G/W/r Agro Zoo

[gohongohon]

well i can tell you that the easiest way to make a deck very hard to play against and easy to play with is the land base. So if you wanted to make the modes different without changing the decks for every mode then i would just mess with the land base. I would make all the very hard mode decks all duel land out and i would make all the easy decks with just basic lands. Sadly I'm not a programmer so i can not help you there but i can build decks if you want and i could make the decks scale with every mode if you want.

"Easy" Basic Green White Agro Deck

26 Creatures:
4 Plated Sliver
4 Elvish Archers
4 Elvish Warrior
4 Muscle Sliver
4 Wild Mongrel
4 Sinew Sliver
2 Avian Changeling

10 Spells:
4 Elvish Fury
4 Giant Growth
2 Oblivion Ring

24 Lands:
12 Forest
12 Plains

Im not sure how well they will play this deck but i think it is a very good agro deck but i put flaws in on it if you dont draw a lot of slivers you will tend to lose to big bodies and if they have a lot of removal you can pick and choose want to do to crush it. Also there is no card advantage and really it is playing to many lands so it could get land flooded and it does not have a lot of removal. But it is also a deck you cant sleep on because really if it drops slivers first 3-4 turns you could be in a lot of trouble.

"Medium" G/W Agro

24 Creatures:
4 Savannah Lions
4 Plated Sliver
4 Muscle Sliver
4 Wild Mongrel
4 Sinew Sliver
4 Avian Changeling

14 Spells:
4 Elvish Fury
4 Giant Growth
4 Oblivion Ring
2 Glorious Anthem

22 Lands:
4 Treetop Villagers
8 Forest
10 Plains

For this deck i added more sliver types and added a great 1 drop in Loins to make the deck faster i also lowered the double green cost as it could lead to bad things. I added Manlands because they are hard to kill with board sweepers up the removal count and added a great 3 drop for the deck in anthem

"Hard" G/W/r Agro

24 Creatures:
4 Savannah Lions
4 Akrasan Squire
4 Muscle Sliver
4 Wild Mongrel
4 Sinew Sliver
4 Isamaru, Hound of Konda

14 Spells:
2 Lightning Helix
4 Giant Growth
4 Oblivion Ring
4 Glorious Anthem

22 Lands:
4 Treetop Villagers
6 Forest
6 Plains
4 Scared Foundry
2 Stomping Grounds

Here we are moving away form the slivers and starting to add even more 1 drops we also added Helix because it is just a great weapon and we changed the land base to support that but not all the way. We uped the Anthems for a little more late game. But really if this deck gets to the late game it really should not have much of a chance but we gave it a little on with the helix

"Very Hard" G/W/r (or Zoo) Agro

28 Creatures:
4 Savannah Lions
4 Akrasan Squire
4 Kird Ape
4 Watchwolf
4 Wild Nacatl
4 Isamaru, Hound of Konda
4 Sigiled Paladin

10 Spells:
4 Lightning Helix
4 Tribal Flames
2 Char
4 Lightning Bolt

18 Lands:
4 Taiga
4 Savannah
2 Plateau
4 Scared Foundry
4 Stomping Grounds

Now i have droped all the slivers and this deck is just plan scary. it has great reach in the burn. Really low Curve soild mana base (though there is only 18 lands but really only need 2 lands to play most of the deck) The thing is in quest mode there is a really good chance this deck is way to good for it. If we do decks like this though we can keep all the pools the same because the decks just get harder and you are going to need more cards.

lol that took a long time tell me what you think and im sure i can come up with every kinda deck in this kinda mode. Also we could go with the idea that the higher lvl you get the harder your opp's are and i could put maybe lvls with all of them just up to you let me know what you would like me to do.

Yeah this is just the decks i did on the other forums but posted here because i really dont know how to up loaded them and stuff.

[juzamjedi]

Knight of the Reliquary is a powerful Spike card.
Tarmogoyf is another powerful Spike card.
Woolly Thoctar is the poster child for Naya.
Path to Exile is another one.
Windswept Heath / Wooded Foothills often goes in decks like this one (especially true for dual land / Tarmogoyf / Knight of the Reliquary synergies)
For this type of deck 2-3 x Ranger of Eos is a good choice.
Mogg Maniac is worth consideration also.

