List of stuff that the AI can't handle
Moderators: BAgate, drool66, Aswan jaguar, gmzombie, stassy, CCGHQ Admins
2 posts
• Page 1 of 1
List of stuff that the AI can't handle
by lujo » 31 Oct 2015, 05:14
To help out anyone trying to make decks.
If someone figures out how to make a deck revolving around these which I can try out and verify as working sensibly I would be thankful. "Occasionally stumbles into doing the right thing" doesn't cut it, card combinations which manage to force the AI to not do stupid things, or benefit from doing stupid things instead of embarrassing itself are where it's at.
As of NB1:
---
Equipment in general - Any number of things with it, spends all it's mana to move it around at any opportunity instead of casting spells, casts it instead of creatures on an empty board, doesn't understand that it can be more aggressive with Equipped creatures, takes them off things it will attack with, puts it onto stuff it won't etc.
Confirmed Examples:
Armory of Iroas
Prowler's Helm
Bone Saw
Lightning Greaves
Prowler's Helm - will occasionally work out in particular matches, but doesn't more often than not it won't.
You name it - sometimes equipment other than these works out fine after fine tuning the deck, but in general there's no telling what the AI will do with it.
Trepanation Blade - is just plain bugged and doesn't trigger.
Stuff that works, somewhat:
Leonin Scimitar (cost and activation cheap enough)
Bonesplitter (same)
Shuko (same)
Trusty Machete (Lujo: seems to kind of work fine in my current Merfolk Shaman tweaked deck. I'm not sure WHY, though.)
Darksteel Axe (Lujo: since it's possible to get the AI to use the Machete somewhat sensibly, I suppose this could also work)
Demonmail Hauberk (Lujo: You CAN make a Demonmail Hauberk work, except the AI doesn't realize it's easier to first put creatures down than it, and it costs 4. I put in a few hours trying to make a Rukh Egg / Dragon Egg deck with it, and it's quite uneven but looks like it could be made to work)
Kitesail can work out fine
Deckbuilding tips:
You tell me.
Lujo's got: Don't put any in. / Don't put too many in. / Don't put any but the first three mentioned ones, which have long since been deemed as too good by Wizards because they have too good cost/equip prices. But hey, with things they way they are now, if it works it works. / There's some way to make cc 1 equip c 2 +2/+1 equipment work, quite well, I'm not really clear on what yet.
---
Creatures that sacrifice other creatures Most sacrifice outlets - will be activated at the strangest times, for example sacrificing an unblocked 4/4 flier to pump a 2/2 discard outlet blocked by a 0 powered wall. Spectacular lunacy includes sacrificing the creature itself whenever it gets the slightest hint of benefiting from it. Trying to get them to work as a basis of a deck is very likely to end in trainwreck.
Confirmed Examples:
Blood Bairn
Plagued Rusalka
Culling the Weak - Lujo: The AI has no idea about the importance of creatures it sacrifices, and I've even seen it sacrifice things with nothing to cast.
Bubbling Cauldron - Lujo: It will sacrifice everything and anything to it whenever it can, dear god.
Infernal Plunge - Lujo: Same thing as culling the weak
Stuff that consistently works in this vein:
No idea.
Deckbuilding Tips
Lujo: Don't bother using them? Use non-creature outlets which don't use creatures as the win condition. Hecatomb, for example, works rather well (although if you make a Shandalar deck with it it'll just murder things).
---
Creatures whith effects when they die / cannon fodder AI will chump-block with other stuff; keep them around forever; pump them instead of other things; sacrifice them to an outlet madly; value them over evasive things; it has little grasp of things being "sacrifice fodder".
Confirmed Examples:
Elder Cathar
Black Cat
Festering Newt
Festering Goblin
Things which work in this vein:
No idea
Deckbuilding tips:
Lujo: They all work if you put them in a deck which can't avoid killing them as collateral damage IF their on-death effects don't affect creatures. Example would be Onulet/Highland game in decks with Earthquake / Wrath of God.
---
Creatures which sacrifice themselves for effects - there's no telling when the AI will choose to activate it, and often very little obvious sense to it.
Confirmed Examples
Frostling
Goblin Firestarter
Sakura-Tribe Elder
Kami of False Hope (will get sacrificed at odd times)
Stuff that works somewhat better:
Kami of Ancient Law
Things that sacrifice to destroy artifacts (haven't tried them out yet, though)
Pain Kami
Deckbuilding Tips
You tell me.
---
Cantrips (cards which Draw a card / Scry in addition to something else they do) - The AI will treat them as card draw primarily (it'll rate the benefit of card drawing as so desirable that it won't care about wasting the rest of the card). This can also lead to trainwercks, and usually does. What it chooses to scry to the bottom of the library can also occasionally be mind-boggling, making giving it the effect risky and random.
