Mahtan/wagic
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

AI Deck cleanup in preparation for the upcoming beta release:

Browse Source 
1. temporarily removed my "Pyromancer" and "Terravore Turmoil" decks. Both performed very badly in my AI comparison done here: http://wololo.net/forum/viewtopic.php?f=3&t=736

While having low-performing decks is not a bad thing per se (actually we *need* some low-performers, as stepping stones for beginners), these two decks currently throw AI deficiencies into the face of the players (targeting problems / fetchlands), and we don't want to do that. Players should feel that they beat the low-perfoming decks because of their own improving skills, not because the AI can't play its own deck.

I'll re-add both decks after the next release, hoping that we can improve the AI until the release after that.

2. Added "Millage" and "Bad Dreams" decks by Niegen. I see that we are in "bug fixes only" mode since yesterday, but I had announced these additions to Niegen two days ago, and he liked the prospect, so I didn't want to cancel the additions.

3. Updated "Kobold Overlord" deck with two flyers, this is actually a week-old change that I just hadn't submitted so far because it was so minor.

I'll also do a general deck audit to prepare for the next release (respective task will be added to Googlecode tracker). Since we're in "bug fixes only" mode, I'll just fix some obvious mistakes - it's always tempting to do last-minute tweaks and improvements, but I wouldn't want a card or deck that was added late, and hardly tested, to cause problems after the release.

This means we can release the next version with 50 AI decks, more than double than the previous version (which had 24). :)
This commit is contained in:
Psyyringe
2009-10-22 11:20:49 +00:00
parent 14fba35ebf
commit 386bcf2bc7
4 changed files with 79 additions and 146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
projects/mtg/bin/Res/ai/baka/deck34.txt
View File

@@ -3,14 +3,15 @@
#DESC: Creature's Decks 1: #DESC: Creature's Decks 1:
#DESC: "The deck that a Kobold Overlord #DESC: "The deck that a Kobold Overlord
#DESC: would (and could) assemble." #DESC: would (and could) assemble."
#DESC: (v1.06) #DESC: (v1.07)
#DESC: #DESC:
#DESC:You may think we are weak #DESC:You may think we are weak
#DESC:But we are many #DESC:But we are many
#DESC:And sometimes we can hire #DESC:And sometimes we can hire
#DESC:A powerful mercenary #DESC:A powerful mercenary
# #
Mountain (10E) *22 Mountain (10E) *20
Kher Keep (TSP) *2 # produces kobold tokens
Crimson Kobolds (LEG) *4 Crimson Kobolds (LEG) *4
Crookshank Kobolds (LEG) *4 Crookshank Kobolds (LEG) *4
Kobold Drill Sergeant (LEG) *4 Kobold Drill Sergeant (LEG) *4
@@ -51,9 +52,3 @@ Mob Justice (STH) *3 # finisher
# for mana acceleration # for mana acceleration
# Removed because the AI doesn't use the mana. # Removed because the AI doesn't use the mana.
# Also ... only works with goblins. Duh. *blush* # Also ... only works with goblins. Duh. *blush*
#
# 2 x Kher Keep (TSP) (ID 113553)
# For bringing additional kobolds into play
# AI taps Kher Keep for mana and then can't
# use that mana to produce a kobold token.
#

76
projects/mtg/bin/Res/ai/baka/deck37.txt
View File

@@ -1,48 +1,28 @@
#NAME:TerravoreTurmoil #NAME:Kinsbaile Cavalier
# #DESC:The knights on their steeds
#DESC: Creature's Decks 4: #DESC:and the knights from the sky
#DESC: "The deck that a Terravore #DESC:unite
#DESC: would (and could) assemble." #DESC:in their crusade for honor
#DESC: (v1.01) Blessing (RV) * 2 #
#DESC: Crusade (RV) * 4 #
#DESC: Your homeland is my lunch. Northern Paladin (RV) * 1 #
# Lost Order of Jarkeld (ICE) * 2 #
Order of the Sacred Torch (ICE) * 1 #
Birds of Paradise (10E) *4 # mana source Soltari Crusader (TMP) * 1 #
Vine Trellis (MRQ) *4 # mana source and defender Crusading Knight (INV) * 1 #
Leonin Skyhunter (MRD) * 1 #
Carven Caryatid (RAV) *3 # card drawer and defender Flagstones of Trokair (TSP) * 2 #
Collective Unconscious (MRQ) *2 # card drawer Paladin en-Vec (10E) * 2 #
Plains (10E) *16 #
Knight of the Reliquary (CFX) *4 # attacker Skyhunter Patrol (10E) * 1 #
Terravore (ODY) *4 # attacker Skyhunter Prowler (10E) * 2 #
Plover Knights (LRW) * 1 #
Armageddon (POR) *4 # puts lands into graveyard Knight of Meadowgrain (LRW) * 2 #
Kudzu (RV) *1 # should be 4, but the AI puts Kinsbaile Cavalier (MOR) * 3 #
# several Kudzus on one land, Wilt-Leaf Cavaliers (SHM) * 3 #
# which causes a crash Kabira Crossroads (ZEN) * 2 #
Winter's Grasp (TMP) *2 # weak replacement for Kudzu Sigiled Paladin (ALA) * 1 #
White Knight (M10) * 2 #
Nature's Lore (ICE) *4 # land fetcher for fast buildup Honor of the Pure (M10) * 4 #
Rampant Growth (10E) *4 # land fetcher for fast buildup Plains (M10) * 4 #
Armored Ascension (M10) * 2 #
Grasslands (MIR) *4 # two land fetchers which put
Mountain Valley (MIR) *4 # themselves into the graveyard,
# feeding the Terravore and the
# Knight
Forest (M10) *8
Plains (10E) *6
Flagstones of Trokair (TSP) *2
#
# Cards considered, but not included:
# Foratog - eats forests (which is good), but the AI doesn't use
# this ability.
# Terramorphic Expanse - crashes game
# Windswept Heath - crashes game
# Elves - more mana sources not needed
#
# Cards removed from the deck:
# 3 more Kudzu- game crashes when a land with two Kudzu on it is
# tapped

