dkates
  • dkates
  • Advanced Member Topic Starter
2015-11-03T13:49:24Z
Well, here we go again. [:tongue:] As I so often do, I'm listing here some issues, suggestions, and such for the AI decks and/or script. As the thread title indicates, I'm currently using AI script version 0.32a, on YGOPro version 1.033.7. Since it's already been mentioned that the AI decks will be updated for the new banlist, I'm not going to say too much about that here. I should note, though, that I haven't done much testing with 0.32a; most of these comments are resulting from testing with 0.32. As such, it is possible that I will note some things that were changed between the two versions, though based on the notes for 0.32a, it doesn't look like much was changed apart from adding the Kozmo deck.

General:
  • For the decks that can bring out Level 8 Synchros, it might be worthwhile to implement PSY-Framelord Omega.
  • In some situations, the AI will activate cards with a Life Point cost that will cause it to pay all of its Life Points (which wasn't an issue until the OCG ruling that while you can't pay a Life Point cost that's greater than your current LP, you can pay one that's exactly equal to your current LP, you'll just lose the game). Offhand, I can only recall seeing this with Instant Fusion.
  • The AI still seems to not pay enough attention to its Life Points and the field situation when deciding what position to Summon monsters in and whether to Normal Summon monsters versus Setting them.
  • I'd have to retest to be sure, but I seem to recall the AI apparently being unaware that Burgesstomas that are being treated as monsters are not affected by monster effects, leading to it trying to use monster effects on them with no success.


Battlin' Boxers:
  • The deck list itself could use an update. In particular, in addition to updating it for the new banlist (which has already been mentioned to be on the docket), I'd suggest taking out Jolt Counter. Some already-scripted cards could do well in some of the slots that will be freed up, though it might also be worthwhile to put in Heroic Challenger - Thousand Blades. I've never really played Battlin' Boxers much, so I don't feel qualified to go too in-depth as far as suggestions on this front.


Blackwings:
  • Already mentioned (and responded to) in a previous report, but this deck could very much use Assault Blackwing - Raikiri the Sudden Shower in its Extra Deck.
  • It would probably require some changes to the script, but the deck could probably make some use of Harpie Harpist. Not sure what to suggest taking out, though, as this deck was not affected by the new banlist and the current Main Deck is already pretty solid.


Bujins:
  • This deck seems to keep revealing issues in the attack logic, probably because the Bujins have so many more combat tricks than most of the AI decks. In particular, being able to better handle multiple combat tricks in one battle would be a huge help. Also, the AI sometimes doesn't use the combat tricks when attacking a Defense Position monster, resulting in it taking unnecessary rebound damage.


Burning Abyss:
  • I believe that I have mentioned this before, but the AI will activate the Graveyard effect of Calcab if able, even if the only available targets are on the AI's side of the field. The AI should only use this effect to target S/T on the player's side of the field, and should not activate the effect if the player does not have any set S/T.
  • I don't play Burning Abyss myself, so take this one with a grain of salt -- but I don't really see Barbar as being all that useful here, at least in terms of its Graveyard effect. I could maybe see it if this were running Leviair the Sea Dragon and/or Burial from a Different Dimension, but since it isn't, Barbar seems to run counter to the way Burning Abyss works.
  • I know that others have mentioned this, but Fiendish Rhino Warrior should be a useful card for this, to replace some of the cards it lost to the new banlist.


Chaos Dragons:
  • I find myself wondering if Peropero Cerberus should be considered for this deck. Sure, it's not LIGHT or DARK, and it's not a Dragon, but given this deck's focus on filling the Graveyard, Peropero could be useful as a way to benefit from the milling.
  • This deck's Extra Deck should probably include Minerva, Lightsworn Saint.
  • The logic for choosing positions on Special Summons could probably use some tweaking. For example, if the player has a monster(s) that the AI doesn't expect to be able to get rid of that turn, it can Special Summon Chaos Sorcerer, Collapserpent and/or Wyverburster in defense, to allow them to act as blockers.


Constellars:
  • I can't help but feel that this deck could use a slightly higher monster count. In particular, I know I said this in an earlier report, but Constellar Siat should be useful here, in place of some of the backrow cards. Also, maybe this deck could use Summoner Monk? I don't know for sure, and it would probably take some tweaking, but a few of the monsters in this deck could be useful when brought out via Summoner Monk.