Confirmed Examples
Spite of Mogis
Defiant Strike
You name it, pretty much. It won't always backfire, but it's often going to be a lousy move made in place of a better move, or not what you put the card in the deck for.
Stuff that still somewhat works:
Lujo: Possibly Repulse and counterspells which draw a card. It's still possible that it will bounce it's own creatures and counter it's own spells, though, I haven't checked.
Deckbuilding tips:
Lujo: Avoid like the plague / Give it only expensive cantrips. This defeats the purpose of cantrips somewhat, but reduces the opportunity for the AI to blow them truly mindlessly. / Don't try to give it "narrow" effects, as that can backfire
---
Removal - Burn that can target players gets thrown to the face if there are no targets which sucks in Green and Blue as well as Red; Bounce and kill is cast with utmost priority over anything else the AI could do including playing creatures to attack with; board sweepers will be spammed on a single target (even if the AI has a board advantage realistically); scaling removal will be chained even if waiting for one more land/creature/whatever-it-counts would let it kill something with one card if it waited just one more turn; things which do minor damage on top of a larger effect will be spammed mindlessly at first opportunity; removal with minor bonus effects like lifegain/no regeneration will be thrown madly (even at own creatures) for the effect; creature-only burn (especially sorceries) will be thrown before combat if it would reduce opponents effective toughness so that IF it blocks something it would die (but the AI sometimes still won't attack); etc. etc. etc. if you can imagine a stupid thing to do with removal besides flat out 1-for-1 killing it's own dudes it'll do it.
Confirmed examples:
You name it. The AI has a mechanism to determine whether it can pay for something that can be classified as removal, a way to tell you it has moved from it's hand to the stack, and an urge to do so if it has it in it's hand and there is a target. From the observers point of view that's what's gonna happen with any and all removal you give it, and every time that's a stupid move it's still gonna do it with seemingly anything you give it.
Stuff that still works somewhat:
Lujo: Spells that distribute damage among several targets (if they can also hit the player) like Arc Lightning, Forked Bolt etc. help red guys because they lessen the effect of wasting cards on foolishness (but are quite overkill as that's basically the most card-efficient burn ever);
Lujo: Expensive removal in general helps because the AI doesn't cast it over creatures so that when it plays the removal it has something to attack with (although this inflates mana curves and possibly creates slow hands).
Lujo: super-efficient-beyond-overkill removal that only targets creatures like Swords to Plowshares or cheap removal auras like Sensory Deprivation get cast early enough that it gets to play stuff instead of removal (although that's like solving someone's aim problems by giving it a bazooka for StP and the AI can be a bit fiddly with Auras).
Lujo: Generally removal up to 2 cc which doesn't target players might not slow it down too much or get wasted. It'll get madly chained if the AI draws a bunch, though.
Lujo: Creatures which do damage/kill stuff as comes-into-play effects like Fire Imp or Flametongue Kavu prevent it from playing removal when it should be playing creatures. Those are also madly efficient cards in any way immaginable, and FtK regularly makes lists of top 10 cards that should've never been made (by folks who know what they're talking about), so yeah, sure solves problems.
---
Wither - AI doesn't understand it can get aggressive with it on creatures, and will cast removal with wither ultra aggressively.
Confirmed examples:
You name it, Blight Sickle, Sickle Ripper, Corrosive Mentor
Works Somewhat:
Puncture Blast - it will get cast on a creature just to debuff it, but at least that's somewhat plausible and not always detrimental.
Deckbuilding tips:
You tell me.
---
Things which can gain flying - The AI currently seems to think that things which can gain flying always have flying so it suicide attacks with them on occasion.
Confirmed Examples:
Goblin Baloon Brigade
Manta Riders
Mystic Visionary
Mystic Penitent
Dross Hopper
---
Attacking - AI will only attack with the vast majority of non-evasive creatures (meaning no flying/landwalk/unblockable/etC) if there's any chance of losing a creature on attack, unless for a certain-looking alpha strike or if it has a pump in hand.
Confirmed examples:
You name it.
Things that work somewhat:
It will always attack with Erg Raiders, for some reason and things that must always attack. It will also somewhat often attack with things which can't block. Things with Deathouch, it they're cheap, will also madly attack. They don't "work" so much as "not not-work in this particular sense".
Deckbuilding tips:
Lujo: Err, well, there's not much you can do. It's the inverse of how MtG is played IRL, IRL "nobody blocks", here "AI doesn't attack". I'll probably write up what seems to work and what doesn't when I'm done tweaking the decks. This can't stay this way forever, it's preposterous.
---
Blocking - the AI seems to make any block which would result in a trade, no matter how horrible the trade. It seems to be, to the casual observer, extra-prone to this if the creature they are about to trade has summoning sickness. You can use any creature as a removal spell this way and it makes playing vs. it a mockery. It also tends to block for trades instead of blocking with 0 powered defenders. It also seems to have 0 brain when choosing what to trade or chump-block with and will often sacrifice evasive creatures in place of sacrifice fodder, or gang-block with several creatures when a single deathtouch creature would do.
