play as good as a living person. This happened in chess, so that now AI can beat the best chess players even without 'fake advantages'.

Well chess engines can beat chess players even with disadvantages these days.
The problem is that with one centralized AI obviously it's rather impossible to program it to recognize some strategies conceived by human players. With chess engines they use opening books and they can assign values to make the computer more aggressive/positional etc but that is only because no matter what they start with the same chess pieces. In elements the AI has to play all sorts of decks.

I think the AI should not be able to play as good as a living person as you say, for that matter it will be some time before we can program it to do as such. Maybe there should be a different AI to play FG decks so that it can recognize some basic elements of anti-FG strategy, or better still create a whole new level 6 with a more intelligent AI and perhaps less fake advantages.

[Xinef]

Yes, I agree that improving the AI would need some changes like new AI levels, and Elements is much more complex than chess to program a good AI... but it's still possible. An artificial neural network could simply learn the deck and learn the decks opponent is using. On the other hand, in mini-max algorithm approach it's possible to evaluate probabilities too, so that you evaluate what cards your opponent might have based on his mark/pillars/some database of decks... and with those probabilities act accordingly, AI could also evaluate random effects like mutation, to check how much it would help him on average... and he could evaluate the probability of drawing a certain card next based on the cards he has already drawn and his knowledge of cards remaining... though it adds up to a huge task to calculate, so I wonder how much calculations would be reasonably possible to do without slowing the game... I'm not a flash programmer (I prefer Java), so I don't know if flash can calculate AI at a reasonable speed, but I guess it could be a problem... at least if it is as slow as it is with graphical effects.

[Bacchus]

I was playing a lvl 5 and he used solar shield. de round after that he used holy light so the shield blocked the spell and casted it forward me. i did heal 10 hp. he did that 3 times