It is currently 13 Dec 2018, 18:48
   
Text Size

Tempest world development

Post MTG Forge Related Programming Questions Here

Moderators: timmermac, friarsol, Blacksmith, KrazyTheFox, Agetian, CCGHQ Admins

Re: Tempest world development

Postby Seravy » 03 Sep 2017, 12:32

Two more important cases :
Creatures owned by AI but controlled by someone else.
Creatures owned by someone else but controlled by AI.
Unfortunately these cases also stack with each other, like auras and ETB effects.
I'll try to think about priorities tomorrow if I have more time.

(creatures with 'sac this if targeted" ability should probably be considered somewhere as well but that's a generic targeting issue not specific for returning effects)
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 03 Sep 2017, 20:51

Meanwhile gave another test run to the spike deck, but Spike Warrior crashes the game so can't do it afterall. One of your changes to AI pump abilities seems to have broken something, it gives a Type Mismatch error (posted the bug report in the snapshot thread.

Next up on the list, Hermit Druid+Living Death deck :
| Open
2 Forest|TMP|1
3 Forest|TMP|2
3 Forest|TMP|3
3 Forest|TMP|4
4 Hermit Druid|STH
4 Living Death|TMP
1 Mirri, Cat Warrior|EXO
2 Mongrel Pack|TMP
4 Pit Spawn|EXO
4 Recurring Nightmare|EXO
1 Spike Feeder|STH
4 Spike Weaver|EXO
4 Survival of the Fittest|EXO
3 Swamp|TMP|1
1 Swamp|TMP|2
5 Swamp|TMP|3
2 Swamp|TMP|4
4 Verdant Force|TMP
4 Wall of Blossoms|STH
2 Wood Elves|EXO


...So, let's see the problems :
The AI didn't activate Survival of the Fittest when it had 3 copies of Pit Spawn in hand.
The AI also didn't activate Recurring Nightmare, not even if it had a Verdant Force in the grave and Wall of Blossoms in hand. It also didn't sac Hermit Druid to return a Pit Spawn or Spike Weaver from the grave. (the former actually making sense because it didn't have 2 swamps to maintain it but I'm pretty sure the AI wouldn't know about that)
So both of these cards seem to be unplayable. One more deck goes into the freezer :D

On the bright side, Living Death works perfectly, and Spike Weaver as well (don't think it leaves mana untapped intentionally for it though, haven't really tested that)
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Agetian » 04 Sep 2017, 04:13

I believe I fixed the crash issue upstream :) Interestingly, it didn't crash for me in the first place in that scenario, so I can't confirm that it's fixed, please take a look :)

Recurring Nightmare: implemented a simple AI logic where the AI would sac a card with lesser CMC to return a card with greater CMC that evaluates to a higher score, currently about the only way I can see this working more or less reliably. :)

Maniacal Rage: promoted it to RemRandomDeck, the AI can use it already (and will always use it), it's just that it'll be trigger-happy with it, so it's only good for aggro decks and stuff.

I'm willing to look into improving Survival of the Fittest for the AI. It's a staple, so it'd be nice to get it going for the AI. What kind of default AI behavior do you see for it?

- Agetian
Agetian
Programmer
 
Posts: 3296
Joined: 14 Mar 2011, 05:58
Has thanked: 632 times
Been thanked: 485 times

Re: Tempest world development

Postby Seravy » 04 Sep 2017, 10:04

Agetian wrote:I'm willing to look into improving Survival of the Fittest for the AI. It's a staple, so it'd be nice to get it going for the AI. What kind of default AI behavior do you see for it?

- Agetian
Let's see...

1.If Living Death or equivalent (Twilight's Call, All-Hallow's Eve comes to mind) is in hand or the deck plays a high number of copies, always use to fill the grave and raise chance of drawing those.

2.If a creature too expensive to play is in hand, discard it to get something that can be played.