Confirmed Examples
You name it. I bet Erg Raiders are present in so many decks because you can't bait the AI into losing them as soon as they hit the board due to their 3 toughness.
Stuff that works somewhat
Nothing will prevent the AI from attempting to do this.
Deckbuilding tips:
Lujo: Just read this, and think about what it means: "Give it things the player can't attack into, so that it won't have a chance to consider blocking". That's how horrible it is at it. There's quite a bunch of ways to do this, but then those things have to also be removal-proof, and they better also be indestructible, just in case, and...
Lujo: On a less defeated note, creatures with higher toughness than power (especially if agressively costed) seem to do the trick somewhat (think Erg Raiders). It makes it more difficult for the AI to lose to many creatures when it madly gang-blocks.
---
Playing MtG - The AI can't attack, block or use removal sensibly. It also can't time various spells and ability activations in a way which doesn't interfere with developing board position. It also can't avoid getting mana screwed via mulligan if it has only one land, or is forced to play it's worst hands. And it has no idea of what role which card is supposed to serve in it's deck or what it's trying to do other than universal, general guidelines.
Confirmed Examples:
You name it.
Stuff that actually works:
Giving it cards which are so fool-proof strong that it can still win despite being unable to attack, block or use removal sensibly. But are also so fool-proof simple that it can't possibly misuse them. And are also not caught in a crossfire of bugs/lunatic general idiosincracies. Also spending days and hours on deckbuilding.
Deckbuilding tips:
Lujo: Spend way more time playing, studying and working on MtG than people who've spent way too much time playing, studying and working on MtG for a quarter of a decade, then spend better part of your waking day on nothing but sifting through stuff until you find combinations the AI doesn't embarrass itself with too hard. Try not going mad, well, more than you already are.
Flanking on attack - AI can't really calculate flanking right, I sometimes won't attack with, say, Fallen Askari even though there's no way he'd die in combat, and no thing to do with it later as it can't block.
Confirmed Examples
Fallen Askari (will occasionally display this)
Stuff that Works Somewhat
Cadaverous Knight (as the AI will sometimes be more agressive with it due to regeneration)
Deckbuilding tips
You tell me.
Lujo: Buffs can help it still get a Flaneker attacking, but that's besides the point - flankers are very deady attackers with no buffs on them and REALLY hard to gang-block in combat with them, if it's not attacking with them it's kinda wasting them.
Creatures that come into play tapped - it's possible for the AI to make mistakes when determining whether it has a blocker or not with creatures that come into play tapped (black has a bunch). If it has one in hand it can think it will be able to block next turn if it plays it and it won't.
Confirmed Examples
Tormented Hero
Stuff that works somewhat
You tell me.
Deckbuilding tips
Lujo: No idea, but I suppose if you don't overdo them in a deck it won't crop up as often.
---
Mana Vault - the AI can't untap it.
Dark Ritual - the AI will generally blow it on anything at all, so don't expect it to get anything expensive out in the early turns.
Shaper Guildmage - AI won't attack with him or activate him in combat somehow. No idea about other guildmages and masters.
Magma Spray - even crazier behavior than most removal for some reason. Looks good if you're trying to attempt to have the AI not throw away it's 1cc cheap removal thing to burn the enemy to the face, it gets spectacularly creative about throwing it away for no gain at all.
Infest - AI is exceptionally weird about when to cast or not to cast this thing.
Kashi-Tribe Reaver - It's regeneration seems to be bugged, so he doesn't in fact regenerate. Or so it seemed to me.
---
Hex - The AI won't sacrifice it's own creatures as part of it, which is a crying shame as it'd make a perfect curve-topper for Witch. If it could use it intelligently, and sac Sengir Autocrat tokens to it :_(
Most "removal" auras - it will cast them post-combat and not benefit from casting them by attacking. It will also generally stack them on the same creature unless individual instances have been reported and addressed.
Whitewater Naiads - since the AI plays both them and removal auras which you'd want to pair up with them post-combat they're very difficult to get any use out of right now.
---
In the same vein as Whitewater Naiads:
Kami of Fire's Roar , Earthshaker - AI seems either unwilling or incapable of using this sort of thing because it generally plays creatures after combat, and these guys have effects which let you attack easier if you do your stuff pre-combat (one makes guys unable to block, the other flat out kills stuff).
---
Playing vs stuff with Protection from Color - the AI seems to go completely mad when it's facing creatures with protection from their mono color. Every time I've seen it not play stuff even though it could it was when I played, say, White Knight vs Mono Black or stuff like that
*Added Culling the Weak, Infernal Plunge, Bubbling Cauldron* up there with "Any sacrifice outlets*
---
Faeries - in theory you could make a striaghtforward "blue skies" thing with just faeries without flash or gimmicks, but that's like making a Rebel deck where the AI can't activate abilities.