Dark World:
  • Not much to say here, other than that the deck could probably make use of some of the Galaxy Xyz in its Extra Deck. (Either 107 or 62, going into Full Armor, and possibly from there into 95 if there's enough room in the Extra.)


Frognarch:
  • Could use some minor tweaks to the logic with Light and Darkness Dragon. Other than that, seems to play pretty well.


Gadget:
  • Could use some updates to its Extra. Otherwise, seems more or less fine. I could theoretically see this being rebuilt to use Pendulums, though that would take a lot of changes to the deck list and a fair amount of scripting.


Harpie:
  • I'd have to retest, but I seem to recall this deck (and only this one) sometimes reacting to its own effects as if they had been activated by the player, causing it to use things such as Maxx "C" and Breakthrough Skill prematurely.


HAT:
  • Could use some updates in its deck list. In particular, I'd be interested to see this incorporate Time-Space Trap Hole. It might also be able to make use of Naturia Sacred Tree, though that might require also adding in some additional monsters that would be compatible with it, and I honestly don't know how well that works.


Heraldic:
  • I've seen you responding to another report suggesting that this get a decklist update. Really, that's all this should need.


Mega Monarch:
  • For my own use, I've been experimenting with a build that main decks Zaborg the Mega Monarch, and includes 3 each of Elder Entity N'Tss and PSY-Framelord Omega in its Extra Deck, with some success. The idea is to use Zaborg to destroy a LIGHT monster (in order of preference, that would be an opponent's monster, The Original Monarch, or Zaborg itself) sending N'Tss and Omega to the Graveyard in order to make use of their effects. (My build also uses Miracle Synchro Fusion and some appropriate Synchros and Fusions for it, though I would probably not recommend that part here.) It does mean that Dominion will sometimes not be locking down the opponent's Extra Deck when it otherwise would, but the tradeoff is that the deck gains an option for dealing with multiple cards on the opponent's field at once and the ability to mess with the opponent's Graveyard. Do you think that play could be implemented here?


Noble Knight:
  • Sometimes makes plays that seem to anticipate it equipping a Noble Arms card, when it doesn't have one available. Might want to have it check for that, and adjust accordingly. Not really sure what to suggest, though. Even for human use, this deck tends to struggle when it gets bad hands.


Qlis:
  • Will sometimes activate Re-Qliate when it doesn't have any other face-up Qli cards, leading to Re-Qliate being wasted.
  • Will sometimes attack a monster clearly anticipating that it will use something to negate its attacking monster's effect (Chalice, I'd assume), but then not do it.
  • I don't play Qlis myself, so I don't know the answer to this, but considering that this deck only rarely goes into its Extra Deck anyway, might it be helpful to remove the Extra Deck entirely and incorporate some of the appropriate Monarch S/T? Many of the Qlis do have the Monarch ATK/DEF stats.


Well, that's it for now, though I will either make further posts on this topic or make a new topic (as appropriate) if/when I find other things to comment on.
Snarky
2015-11-03T15:31:42Z
Oh my...

Not commenting on all of it, some of them I already answered, others I just don't have an answer to right now.

Originally Posted by: dkates 

  • For the decks that can bring out Level 8 Synchros, it might be worthwhile to implement PSY-Framelord Omega.


  • Definitely, at some point. Not the easiest card to use for the AI, though.
    Quote:


  • In some situations, the AI will activate cards with a Life Point cost that will cause it to pay all of its Life Points (which wasn't an issue until the OCG ruling that while you can't pay a Life Point cost that's greater than your current LP, you can pay one that's exactly equal to your current LP, you'll just lose the game). Offhand, I can only recall seeing this with Instant Fusion.


  • What a stupid ruling, right? -.-
    I fixed most of these, I think, but some might still be there. I'll look after Instant Fusion. Also, obviously, that will affect any non-scripted card with a LP cost as well, and unfortunately I cannot check for costs in a generic sense, so that issue is here to stay.
    Quote:


  • The AI still seems to not pay enough attention to its Life Points and the field situation when deciding what position to Summon monsters in and whether to Normal Summon monsters versus Setting them.