3.If it's possible to discard a creature to get something significantly more valuable and both can be played, do it as well.

4.Once we have "play priorities" for spells, this (and all other tutor effects) could search for the creature card that has the highest such priority, for now, get the highest "value" creature, however that is calculated. (it could also be coded to discard cards with a play priority below a threshold to allow replacing useless creatures)

5.For non-Tempest purposes, always search for Squee, Goblin Nabob if not in hand or grave yet, and always discard that and nothing else if it is.

6.Furthermore, discard anything with a "can be played from gave" or "while it's in grave" effect.

I think that's about everything I can think of.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Agetian » 04 Sep 2017, 10:40

Nice! I'll work on these.

- Agetian
Agetian
Programmer
 
Posts: 3296
Joined: 14 Mar 2011, 05:58
Has thanked: 632 times
Been thanked: 485 times

Re: Tempest world development

Postby Seravy » 04 Sep 2017, 16:25

One more idea :
7. If the card has Madness, is good to play, and the cost can be paid after tapping the G for the search, discard and play it.

And as a generic rule, any discard cost should be paid (or discard choice made for effects like Anvil of Bogardan) as 1,2,5, 6 and 7. (ie, discard squee, grave cards and madness cards, stuff for reanimation if such is available, and cards with an unpayable cost.)
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 10 Sep 2017, 11:33

Sorry, I haven't had much time to work on this the past few days. I'll continue as soon as I can.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 10 Sep 2017, 17:58

Meanwhile, you asked for how to pick a target for bounce effects. I tried to include everything that can be relevant, but I'm probably still forgetting some corner cases.

// Note that targets totalling less than zero priority will not get picked, thus it avoids unfavorable plays.
Bestsofar=0;Chosentarget=-1
For each target {
Priority=0
If Owned by enemy but controlled by AI : Invalid target, skip
If controlled by enemy and enemy can pay X where X is at least as much as the available Erratic Portal activations : Invalid target, skip
+100 If Owned by AI but controlled by enemy
+CMC-2 If Controlled by enemy
+3*Power if controlled by enemy, is attacking and unblocked, +1000 if power>=AI's life remaining
+20 if controlled by enemy and costs more than that enemy can afford after untapping all their permenants.
+5*CMC If Controlled by AI and is about to die to combat damage without killing its blockers/creatures it is blocking, or if it is about to die to a spell or ability
+10 If controlled by AI and has beneficial "comes into play" effect
-15 If controlled by AI and has detrimental "leaves play" effect
+15 If controlled by enemy and has detrimental "leaves play" effect
+5 If controlled by AI and has beneficial "leaves play" effect
-10 If controlled by enemy and has beneficial "leaves play" effect
-25 If controlled by AI and has detrimental "comes into play" effect
-20 If controlled by enemy and has beneficial "comes into play" effect
+20 If controlled by enemy and has detrimental "comes into play" effect
+3*CMC of Auras attached to permanent controlled by enemy
+Equip cost of Equipments attached to permanent controlled by enemy
-Equip cost of Equipments attached to permanent controlled by AI
If total priority>best so far then {bestsofar = totalpriority, chosentarget=this target}
}
For timing, use end of (last enemy) turn, or when one of the AI's permanents is about to be destroyed, or when there are unblocked enemy attackers.
Instead of CMC, if such a thing exists, use "in play value" of the permanent, excluding value granted to it by auras/equipment.
In that case standardize that number to a 0-10 range where 0 is a worthless permanent, 10 is an absolutely amazing game changer.
You might need to tweak the numbers slightly but be careful with that, you always want to have a result of 0 or below if a target is not worth it such as bouncing something that has both a good and a bad "enters play" effect.
If the bounce effect is sorcery, prefer Main 1 for enemy controlled, Main 2 for own targets to maximize combat damage potential.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 10 Sep 2017, 19:02