In practice it mangles faeries beyond belief - never holds mana to cast them at the end of the opponents turn, casts them in it's second main phase, casts Spellstutter Sprite as a regular creature, spams Quickling and Faerie Impostor as soon as it can play them, then recasts whatever it bounced back, list probably goes on.
Homarid Spawning Bed - Currently crashes the game AFAIK.
---
Pulse Tracker - the AI is madly reluctant to attack with it.
Captain of the Mists - is bugged ATM in not letting you untap permanents other than creatures
Armageddon Clock is bugged in several ways and non-functioning
Restless Dead - the AI is counter-productively suicidal with it. Yes, I'm amazed too.
Corrupt - the AI tends to prefer casting it at creatures before even considering throwing it at the head.
Sign in Blood - the AI is very trigger happy about casting it at the opponent's head, which very likely isn't what it's there for and makes for spectacular missplays with concequences. Use Night's Whisper instead even if it's not on theme.
Agadeem Occultist - is currently bugged as if his ability was a free activated one instead of a tapping ability. It's a bit hilarious to see, but don't put it in decks, seriously XD
Artificer's Hex - straight up doesn't trigger during upkeep which is the whole point.
Fate Unraveller - it's a bit of a dick move to put one in any Shandalar deck (yours or AI), but I tried anyway, and the AI was madly agressive with it, attacking all the time, even into suicide.
If someone figures out how to make a deck revolving around these which I can try out and verify as working sensibly I would be thankful. "Occasionally stumbles into doing the right thing" doesn't cut it, card combinations which manage to force the AI to not do stupid things, or benefit from doing stupid things instead of embarrassing itself are where it's at.
As of NB1:
---
Equipment in general - Any number of things with it, spends all it's mana to move it around at any opportunity instead of casting spells, casts it instead of creatures on an empty board, doesn't understand that it can be more aggressive with Equipped creatures, takes them off things it will attack with, puts it onto stuff it won't etc.
Confirmed Examples:
Armory of Iroas
Prowler's Helm
Bone Saw
Lightning Greaves
Prowler's Helm - will occasionally work out in particular matches, but doesn't more often than not it won't.
You name it - sometimes equipment other than these works out fine after fine tuning the deck, but in general there's no telling what the AI will do with it.
Trepanation Blade - is just plain bugged and doesn't trigger.
Stuff that works, somewhat:
Leonin Scimitar (cost and activation cheap enough)
Bonesplitter (same)
Shuko (same)
Trusty Machete (Lujo: seems to kind of work fine in my current Merfolk Shaman tweaked deck. I'm not sure WHY, though.)
Darksteel Axe (Lujo: since it's possible to get the AI to use the Machete somewhat sensibly, I suppose this could also work)
Demonmail Hauberk (Lujo: You CAN make a Demonmail Hauberk work, except the AI doesn't realize it's easier to first put creatures down than it, and it costs 4. I put in a few hours trying to make a Rukh Egg / Dragon Egg deck with it, and it's quite uneven but looks like it could be made to work)
Kitesail can work out fine
Deckbuilding tips:
You tell me.
Lujo's got: Don't put any in. / Don't put too many in. / Don't put any but the first three mentioned ones, which have long since been deemed as too good by Wizards because they have too good cost/equip prices. But hey, with things they way they are now, if it works it works. / There's some way to make cc 1 equip c 2 +2/+1 equipment work, quite well, I'm not really clear on what yet.
---
Confirmed Examples:
Blood Bairn
Plagued Rusalka
Culling the Weak - Lujo: The AI has no idea about the importance of creatures it sacrifices, and I've even seen it sacrifice things with nothing to cast.
Bubbling Cauldron - Lujo: It will sacrifice everything and anything to it whenever it can, dear god.
Infernal Plunge - Lujo: Same thing as culling the weak
Stuff that consistently works in this vein:
No idea.
Deckbuilding Tips
Lujo: Don't bother using them? Use non-creature outlets which don't use creatures as the win condition. Hecatomb, for example, works rather well (although if you make a Shandalar deck with it it'll just murder things).
---
Creatures whith effects when they die / cannon fodder AI will chump-block with other stuff; keep them around forever; pump them instead of other things; sacrifice them to an outlet madly; value them over evasive things; it has little grasp of things being "sacrifice fodder".
Confirmed Examples:
Elder Cathar
Black Cat
Festering Newt
Festering Goblin
Things which work in this vein:
No idea
Deckbuilding tips:
Lujo: They all work if you put them in a deck which can't avoid killing them as collateral damage IF their on-death effects don't affect creatures. Example would be Onulet/Highland game in decks with Earthquake / Wrath of God.
---
Creatures which sacrifice themselves for effects - there's no telling when the AI will choose to activate it, and often very little obvious sense to it.
