Card AI (Improvements) Requests
by mtgrares
Moderators: timmermac, Blacksmith, KrazyTheFox, Agetian, friarsol, CCGHQ Admins
Re: Card AI (Improvements) Requests
by Sloth » 18 Jul 2012, 20:53
Fixed! Thanks Diogenes.Diogenes wrote:Here's another one: the AI just enchanted a Misty Rainforest with Leafdrake Roost, and then as soon as the attachment resolved (same phase) sacrificed the Misty Rainforest. The choice of land for the Roost wasn't bad, actually, since it didn't need any mana fixing.
Fixed! Thanks Milod. (Note that AI handling of exalted still needs some work).Milod wrote:duty-bound death most of time doesent attack even if i dont got any monsters in play or only got a planeswalker in play it could attack for 1 because it has exalted plus ai has exalted lands in play for more attack bonus
-
Sloth - Programmer
- Posts: 3498
- Joined: 23 Jun 2009, 19:40
- Has thanked: 125 times
- Been thanked: 507 times
Re: Card AI (Improvements) Requests
by Milod » 19 Jul 2012, 09:49
when there are no more swamps to search for lililan of dark realms ability she always stays on 8 loyality vs my deck with no creatures it could at least use her ultimate for 7
Re: Card AI (Improvements) Requests
by Milod » 19 Jul 2012, 10:42
ai doesent make use of faith reward
Re: Card AI (Improvements) Requests
by Diogenes » 19 Jul 2012, 12:59
I've noticed that the AI regularly pays two life to bring pain-lands into play untapped, but the majority of the time it does not use those lands on that turn. Quite often I suspect it can't afford to play anything in its hand. I'll use dev mode to check its hand the next time it does this.
Edit: Got it in one! First turn AI drops an Overgrown Tomb with nothing under cmc 2 in hand.
Edit2: Next game, though, with two lands in play it allowed an Overgrown Tomb to come in tapped. This is unfortunate as the cheapest card in its hand (Tatterkite) is , and it's sitting on two useful four-cmc cards. This is a situation where it should take the two damage. It looks like it'll take the damage to get an early mana base down, but can't determine whether the land would enable it to play cards from its hand. Would something like this be possible Sloth?
Edit: Got it in one! First turn AI drops an Overgrown Tomb with nothing under cmc 2 in hand.
Edit2: Next game, though, with two lands in play it allowed an Overgrown Tomb to come in tapped. This is unfortunate as the cheapest card in its hand (Tatterkite) is , and it's sitting on two useful four-cmc cards. This is a situation where it should take the two damage. It looks like it'll take the damage to get an early mana base down, but can't determine whether the land would enable it to play cards from its hand. Would something like this be possible Sloth?
Re: Card AI (Improvements) Requests
by Sloth » 20 Jul 2012, 16:40
Fixed! Thanks Diogenes.Diogenes wrote:Here's another one: the AI just enchanted a Misty Rainforest with Leafdrake Roost, and then as soon as the attachment resolved (same phase) sacrificed the Misty Rainforest. The choice of land for the Roost wasn't bad, actually, since it didn't need any mana fixing.
Fixed! Thanks Milod.Milod wrote:duty-bound death most of time doesent attack even if i dont got any monsters in play or only got a planeswalker in play it could attack for 1 because it has exalted plus ai has exalted lands in play for more attack bonus
This was some very old code that determined randomly to pay the life or not. The AI will now never pay the life until someone wants to add some smart code (maybe when RtR comes out).Diogenes wrote:I've noticed that the AI regularly pays two life to bring pain-lands into play untapped, but the majority of the time it does not use those lands on that turn. Quite often I suspect it can't afford to play anything in its hand. I'll use dev mode to check its hand the next time it does this.
Edit: Got it in one! First turn AI drops an Overgrown Tomb with nothing under cmc 2 in hand.
Edit2: Next game, though, with two lands in play it allowed an Overgrown Tomb to come in tapped. This is unfortunate as the cheapest card in its hand (Tatterkite) is , and it's sitting on two useful four-cmc cards. This is a situation where it should take the two damage. It looks like it'll take the damage to get an early mana base down, but can't determine whether the land would enable it to play cards from its hand. Would something like this be possible Sloth?