Next deck. Shouldn't be causing AI problems, it's almost all creatures...
| Open
3 Disenchant|TMP
2 Cursed Scroll|TMP
3 Gerrard's Battle Cry|TMP
5 Plains|TMP|1
5 Plains|TMP|2
5 Plains|TMP|3
5 Plains|TMP|4
4 Soltari Champion|STH
2 Soltari Crusader|TMP
4 Soltari Emissary|TMP
4 Soltari Foot Soldier|TMP
2 Soltari Lancer|TMP
4 Soltari Monk|TMP
4 Soltari Priest|TMP
4 Soltari Trooper|TMP
4 Soltari Visionary|EXO


Observed no AI problem so far, the deck worked fairly well but had problems with playing the 3 mana cards, and generally seemed to need more lends. This is a candidate for medium difficulty, and I'll attempt to make a "hard" version by removing some 3 mana creatures and adding more lands and maybe Cursed Scrolls.
I haven't found out if the AI can or cannot use Gerrard's Battle Cry, as it never had the free mana to do so - we'll see it when testing the hard version.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 10 Sep 2017, 19:47

One more, built around Humility.
| Open
3 Bullwhip|STH
3 Gerrard's Battle Cry|TMP
3 Hero's Resolve|TMP
4 Humility|TMP
4 Orim's Prayer|TMP
4 Paladin en-Vec|EXO
4 Pegasus Stampede|EXO
7 Plains|TMP|1
9 Plains|TMP|2
4 Plains|TMP|3
3 Plains|TMP|4
4 Spirit Mirror|TMP
4 Winds of Rath|TMP
4 Youthful Knight|STH


Spirit Mirror : The AI should prioritize trading on blocks using the reflection token, even if another exactly same value creature exists. (which is likely the case, during Humility)
It also should prioritize blocking with the token even if it doesn't kill the attacker, just to save some health. The token comes back next turn anyway.
Both of these rules should only apply if the AI still has the mirror in play.

Surprisingly, I won 6 against this deck and lost none, even though I play creatures mainly and no enchantment removal. It's just too slow to compete and will need to stay Easy difficulty I'm afraid.

I still haven't seen the AI activate Gerrard's Battle Cry even once but I don't think it had very optimal situations to do so.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 12 Sep 2017, 12:11

The burn deck :
| Open
2 Cursed Scroll|TMP
4 Ensnaring Bridge|STH
4 Fanning the Flames|STH
4 Fireslinger|TMP
4 Kindle|TMP
4 Lightning Blast|TMP
4 Mogg Fanatic|TMP
2 Mountain|TMP|1
7 Mountain|TMP|2
4 Mountain|TMP|3
4 Mountain|TMP|4
4 Price of Progress|EXO
4 Rolling Thunder|TMP
1 Seismic Assault|EXO
4 Sudden Impact|TMP
4 Wasteland|TMP


AI played Price of Progress when I had no nonbasic lands.

Ideally, the AI should
-not play the card if the player controls equal or fewer nonbasic lands
-but always play the card if the player would die to the damage or would take a larger percentage of their total life as damage, even if this overrides the above
-not play the card in the first few turns of a game to give time to the player to play all their lands, allowing to deal maximal damage.

I see a lot of "svar X not found" messages in the console for Sudden Impact. AI is not casting the card when it has it in hand.

AI still taps Fireslinger to damage me and itself when I'm ahead in total life.

This deck is unplayable for now.

Evacuation - Scrivener deck
| Open
4 Counterspell|TMP
4 Dismiss|TMP
4 Ertai, Wizard Adept|EXO
4 Evacuation|STH
6 Island|TMP|1
6 Island|TMP|2
6 Island|TMP|3
4 Island|TMP|4
4 Legacy's Allure|TMP
4 Mana Leak|STH
2 Power Sink|TMP
4 Propaganda|TMP
4 Scrivener|EXO
2 Thumbscrews|TMP
2 Treasure Trove|EXO


The idea is to use Evacuation repeatedly, with Scrivener returning it from the grave. Probably not a strong deck but interesting.