Confirmed Examples
Frostling
Goblin Firestarter
Sakura-Tribe Elder
Kami of False Hope (will get sacrificed at odd times)
Stuff that works somewhat better:
Kami of Ancient Law
Things that sacrifice to destroy artifacts (haven't tried them out yet, though)
Pain Kami
Deckbuilding Tips
You tell me.
---
Cantrips (cards which Draw a card / Scry in addition to something else they do) - The AI will treat them as card draw primarily (it'll rate the benefit of card drawing as so desirable that it won't care about wasting the rest of the card). This can also lead to trainwercks, and usually does. What it chooses to scry to the bottom of the library can also occasionally be mind-boggling, making giving it the effect risky and random.
Confirmed Examples
Spite of Mogis
Defiant Strike
You name it, pretty much. It won't always backfire, but it's often going to be a lousy move made in place of a better move, or not what you put the card in the deck for.
Stuff that still somewhat works:
Lujo: Possibly Repulse and counterspells which draw a card. It's still possible that it will bounce it's own creatures and counter it's own spells, though, I haven't checked.
Deckbuilding tips:
Lujo: Avoid like the plague / Give it only expensive cantrips. This defeats the purpose of cantrips somewhat, but reduces the opportunity for the AI to blow them truly mindlessly. / Don't try to give it "narrow" effects, as that can backfire
---
Removal - Burn that can target players gets thrown to the face if there are no targets which sucks in Green and Blue as well as Red; Bounce and kill is cast with utmost priority over anything else the AI could do including playing creatures to attack with; board sweepers will be spammed on a single target (even if the AI has a board advantage realistically); scaling removal will be chained even if waiting for one more land/creature/whatever-it-counts would let it kill something with one card if it waited just one more turn; things which do minor damage on top of a larger effect will be spammed mindlessly at first opportunity; removal with minor bonus effects like lifegain/no regeneration will be thrown madly (even at own creatures) for the effect; creature-only burn (especially sorceries) will be thrown before combat if it would reduce opponents effective toughness so that IF it blocks something it would die (but the AI sometimes still won't attack); etc. etc. etc. if you can imagine a stupid thing to do with removal besides flat out 1-for-1 killing it's own dudes it'll do it.
Confirmed examples:
You name it. The AI has a mechanism to determine whether it can pay for something that can be classified as removal, a way to tell you it has moved from it's hand to the stack, and an urge to do so if it has it in it's hand and there is a target. From the observers point of view that's what's gonna happen with any and all removal you give it, and every time that's a stupid move it's still gonna do it with seemingly anything you give it.
Stuff that still works somewhat:
Lujo: Spells that distribute damage among several targets (if they can also hit the player) like Arc Lightning, Forked Bolt etc. help red guys because they lessen the effect of wasting cards on foolishness (but are quite overkill as that's basically the most card-efficient burn ever);
Lujo: Expensive removal in general helps because the AI doesn't cast it over creatures so that when it plays the removal it has something to attack with (although this inflates mana curves and possibly creates slow hands).
Lujo: super-efficient-beyond-overkill removal that only targets creatures like Swords to Plowshares or cheap removal auras like Sensory Deprivation get cast early enough that it gets to play stuff instead of removal (although that's like solving someone's aim problems by giving it a bazooka for StP and the AI can be a bit fiddly with Auras).
Lujo: Generally removal up to 2 cc which doesn't target players might not slow it down too much or get wasted. It'll get madly chained if the AI draws a bunch, though.
Lujo: Creatures which do damage/kill stuff as comes-into-play effects like Fire Imp or Flametongue Kavu prevent it from playing removal when it should be playing creatures. Those are also madly efficient cards in any way immaginable, and FtK regularly makes lists of top 10 cards that should've never been made (by folks who know what they're talking about), so yeah, sure solves problems.
---
Wither - AI doesn't understand it can get aggressive with it on creatures, and will cast removal with wither ultra aggressively.
Confirmed examples:
You name it, Blight Sickle, Sickle Ripper, Corrosive Mentor
Works Somewhat:
Puncture Blast - it will get cast on a creature just to debuff it, but at least that's somewhat plausible and not always detrimental.
Deckbuilding tips:
You tell me.
---
Things which can gain flying - The AI currently seems to think that things which can gain flying always have flying so it suicide attacks with them on occasion.
Confirmed Examples:
Goblin Baloon Brigade
Manta Riders
Mystic Visionary
Mystic Penitent
Dross Hopper
---
Attacking - AI will only attack with the vast majority of non-evasive creatures (meaning no flying/landwalk/unblockable/etC) if there's any chance of losing a creature on attack, unless for a certain-looking alpha strike or if it has a pump in hand.
Confirmed examples:
You name it.