[gohongohon]

As for the cards you are talking about Knight im not playing because no Onslaught lands.
Tarmogofy is fine but it does not really shine without FoW (which is why he is played in those decks)
Thoctar is too slow
Really all 3 drops are too slow
Ranger is also too slow we want to kill by turn 4 or 5
Path to Exile only deals with creatures and cant burn

[GandoTheBard]

As for the cards you are talking about Knight im not playing because no Onslaught lands.
Tarmogofy is fine but it does not really shine without FoW (which is why he is played in those decks)
Thoctar is too slow
Really all 3 drops are too slow
Ranger is also too slow we want to kill by turn 4 or 5
Path to Exile only deals with creatures and cant burn

Not sure I entirely agree with these assumptions. That is that the AI will win by turn 4-5 with the current listed decks. If it is possible, well thats another story but the way the AI plays is rather backwards so don't expect quick kills. Better to build around the AI's capabilities than build a deck that a human would kill with and hope the AI catches on.

[juzamjedi]

If you do not have the Onslaught lands yet you could go with Terramorphic Expanse.
The best change in my mind would be adding the Tarmogoyf to the top level deck. There are very few decks I have seen that wouldn't like to have Tarmogoyf for some cheap fat.

No worries either way; just trying to help.

[DennisBergkamp]

I could add the onslaught lands... Should be easy to code, but then again, the AI won't benefit that much from these, since after fetching a land, the shuffle will get rid of the land stack.

[GandoTheBard]

how bout a restacking routine that checks how much mana the AI has in play already and then stacks the deficit over a shuffled deck. IE the AI has 3 mountains and 1 plains in play...in theory he wants at least 7 lands and 2 of each color so the routine counts the lands in play on his side...searches for more lands and assigns ones that fit the criteria to the stacked portion then shuffles the rest back into the deck underneath the stacked portion. This would make the AI a super cheater but it would solve the problem of accelerated mana searchers like Sakura-Tribe Elder.

[Chris H.]

@ gohongohon:

The computer uses a system called land stacking and as such the mana system for the computer is different than for the human player. This is true if you have the land stacking option checked. If the land stacking is not checked, the mana will be randomly shuffled into the library like we would normally expect.

With land stacking set to on, the computer will put seven lands into play and the rest of the lands are placed at the bottom of the library. Some have noticed in the past situations where the computer played six lands rather than seven.



You may want to consider replacing the Treetop Villager cards until they get repaired.

Isamaru, Hound of Konda is a Legendary creature and you may want to consider having no more than two in the deck.

The AI code for Lightning Helix, Tribal Flames, Char and Lightning Bolt is somewhat weak at this time. The computer tends to use these spells only on flying creature that it can kill with the damage from just a single spell.

I have not checked the source code for Swords to Plowshares, but the computer may be able to kill more creatures if you use this spell instead. The computer also might be able to make better use of the Crib Swap spell.

I think that Tribal Flames requires just Basic Lands. If so, you can replace the Sacred Foundry and the Stomping Ground with Plains, Mountain and Forest.

With the land stacking flag checked, you will see the computer play it's lands in this order:

1st Taiga, 2nd Plateau, 3rd Forest, 4th Mountain, 5th Plains, 6th Savannah, 7th Taiga



I like your decks and with just a wee bit of adjustment you should be able to account for the land staking system and also how well the computer can or can nor use certain cards.

At some point the in the future, the coders may give us the ability to, say, divide all of the computer's AI tuned decks into 2 groups. An easy group and a hard group. We may be getting close to having enough decks to do this.

Some thought has gone into having as many as 5 different pools of computer AI tuned decks. If we did it at this time we would find that we only have 5 or so decks for each of these 5 different pools.

[GandoTheBard]

Re the legend thing Chris. It is not always a good idea to use 1/2 ofs when you really want to draw a particular card...so Id not force people to a rule of 1/2s without a good reason such as the deck having good fetch. In the case of Isamaru ...well he isnt all that great. Mainly just a slightly tougher Savannah Lions.

[gohongohon]

yeah really the burn is not to kill creatures it is to burn the face unless say you got T1 1 drop T2 1 drop 1 drop T3 1 drop and burn body that is game really. I guess the thing is i dont understand what the computer can do and can not do so i guess if i knew then i could build better decks for it. I guess if there is land weaving this deck does not want 7 lands lol it wants like 3 at the most.

As for crib swip you dont want because it is giving them a blocker when really what is the point anything they have down by T3 we are out classing. And Swords is giving them life which is something we dont want if we are killing fast. As far as the goyf i dont mind you adding it but with out the fetch lands you really need a lot thing in there to give it power 3. Instant Sorcery and Creature. Once we get fetchlands then we add goyf in there for sure because it is easy to get 3 power out of it. This is not my ideal list there is just cards we dont have like jitte chain lightning stuff like that. But with land stacking then this deck is just not that good if he is going to have or draw 7 lands.