Ertai, Wizard Adept
The AI shouldn't use this for attacking if it has 4 untapped mana. If having the option to counter a spell with this or a spell in hand, it should use this.

The AI is overly aggressive on offense. Not sure if it's the default template only, but it attacks with a 2/2 First Strike creature when I have a 3/2 and it's at 10 life while I'm at 18. That's a bad exchange of blows, it should keep it as blocker. Dealing 2 damage in exchange for taking 3 should only be considered an option if the ratio of life totals is more than 3/2 in the AI's favor.
(we kept exchanging blows until the AI reached the defined 5 life where is starts to prefer leaving blockers)

Even with these flaws, the deck won 3 out of 8 games, so it'll be tested again for medium difficulty.

Next - black shadow creature deck
| Open
1 Coat of Arms|EXO
1 Commander Greven il-Vec|TMP
2 Cursed Scroll|TMP
2 Dark Banishing|TMP
2 Dauthi Cutthroat|EXO
4 Dauthi Horror|TMP
4 Dauthi Marauder|TMP
4 Dauthi Mercenary|TMP
4 Dauthi Slayer|TMP
4 Dauthi Warlord|EXO
4 Diabolic Edict|TMP
2 Extinction|TMP
1 Swamp|TMP|1
6 Swamp|TMP|2
3 Swamp|TMP|3
7 Swamp|TMP|4
1 Volrath's Stronghold|STH
4 Wall of Souls|STH
4 Wasteland|TMP


This deck is a monster - I lost 4 out of 6. Definitely medium or above. The games never lasted long enough for Greven to be playable so idk if the AI can use the card or not but it seems useless for the deck, I'll remove him. I'll split this to two decks : a harder version with 4 scrolls and more efficient removal spells, and another with no scrolls and some higher cost removal (Extinction). Hoping for the pair to make it to hard/very hard.

Next "Dark Druids" :
| Open
4 Avenging Druid|EXO
2 Carrionette|TMP
4 Dauthi Horror|TMP
4 Dauthi Slayer|TMP
3 Elvish Berserker|EXO
3 Forest|TMP|1
2 Forest|TMP|2
2 Forest|TMP|3
5 Forest|TMP|4
4 Hermit Druid|STH
4 Muscle Sliver|TMP
4 Oath of Ghouls|EXO
2 Overrun|TMP
2 Pit Spawn|EXO
1 Plaguebearer|EXO
2 Skyshroud War Beast|EXO
3 Swamp|TMP|1
2 Swamp|TMP|2
3 Swamp|TMP|3
2 Trained Armodon|TMP
2 Verdant Force|TMP

The idea is to take advantage of Oath of Ghouls in an almost all creature deck that also fills the grave for an unstoppable horde of creatures.

Skyshroud War Beast
Shouldn't be played if the player has zero nonbasic lands.

6 wins, 1 loss, this deck is weak. Categorizing as easy. Don't know if the AI can activate Carrionette or not.

Overall status :

Still untested : 11 decks
Easy : 2 decks
Medium or higher : 6 decks
Needs retesting due to AI problems : 8 decks
Needs to be built : ~10 more decks
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 16 Sep 2017, 08:40

Continuing...

| Open
4 Bottomless Pit|STH
4 Bounty Hunter|TMP
2 Coffin Queen|TMP
4 Cursed Scroll|TMP
4 Dark Ritual|TMP
2 Dauthi Mindripper|TMP
2 Dungeon Shade|STH
4 Ensnaring Bridge|STH
4 Megrim|STH
4 Null Brooch|EXO
4 Scalding Tongs|TMP
6 Swamp|TMP|1
1 Swamp|TMP|2
4 Swamp|TMP|3
7 Swamp|TMP|4
4 Wasteland|TMP


Deck built on the "empty own hand" mechanic. Null Brooch, Ensnaring Bridge, Bottomless Pit, Cursed Scroll.