Things that work somewhat:
It will always attack with Erg Raiders, for some reason and things that must always attack. It will also somewhat often attack with things which can't block. Things with Deathouch, it they're cheap, will also madly attack. They don't "work" so much as "not not-work in this particular sense".
Deckbuilding tips:
Lujo: Err, well, there's not much you can do. It's the inverse of how MtG is played IRL, IRL "nobody blocks", here "AI doesn't attack". I'll probably write up what seems to work and what doesn't when I'm done tweaking the decks. This can't stay this way forever, it's preposterous.
---
Blocking - the AI seems to make any block which would result in a trade, no matter how horrible the trade. It seems to be, to the casual observer, extra-prone to this if the creature they are about to trade has summoning sickness. You can use any creature as a removal spell this way and it makes playing vs. it a mockery. It also tends to block for trades instead of blocking with 0 powered defenders. It also seems to have 0 brain when choosing what to trade or chump-block with and will often sacrifice evasive creatures in place of sacrifice fodder, or gang-block with several creatures when a single deathtouch creature would do.
Confirmed Examples
You name it. I bet Erg Raiders are present in so many decks because you can't bait the AI into losing them as soon as they hit the board due to their 3 toughness.
Stuff that works somewhat
Nothing will prevent the AI from attempting to do this.
Deckbuilding tips:
Lujo: Just read this, and think about what it means: "Give it things the player can't attack into, so that it won't have a chance to consider blocking". That's how horrible it is at it. There's quite a bunch of ways to do this, but then those things have to also be removal-proof, and they better also be indestructible, just in case, and...
Lujo: On a less defeated note, creatures with higher toughness than power (especially if agressively costed) seem to do the trick somewhat (think Erg Raiders). It makes it more difficult for the AI to lose to many creatures when it madly gang-blocks.
---
Playing MtG - The AI can't attack, block or use removal sensibly. It also can't time various spells and ability activations in a way which doesn't interfere with developing board position. It also can't avoid getting mana screwed via mulligan if it has only one land, or is forced to play it's worst hands. And it has no idea of what role which card is supposed to serve in it's deck or what it's trying to do other than universal, general guidelines.
Confirmed Examples:
You name it.
Stuff that actually works:
Giving it cards which are so fool-proof strong that it can still win despite being unable to attack, block or use removal sensibly. But are also so fool-proof simple that it can't possibly misuse them. And are also not caught in a crossfire of bugs/lunatic general idiosincracies. Also spending days and hours on deckbuilding.
Deckbuilding tips:
Lujo: Spend way more time playing, studying and working on MtG than people who've spent way too much time playing, studying and working on MtG for a quarter of a decade, then spend better part of your waking day on nothing but sifting through stuff until you find combinations the AI doesn't embarrass itself with too hard. Try not going mad, well, more than you already are.
Flanking on attack - AI can't really calculate flanking right, I sometimes won't attack with, say, Fallen Askari even though there's no way he'd die in combat, and no thing to do with it later as it can't block.
Confirmed Examples
Fallen Askari (will occasionally display this)
Stuff that Works Somewhat
Cadaverous Knight (as the AI will sometimes be more agressive with it due to regeneration)
Deckbuilding tips
You tell me.
Lujo: Buffs can help it still get a Flaneker attacking, but that's besides the point - flankers are very deady attackers with no buffs on them and REALLY hard to gang-block in combat with them, if it's not attacking with them it's kinda wasting them.
Creatures that come into play tapped - it's possible for the AI to make mistakes when determining whether it has a blocker or not with creatures that come into play tapped (black has a bunch). If it has one in hand it can think it will be able to block next turn if it plays it and it won't.
Confirmed Examples
Tormented Hero
Stuff that works somewhat
You tell me.
Deckbuilding tips
Lujo: No idea, but I suppose if you don't overdo them in a deck it won't crop up as often.
---
Mana Vault - the AI can't untap it.
Dark Ritual - the AI will generally blow it on anything at all, so don't expect it to get anything expensive out in the early turns.
Shaper Guildmage - AI won't attack with him or activate him in combat somehow. No idea about other guildmages and masters.
Magma Spray - even crazier behavior than most removal for some reason. Looks good if you're trying to attempt to have the AI not throw away it's 1cc cheap removal thing to burn the enemy to the face, it gets spectacularly creative about throwing it away for no gain at all.
Infest - AI is exceptionally weird about when to cast or not to cast this thing.
Kashi-Tribe Reaver - It's regeneration seems to be bugged, so he doesn't in fact regenerate. Or so it seemed to me.
---
Hex - The AI won't sacrifice it's own creatures as part of it, which is a crying shame as it'd make a perfect curve-topper for Witch. If it could use it intelligently, and sac Sengir Autocrat tokens to it :_(
Most "removal" auras - it will cast them post-combat and not benefit from casting them by attacking. It will also generally stack them on the same creature unless individual instances have been reported and addressed.