[Chris H.]

yeah really the burn is not to kill creatures it is to burn the face unless say you got T1 1 drop T2 1 drop 1 drop T3 1 drop and burn body that is game really. I guess the thing is i dont understand what the computer can do and can not do so i guess if i knew then i could build better decks for it. I guess if there is land weaving this deck does not want 7 lands lol it wants like 3 at the most.

As for crib swip you dont want because it is giving them a blocker when really what is the point anything they have down by T3 we are out classing. And Swords is giving them life which is something we dont want if we are killing fast. As far as the goyf i dont mind you adding it but with out the fetch lands you really need a lot thing in there to give it power 3. Instant Sorcery and Creature. Once we get fetchlands then we add goyf in there for sure because it is easy to get 3 power out of it. This is not my ideal list there is just cards we dont have like jitte chain lightning stuff like that. But with land stacking then this deck is just not that good if he is going to have or draw 7 lands.

I think that the computer will only use a single burn spell against the human if it will do enough damage by itself to win the game. Otherwise it holds onto the spell waiting for a flying creature that it can kill.

A couple of months ago i noticed that the computer was discarding burn spells at the end of turn. A change in the coding at that time now allows the computer to target the human if the spell will win the game or if the burn spell would have to be discarded.

I think that the programmers will likely improve the computers artificial intelligence with the burn spells at some point in the future. I have seen some improvement in the recent past in this area.

Writing the code so a human can use a card can be fairly complicated. Writing the code so the computer can use a card to great effect can be substantially more complicated.

I don't have great skills in designing or balancing decks for the human player. You and Gando can see the subtleties that I miss. I have spent some time recently creating some of the AI decks that you may have seen as part of the 100 deck archive. I am willing to pass along the knowledge that I gained about building AI decks.

[gohongohon]

yeah i see your point because im building these decks on how me (human lol) would want to crush people. Even if it does not point at the head with the burn spells aslong as they are pointing at blockers it is not too bad. Is there a way to have the AI just add up all the burn in there hand and see if it is more when there life and win? How do i test with the decks in the AI hands?

[Chris H.]

yeah i see your point because im building these decks on how me (human lol) would want to crush people. Even if it does not point at the head with the burn spells aslong as they are pointing at blockers it is not too bad. Is there a way to have the AI just add up all the burn in there hand and see if it is more when there life and win? How do i test with the decks in the AI hands?

At first, Gando and I would just design a deck for the computer and then let the computer play the deck against us and see what happens. We gained some experience just by letting the computer try out different cards. You can put together a small test deck with just limited set of cards to see if the computer will use an ability of a card.

In one test I discovered that the computer can use the Royal Assassin to good effect but it would not use the Squall Drifter to tap one of my creatures to set it up for the kill. If you have an idea for a card combo like this you can set up a test deck and try it out.

I have been looking at the source code and I am starting to recognize some of the code. The file CardFactory.java can help you determine if the computer can use a card. Below is the code for Squall Drifter:

Code: Select all

 if(cardName.equals("Squall Drifter"))
    {
      final SpellAbility ability = new Ability_Tap(card, "W")
      {
        public void resolve()
        {
          Card c = getTargetCard();
          c.tap();
        }
        public boolean canPlayAI() {return false;}
      };//SpellAbility
      card.addSpellAbility(ability);
      ability.setDescription("W, tap: Tap target creature.");

      ability.setBeforePayMana(CardFactoryUtil.input_targetCreature(ability));

The line which states "public boolean canPlayAI() {return false;}" says in essence the computer can not use the special ability for this card.

Gando and I can also pass along our knowledge. You can always ask us if we have any knowledge in reference to certain cards that you are planning to use in an AI deck.

It is also important to remember that the Forge project continues to improve over time. People are working on it. In upcoming weeks we will see some bug fixes and new cards with Flanking, Protection from Color and Hybrid mana.

You can if your interested, look at the AI decks that are included in the quest mode. The cards included in these decks work fairly well at this time.

The programmers will improve the AI use of the burn spells in time. At this point, 16 burn spells in an AI deck may be too many. You could import this deck into the constructed deck format and then play against it.

[gohongohon]

you are right about the burn and i also noticed that there attacking with exlated was wrong like they had 2 squires and would attack with both and not just one. Also im still having a problem with the wild's not being 3/3s with the 3 types out so im not sure on that one there.

[Chris H.]

you are right about the burn and i also noticed that there attacking with exlated was wrong like they had 2 squires and would attack with both and not just one. Also im still having a problem with the wild's not being 3/3s with the 3 types out so im not sure on that one there.

Dennis has made some improvements to the AI exalted code. It is getting better.

I ran a test with the Wild Nacatl and came across a different problem. It is a recent addition. I will report my findings over on the bug board. Keep my fingers crossed.