Dungeon Shade - I blocked this with a 2/1 and the AI had 4 untapped swamps but didn't pump it.

Mox Diamond - Not in the deck currently, any chance for AI support on this card? I think it could improve it significantly by speeding up both mana and hand depletion.

In the current form, I won 6 out of 7 - so this should be an easy opponent but I'll bump it up to medium because it has a lot of frustrating/hard to deal with mechanics. I'm also trying to build a faster and more efficient version for hard, not sure if possible.

Next :
| Open
4 Capsize|TMP
4 Fanning the Flames|STH
3 Island|TMP|1
3 Island|TMP|2
4 Island|TMP|3
2 Island|TMP|4
4 Memory Crystal|EXO
3 Mountain|TMP|1
6 Mountain|TMP|2
2 Mountain|TMP|3
4 Ruby Medallion|TMP
4 Sapphire Medallion|TMP
2 Searing Touch|TMP
2 Shattering Pulse|EXO
3 Wall of Diffusion|TMP
2 Wall of Razors|STH
4 Wall of Tears|STH
4 Whispers of the Muse|TMP


Capsize - The AI taps itself out by casting this during upkeep. Should be used at end of turn or at least during enemy combat phase unless it's removing a blocker. This is especially problematic because the AI was holding Fanning the Flames and could have killed those same creatures instead.

For now this deck is marked unplayable.

Next :
| Open
4 Avenging Druid|EXO
3 Burgeoning|STH
2 Dirtcowl Wurm|TMP
4 Evincar's Justice|TMP
7 Forest|TMP|1
5 Forest|TMP|2
3 Forest|TMP|3
2 Forest|TMP|4
4 Hermit Druid|STH
4 Horn of Greed|STH
4 Nature's Revolt|TMP
1 Overrun|TMP
4 Rampant Growth|TMP
3 Spidersilk Armor|MMQ
3 Swamp|TMP|2
5 Swamp|TMP|3
2 Swamp|TMP|4


Deck built around Nature's Revolt - lots of land acceleration. Unfortunately had to use a card from outside the block (Spidersilk Armor) because there is no card in the block with the effect "all creatures get +0/+1 or +1/+1".

No AI problems - deck works and if it gets the combo in play, it wins - but it's very slow so easy difficulty.

Next :
| Open
4 Commander Greven il-Vec|TMP
4 Grave Pact|STH
4 Helm of Possession|TMP
2 Hornet Cannon|STH
4 Lab Rats|STH
4 Oath of Lieges|EXO
4 Pegasus Stampede|EXO
4 Plains|TMP|1
1 Plains|TMP|2
1 Plains|TMP|3
5 Plains|TMP|4
2 Selenia, Dark Angel|TMP
2 Soltari Visionary|EXO
4 Spirit Mirror|TMP
2 Staunch Defenders|TMP
1 Swamp|TMP|1
3 Swamp|TMP|2
4 Swamp|TMP|3
3 Swamp|TMP|4
2 Wall of Essence|STH


Doesn't seem to draw enough creatures to use the pact and helm, adding 2 more walls of essence.

Hornet Cannon - AI uses it during upkeep, shouldn't. Use this during the beginning of combat phase only (own if planning to send the token to attack, enemy if block). However I think the deck works better without the card anyway, so I'm removing it.

With the above changes, I lost 3 out of 5, this deck definitely works and goes into the "at least medium" folder.

Next deck has Price of Progress and Sudden Impact again - I haven't seen any commits on AI for those yet so moving this to the unplayable folder.