Whitewater Naiads - since the AI plays both them and removal auras which you'd want to pair up with them post-combat they're very difficult to get any use out of right now.
---
In the same vein as Whitewater Naiads:
Kami of Fire's Roar , Earthshaker - AI seems either unwilling or incapable of using this sort of thing because it generally plays creatures after combat, and these guys have effects which let you attack easier if you do your stuff pre-combat (one makes guys unable to block, the other flat out kills stuff).
---
Playing vs stuff with Protection from Color - the AI seems to go completely mad when it's facing creatures with protection from their mono color. Every time I've seen it not play stuff even though it could it was when I played, say, White Knight vs Mono Black or stuff like that
*Added Culling the Weak, Infernal Plunge, Bubbling Cauldron* up there with "Any sacrifice outlets*
---
Faeries - in theory you could make a striaghtforward "blue skies" thing with just faeries without flash or gimmicks, but that's like making a Rebel deck where the AI can't activate abilities.
In practice it mangles faeries beyond belief - never holds mana to cast them at the end of the opponents turn, casts them in it's second main phase, casts Spellstutter Sprite as a regular creature, spams Quickling and Faerie Impostor as soon as it can play them, then recasts whatever it bounced back, list probably goes on.
Homarid Spawning Bed - Currently crashes the game AFAIK.
---
Pulse Tracker - the AI is madly reluctant to attack with it.
Captain of the Mists - is bugged ATM in not letting you untap permanents other than creatures
Armageddon Clock is bugged in several ways and non-functioning
Restless Dead - the AI is counter-productively suicidal with it. Yes, I'm amazed too.
Corrupt - the AI tends to prefer casting it at creatures before even considering throwing it at the head.
Sign in Blood - the AI is very trigger happy about casting it at the opponent's head, which very likely isn't what it's there for and makes for spectacular missplays with concequences. Use Night's Whisper instead even if it's not on theme.
Agadeem Occultist - is currently bugged as if his ability was a free activated one instead of a tapping ability. It's a bit hilarious to see, but don't put it in decks, seriously XD
Artificer's Hex - straight up doesn't trigger during upkeep which is the whole point.
Fate Unraveller - it's a bit of a dick move to put one in any Shandalar deck (yours or AI), but I tried anyway, and the AI was madly agressive with it, attacking all the time, even into suicide.
---
My Shandalar deck pack folder is avaliable here:Dropbox
Leave feedback on particular decks here: Google doc
Ask for instructions, give feedback and complaints here: Thread
My Shandalar deck pack folder is avaliable here:Dropbox
Leave feedback on particular decks here: Google doc
Ask for instructions, give feedback and complaints here: Thread
Re: List of stuff that the AI can't handle
by lujo » 07 Nov 2015, 09:54
This deserves it's own post:
BUILDAROUNDS - Cards you put in decks, enchantments especially, which you build decks around. The combination of it's lunacy makes it unable to cash in on any of the sinergies present in this deck:
.13472 3 Lagonna-Band Elder
.13925 3 Lagonna-Band Trailblazer
.897 3 Windseeker Centaur
.189 4 Plateau
.188 8 Plains
.164 8 Mountain
.14118 3 Inferno Fist
.8303 3 Taste for Mayhem
.3053 2 Thundermare
.3528 4 Onslaught
.1605 2 Serra's Blessing
.229 3 Smoke
.13372 1 Anax and Cymede
.10250 3 Cerodon Yearling
.7881 2 Hunted Lammasu
.10355 2 Stun Sniper
.4199 2 Tethered Griffin
.5334 4 Battlefield Forge
That deck wants to do one thing - play it's stuff and then attack. The AI passes the attack phase, then plays it's stuff. Then does nothing. If you would like to build a deck around something that makes the AI more capable to attack other than making things unblockable or boosting their power and toughness to the point where the creature is effectively unkillable in combat - it won't work. Take that deck, play it for a few games, notice everything that goes on with you just playing stuff, there's a million things you can do and you don't have to think about any of it while you're playing - their stuff gets tapped down, your guys attack. Or you just attack and trade favorably.
Then give it to the AI. Then go cry like I want to. It will win, oh yes, IF it casts the auras on the fliers AND makes them vigilant AND it has no chance of ever losing a flier if it attacks.
Or take my recent Ape Lord deck, take out any utility and add 4 Fires of Yavimaya, play it a few times vs the AI, then give it to the AI and watch what happens.
Deckbuilding tips:
Don't make decks for Shandalar. Just don't. If you do, what you gotta do is: give the AI creatures which can't be blocked, and possibly can't be killed, either. Only give the AI blockable creatures if their stats are significantly higher then their casting cost and it's casting cost is reasonable. Always give the AI enough pump that it'll never lose anything in offensive combat, but make sure you also give it creatures it can't lose when it madly blocks with them, and also which can't die to deathtouch. Best course of action - only ever give it fliers, but also give it 4-8 straight up unblockable/horsemanship/shadow dudes in case it has to play against a player with many fliers. If the deck is too strong, you can easily nerf it - just give it some non-fliers which can't win the game on it's own.