113
projects/mtg/bin/Res/ai/baka/deck38.txt
View File

@@ -1,91 +1,24 @@
#NAME: Pyromancer #NAME:Bad Dreams
# #DESC:Vision without action is a daydream.
#DESC: Creature's Decks 5: #DESC:Action without vision is a nightmare.
#DESC: "The deck that a Prodigal Pyromancer
#DESC: would (and could) assemble"
#DESC: (v1.01)
#DESC: #DESC:
#DESC: Well protected by solid stone #DESC:Japanese Proverb
#DESC: I scorch you with lances of fire #DESC:
# #DESC:v1.0 2009 10 01
Black Vise (RV) *1
Prodigal Pyromancer (10E) *1 # actually the deck evolved to a Wheel of Fortune (RV) *1
# state where it would fare better Howling Mine (M10) *4
# without its main creature ... Underworld Dreams (10E) *4
Spiteful Visions (SHM) *4
Wall of Diffusion (TMP) *4 # 12 walls for defense Sudden Impact (10E) *4
Wall of Earth (LEG) *4 Font of Mythos (ALA) *2
Wall of Stone (RV) *4 Damnation (PLC) *4
Infest (ALA) *4
Beacon of Destruction (10E) *3 # 19 direct damage spells Lightning Bolt (M10) *4
Flame Burst (ODY) *4 Terminate (ARB) *4
Kindle (TMP) *4 Swamp (RAV) *4
Lava Spike (CHK) *4 Swamp (TSP) *4
Lightning Bolt (M10) *4 Swamp (LRW) *4
Swamp (ALA) *4
Howling Mine (10E) *4 # accelerates the game, makes more Mountain (RAV) *4
# damage spells available Mountain (TSP) *4
Mountain (10E) *24
# Cards considered, but not included:
# Shock - nice damager per cost ratio, but 2 damage is too little,
# especially since the AI likes to use it on creatures with
# toughness >2
# Steam Blast - nice effect on all creatures and players, but
# damage amplitude of 2 is just a bit too little.
#
# Aetherflame Wall - no need for a 4th wall
# Ball Lightning - powerful card, but the AI sometimes just casts
# it and then doesn't attack with it, so the card is wasted
# Blaze - AI currently casts spells with "X" cost for too weak an
# effect
# Char - damage per cost ratio too low, and damages caster as well
# Earthquake - difficult to fine-tune for the AI: too little mana
# and the effect might be wasted, too much mana and the effect
# might be dangerous for the AI and its creatures as well.
# Fault Line - more expensive than Earthquake for same effect
# Fire Tempest - very expensive, and too dangerous for the AI
# Fireball - AI doesn't choose targets well
# Flame Javelin - 3 mana for 4 damage is too expensive
# Flamewave Invoker - ability too expensive compared to Beacon of
# Destruction
# Lava Axe - Too expensive for one-time use, esp. since red mana
# acceleration is currently not possible
# Lightning Blast - Too expensive
# Puncture Blast - damage per cost ratio too low. Wither damage
# would help against big creatures, but so do the walls, and the
# opponent hopefully won't last long enough to bring many big
# creatures out anyway.
# Puncture Bolt - effect too weak
# Scorching Spear - effect too weak
# Sonic Burst - AI doesn't make smart discarding decisions
# Zap - too little damage, the AI often uses it on a creature
# that doesn't get killed by it, so the card is wasted.
# Jayemdae Tome - casting cost and ability cost too expensive in
# that deck; Howling Mine serves the same purpose, and hopefully
# the opponent will be dead before he can make much use of the
# cards he draws.
# Dancing Scimitar - Perhaps add as a defense against flyers as
# long as Aether Membrane isn't available?
# Cards removed from the deck:
# Volcanic Hammer - was a stand-in for Kindle as long is that was
# buggy.
# Notes:
#
# The AI doesn't target the damage spells well, often wasting them
# on creatures that are tough enough to survive the damage.
#
# Oddly, the AI often doesn't block attackers although it could.
#
# Also, the AI summons walls even if no enemy creatures are on the
# battlefield. This could be useful in case the opponent has
# creatures with Haste, but nevertheless the AI would fare better
# if it prioritized other spells over wall summoning as long as no
# enemy creatures are out.
#
# The deck would profit massively from Mana Flare and Aether
# Membrane, but neither card is currently implemented.

25
projects/mtg/bin/Res/ai/baka/deck50.txt Normal file
View File

@@ -0,0 +1,25 @@
#NAME:Millage
#DESC:More than one mage has been
#DESC:driven insane by the sound
#DESC:of the millstone relentlessly
#DESC:grinding away.
#DESC:
#DESC:v1.0 2009 10 01
Wrath of God (10E) *4
Damnation (PLC) *4
Evacuation (10E) *3
Moat (LEG) *2
Tome Scour (M10) *4
Howling Mine (M10) *4
Glimpse the Unthinkable (RAV) *4
Dream Fracture (EVE) *4
Memory Erosion (ALA) *4
Traumatize (M10) *2
Forced Fruition (LRW) *1
Island (RAV) *4
Island (TSP) *4
Island (LRW) *2
Swamp (RAV) *4
Swamp (TSP) *3
Plains (RAV) *4
Plains (TSP) *3