Next one :
| Open
4 Apes of Rath|TMP
4 Charging Rhino|TMP
7 Forest|TMP|1
4 Forest|TMP|2
7 Forest|TMP|3
5 Forest|TMP|4
1 Horned Sliver|TMP
2 Mirri, Cat Warrior|EXO
3 Mongrel Pack|TMP
4 Muscle Sliver|TMP
1 Plated Rootwalla|EXO
2 Rabid Wolverines|EXO
2 Skyshroud Archer|STH
2 Skyshroud Elite|EXO
2 Spike Weaver|EXO
2 Spined Wurm|STH
4 Trained Armodon|TMP
4 Wall of Blossoms|STH


Simple green creature deck with big, efficient creatures. Probably easy tier.

Won 6 out of 8. Suitable for easy but has potential for medium if we add some spells.

next :
| Open
4 Apes of Rath|TMP
4 Craven Giant|STH
3 Forest|TMP|1
3 Forest|TMP|2
1 Forest|TMP|3
3 Forest|TMP|4
3 Jackalope Herd|EXO
2 Mountain|TMP|1
2 Mountain|TMP|2
3 Mountain|TMP|3
1 Mountain|TMP|4
4 Pandemonium|EXO
4 Rampant Growth|TMP
3 Rathi Dragon|TMP
4 Segmented Wurm|TMP
4 Skyshroud Elf|TMP
4 Spined Wurm|STH
4 Wall of Razors|STH
4 Wild Wurm|TMP


Pandemonium with creatures that have high power for a low cost.

This works very poorly. The AI should learn to play Pandemonium first from its hand otherwise the deck just doesn't work.
Aside from that it suffers from the same problem as the green deck : cost efficient big creatures with no abilities just isn't enough to win a game - the player will leave blockers and sneak in with flying or shadow attackers, or play removal - a few terrors and 2/2s beats a deck like this, unless it draws the Pandemonium and plays it before the creatures, even then it's probably slow.
So I'll put this into easy - even if the AI learns to play Pandemonium first it's still a weak deck.

Next one :
| Open
4 Blood Frenzy|TMP
4 Canyon Wildcat|TMP
4 Fireslinger|TMP
4 Giant Strength|TMP
4 Jackal Pup|TMP
4 Maniacal Rage|EXO
4 Mogg Fanatic|TMP
4 Mogg Maniac|STH
5 Mountain|TMP|1
5 Mountain|TMP|2
6 Mountain|TMP|4
4 Raging Goblin|EXO
4 Rathi Dragon|TMP
4 Wasteland|TMP

Red aggro, creature based deck.

Won 5 out of 6, this deck isn't working as well as I hoped.
This seems to be the case where Profile=Reckless might improve performance, I'll retest with that, and I'll also add some removal and Cursed Scrolls.

The AI is still accelerating its own demise by using Fireslinger to deal damage to players when it's lower on health himself. This needs to be fixed - while not a major concern in this deck, it is blocking at least two other decks from inclusion.

After modifications, I lost 4 out of 7, so the improved version will be tested for medium or higher difficulty in the next loop. The unmodified one will be included as easy difficulty.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Seravy » 16 Sep 2017, 13:47

Next deck :
| Open
4 Death Pits of Rath|TMP
4 Disenchant|TMP
4 Evincar's Justice|TMP
4 Oath of Lieges|EXO
2 Orim's Prayer|TMP
4 Paladin en-Vec|EXO
2 Pit Spawn|EXO
7 Plains|TMP|1
2 Plains|TMP|2
1 Plains|TMP|3
2 Plains|TMP|4
2 Selenia, Dark Angel|TMP
4 Soltari Monk|TMP
2 Spirit Mirror|TMP
4 Swamp|TMP|1
2 Swamp|TMP|2
3 Swamp|TMP|3
3 Swamp|TMP|4
4 Youthful Knight|STH


Death Pits of Rath with first strike creatures and Evincar's Justice.