Don't, under any circumstances whatsoever, give it an enchantment which works if it plays stuff in the pre-combat main phase, because it won't and the deck won't work. Don't try to build around anything - just give it all fliers, 4 unblockable, 4 pump spells, one thing to make the oppoenent unable to block with any creatures, and some removal.
And while we're at removal - either give it 20 removal and some bombs and 4 card draw, or only give it removal that kills 3 cards for one, or a creature which is also removal. Otherwise you just can't be sure. Really, you can't. Also, about 3-4 removal which costs more than anything else in the deck can work, just make sure it can't possibly play it before it's played everything else it can draw.
And speaking of pump spells - they need to be strong and cheap enough that nothing it has can possibly die and will surely kill something, but it also better draw a card because it will be cast as burn leaving it down a card if it doesn't just kill the opponent. A safer and more sensible choice would be Overrun - it won't attack anyway, so it might as well win right there if it does. Make sure to put Levitation in too, just in case.
BUILDAROUNDS - Cards you put in decks, enchantments especially, which you build decks around. The combination of it's lunacy makes it unable to cash in on any of the sinergies present in this deck:
.13472 3 Lagonna-Band Elder
.13925 3 Lagonna-Band Trailblazer
.897 3 Windseeker Centaur
.189 4 Plateau
.188 8 Plains
.164 8 Mountain
.14118 3 Inferno Fist
.8303 3 Taste for Mayhem
.3053 2 Thundermare
.3528 4 Onslaught
.1605 2 Serra's Blessing
.229 3 Smoke
.13372 1 Anax and Cymede
.10250 3 Cerodon Yearling
.7881 2 Hunted Lammasu
.10355 2 Stun Sniper
.4199 2 Tethered Griffin
.5334 4 Battlefield Forge
That deck wants to do one thing - play it's stuff and then attack. The AI passes the attack phase, then plays it's stuff. Then does nothing. If you would like to build a deck around something that makes the AI more capable to attack other than making things unblockable or boosting their power and toughness to the point where the creature is effectively unkillable in combat - it won't work. Take that deck, play it for a few games, notice everything that goes on with you just playing stuff, there's a million things you can do and you don't have to think about any of it while you're playing - their stuff gets tapped down, your guys attack. Or you just attack and trade favorably.
Then give it to the AI. Then go cry like I want to. It will win, oh yes, IF it casts the auras on the fliers AND makes them vigilant AND it has no chance of ever losing a flier if it attacks.
Or take my recent Ape Lord deck, take out any utility and add 4 Fires of Yavimaya, play it a few times vs the AI, then give it to the AI and watch what happens.
Deckbuilding tips:
Don't make decks for Shandalar. Just don't. If you do, what you gotta do is: give the AI creatures which can't be blocked, and possibly can't be killed, either. Only give the AI blockable creatures if their stats are significantly higher then their casting cost and it's casting cost is reasonable. Always give the AI enough pump that it'll never lose anything in offensive combat, but make sure you also give it creatures it can't lose when it madly blocks with them, and also which can't die to deathtouch. Best course of action - only ever give it fliers, but also give it 4-8 straight up unblockable/horsemanship/shadow dudes in case it has to play against a player with many fliers. If the deck is too strong, you can easily nerf it - just give it some non-fliers which can't win the game on it's own.
Don't, under any circumstances whatsoever, give it an enchantment which works if it plays stuff in the pre-combat main phase, because it won't and the deck won't work. Don't try to build around anything - just give it all fliers, 4 unblockable, 4 pump spells, one thing to make the oppoenent unable to block with any creatures, and some removal.
And while we're at removal - either give it 20 removal and some bombs and 4 card draw, or only give it removal that kills 3 cards for one, or a creature which is also removal. Otherwise you just can't be sure. Really, you can't. Also, about 3-4 removal which costs more than anything else in the deck can work, just make sure it can't possibly play it before it's played everything else it can draw.
And speaking of pump spells - they need to be strong and cheap enough that nothing it has can possibly die and will surely kill something, but it also better draw a card because it will be cast as burn leaving it down a card if it doesn't just kill the opponent. A safer and more sensible choice would be Overrun - it won't attack anyway, so it might as well win right there if it does. Make sure to put Levitation in too, just in case.
---
My Shandalar deck pack folder is avaliable here:Dropbox
Leave feedback on particular decks here: Google doc
Ask for instructions, give feedback and complaints here: Thread
My Shandalar deck pack folder is avaliable here:Dropbox
Leave feedback on particular decks here: Google doc
Ask for instructions, give feedback and complaints here: Thread
2 posts
• Page 1 of 1
Who is online
Users browsing this forum: No registered users and 24 guests