-
Sloth - Programmer
- Posts: 3498
- Joined: 23 Jun 2009, 19:40
- Has thanked: 125 times
- Been thanked: 507 times
Re: Card AI (Improvements) Requests
by muppet » 20 Jul 2012, 17:22
I'd have thought a default of always paying the life was better actually ?
Re: Card AI (Improvements) Requests
by friarsol » 20 Jul 2012, 18:36
If there's going to be absolutely no logic, it's better if the AI can't kill themselves with the default choice.muppet wrote:I'd have thought a default of always paying the life was better actually ?
- friarsol
- Global Moderator
- Posts: 7593
- Joined: 15 May 2010, 04:20
- Has thanked: 243 times
- Been thanked: 965 times
Re: Card AI (Improvements) Requests
by cc-drake » 20 Jul 2012, 20:01
AI shouldn't activate the Spikes (like Spike Weaver) to move +1/+1 counters onto itself...
Re: Card AI (Improvements) Requests
by ZzzzSleep » 21 Jul 2012, 00:03
Unless it has Doubling Season in play, then it's an awesome move! But if we don't want to check for that, then defaulting to not "moving" a counter to itself sounds like a good idea.cc-drake wrote:AI shouldn't activate the Spikes (like Spike Weaver) to move +1/+1 counters onto itself...
Re: Card AI (Improvements) Requests
by muppet » 21 Jul 2012, 10:33
It can't kill itself you can't pay life you don't have.
Re: Card AI (Improvements) Requests
by Sloth » 21 Jul 2012, 10:54
If a player has 2 life he can pay 2 life.muppet wrote:It can't kill itself you can't pay life you don't have.
-
Sloth - Programmer
- Posts: 3498
- Joined: 23 Jun 2009, 19:40
- Has thanked: 125 times
- Been thanked: 507 times
Re: Card AI (Improvements) Requests
by muppet » 21 Jul 2012, 12:56
ahh right good point. Anyway my point is if the shock lands read these come into play tapped always they would never see constructed play.
How about a super easy algorithm for example if it is in the first 5 turns of the game always pay the life if it is later don't.
How about a super easy algorithm for example if it is in the first 5 turns of the game always pay the life if it is later don't.
Re: Card AI (Improvements) Requests
by friarsol » 21 Jul 2012, 14:06
In the WG Easy deck, when the AI plays Journeyer's Kite it pretty much doesn't do anything else for the next 4-5 turns. Since this is a reusable instant ability, we probably should reduce how often it gets played during the AIs turn, since it doesn't do anything to the board at all.
- friarsol
- Global Moderator
- Posts: 7593
- Joined: 15 May 2010, 04:20
- Has thanked: 243 times
- Been thanked: 965 times
Re: Card AI (Improvements) Requests
by cc-drake » 21 Jul 2012, 18:46
AI shouldn't cast Sculpting Steel when there is no artifact on the battlefield
Re: Card AI (Improvements) Requests
by Sloth » 22 Jul 2012, 18:11
I've added some code that will make the AI pay the 2 life sometimes:muppet wrote:ahh right good point. Anyway my point is if the shock lands read these come into play tapped always they would never see constructed play.
How about a super easy algorithm for example if it is in the first 5 turns of the game always pay the life if it is later don't.
If the AI controls X lands (with the new Shockland) and has a card with converted cost X in hand it will pay the 2 life.
That should be acceptable, no?
This should be fixed with r16299.friarsol wrote:In the WG Easy deck, when the AI plays Journeyer's Kite it pretty much doesn't do anything else for the next 4-5 turns. Since this is a reusable instant ability, we probably should reduce how often it gets played during the AIs turn, since it doesn't do anything to the board at all.
-
Sloth - Programmer
- Posts: 3498
- Joined: 23 Jun 2009, 19:40
- Has thanked: 125 times
- Been thanked: 507 times
Who is online
Users browsing this forum: No registered users and 152 guests