I won 6 out of 8 - while the deck worked well, it still is only easy difficulty level. I suspect my test deck is just too powerful - it doesn't seem so but I have lots of efficient cheap creatures (2/2s for 2 or 3/1s for 3), and some decent other cards on top of that (a 4/4 unblockable, a sol ring etc), plus quite a few removal (terror etc). I think I'm going to roll another test deck after I complete every test using this one, and do it all over again. Maybe playing black/white when testing isn't fair, those colors are far too good? Too much removal and weenies as well as plenty of flying/otherwise unblockable things.

Next deck...
| Open
2 Capsize|TMP
1 Ertai, Wizard Adept|EXO
4 Island|TMP|1
4 Island|TMP|2
4 Island|TMP|4
4 Legacy's Allure|TMP
4 Mana Leak|STH
2 Master Decoy|TMP
3 Paladin en-Vec|EXO
5 Plains|TMP|1
4 Plains|TMP|2
1 Plains|TMP|3
2 Plains|TMP|4
1 Power Sink|TMP
4 Propaganda|TMP
4 Sky Spirit|TMP
3 Soltari Visionary|EXO
2 Spirit Mirror|TMP
2 Treasure Trove|EXO
4 Winds of Rath|TMP

Blue-White control with no particular extra tricks or combos.

AI uses Master Decoy to attack and deal 1 damage even if it does leave mana untapped and I can swing back for more than 2 damage next turn. This creature should never attack unless the enemy has no creatures in play, or the AI expects to have no untapped mana after the turn.

7 wins, 1 loss, easy difficulty. Haven't observed any AI problems here.

And the last already built deck :
| Open
2 Acidic Sliver|STH
2 Clot Sliver|TMP
3 Coat of Arms|EXO
4 Crystalline Sliver|STH
4 Forest|TMP|1
4 Forest|TMP|2
2 Forest|TMP|4
4 Heartstone|STH
4 Hermit Druid|STH
1 Island|TMP|1
2 Island|TMP|2
1 Island|TMP|4
1 Mnemonic Sliver|TMP
1 Mountain|TMP|3
2 Mountain|TMP|4
4 Muscle Sliver|TMP
3 Plains|TMP|2
1 Plains|TMP|3
4 Rampant Growth|TMP
4 Sliver Queen|STH
3 Swamp|TMP|4
2 Victual Sliver|STH
2 Winged Sliver|TMP


Would benefit greatly from some Mox Diamonds I believe if the AI could learn to play those.

6 wins, 1 loss, I guess I must try and see how the AI plays Mox Diamond before making a decision on this one.

...well, the AI does play it if it's holding an additional land after the normal land play. I suggest making a new AI logic "PlayBeforeLands" for this and the "discard a land" Wurm from the portal set and any similar card - with only that much change, Mox Diamond would be playable. Until then, this will wait in the unplayable folder.

And now, time to build more decks, I have a long list of cards I need to look at if anything can be made for them.
Seravy
 
Posts: 363
Joined: 26 Oct 2016, 21:23
Has thanked: 5 times
Been thanked: 27 times

Re: Tempest world development

Postby Agetian » 16 Sep 2017, 14:34

I'll see if I can figure out Mox Diamond AI logic ;)

- Agetian
Agetian
Programmer
 
Posts: 3296
Joined: 14 Mar 2011, 05:58
Has thanked: 632 times
Been thanked: 485 times

Re: Tempest world development

Postby Agetian » 16 Sep 2017, 15:48

Mox Diamond is now AI playable. Also, added a similar AI logic to Thundering Wurm.

- Agetian
Agetian
Programmer
 
Posts: 3296
Joined: 14 Mar 2011, 05:58
Has thanked: 632 times
Been thanked: 485 times

PreviousNext

Return to Developer's Corner

Who is online

Users browsing this forum: No registered users and 9 guests


Who is online

In total there are 9 users online :: 0 registered, 0 hidden and 9 guests (based on users active over the past 10 minutes)
Most users ever online was 279 on 11 Jul 2013, 22:03

Users browsing this forum: No registered users and 9 guests

Login Form