  • Yeah, most monsters don't check for anything but a list, which monster should be summoned in what position. If I factor in current life points, current ATK of opponent's monsters and especially, if and which of these monsters the AI expects to be able to remove, or if the AI needs to pay more life points this turn, it just gets overwhelmingly complex. I have no idea, how to do that properly, so I'll probably just stick to the current logic for now.
    Quote:


  • I'd have to retest to be sure, but I seem to recall the AI apparently being unaware that Burgesstomas that are being treated as monsters are not affected by monster effects, leading to it trying to use monster effects on them with no success.


  • Maybe. Will look into it.

    Quote:

  • It would probably require some changes to the script, but the deck could probably make some use of Harpie Harpist. Not sure what to suggest taking out, though, as this deck was not affected by the new banlist and the current Main Deck is already pretty solid.


  • The guys at pojo disagreed on this mostly, Harpist was deemed too slow for a fast swarm/OTK deck like Blackwings. Some of them don't even play Pinaki anymore, which I thought to be an amazing card for the archetype, but end phase adds are largely unneeded, if you have access to 3 Whirlwind.

    Quote:

  • I believe that I have mentioned this before, but the AI will activate the Graveyard effect of Calcab if able, even if the only available targets are on the AI's side of the field. The AI should only use this effect to target S/T on the player's side of the field, and should not activate the effect if the player does not have any set S/T.


  • The intend was to re-enable the BA self-summon effects. The execution is done poorly, but it can be useful to return your own S/T.
    Quote:


  • I don't play Burning Abyss myself, so take this one with a grain of salt -- but I don't really see Barbar as being all that useful here, at least in terms of its Graveyard effect. I could maybe see it if this were running Leviair the Sea Dragon and/or Burial from a Different Dimension, but since it isn't, Barbar seems to run counter to the way Burning Abyss works.


  • Barbar is just a name in the early game, good enough for that initial Dante. And lategame he is an awesome finisher. 900 damage just for sending him to the grave is not bad at all. Others run Cowboy in the extra for this, arguably way harder to make use of, and weaker burn effect. The AI doesn't use his effect early to not waste resources, and I have found myself being finished by this card more than once. Definitely good enough as a 1-of, in my opinion.

    Quote:


  • The logic for choosing positions on Special Summons could probably use some tweaking. For example, if the player has a monster(s) that the AI doesn't expect to be able to get rid of that turn, it can Special Summon Chaos Sorcerer, Collapserpent and/or Wyverburster in defense, to allow them to act as blockers.


  • As mentioned, this "expect to be able to get rid of" part is actually very hard to check.

    Quote:


    Mega Monarch:

    • For my own use, I've been experimenting with a build that main decks Zaborg the Mega Monarch, and includes 3 each of Elder Entity N'Tss and PSY-Framelord Omega in its Extra Deck, with some success. The idea is to use Zaborg to destroy a LIGHT monster (in order of preference, that would be an opponent's monster, The Original Monarch, or Zaborg itself) sending N'Tss and Omega to the Graveyard in order to make use of their effects. (My build also uses Miracle Synchro Fusion and some appropriate Synchros and Fusions for it, though I would probably not recommend that part here.) It does mean that Dominion will sometimes not be locking down the opponent's Extra Deck when it otherwise would, but the tradeoff is that the deck gains an option for dealing with multiple cards on the opponent's field at once and the ability to mess with the opponent's Graveyard. Do you think that play could be implemented here?


    That is a completely different variant and would require quite some changes to the build, as having an extra clashes with Dominion, arguably one of the most important cards in the current AI build.

    If I add a Zaborg build, it will be a different deck, since that plays out in a completely different way. However, I will work on other decks first, before I implement a 3rd variation of Monarchs :D

    Thanks again for your extensive notes. Did you have a change to try out the Kozmo deck yet?

    dkates
    • dkates
    • Advanced Member Topic Starter
    2015-11-03T16:51:21Z
    Originally Posted by: Snarky 

    Thanks again for your extensive notes. Did you have a change to try out the Kozmo deck yet?



    I've only dueled against it once so far, but based on what I saw in that duel, it looks like the AI plays the deck pretty well, and looking at the deck list, it looks like a very good build.

    Also, a minor general note that should be easy enough to take care of. I assume that Solemn Notice has a blacklist. I'd suggest adding the Graveyard effect of PSY-Framelord Omega to that blacklist.
    Snarky
    2015-11-03T17:28:35Z
    Originally Posted by: dkates 


    Also, a minor general note that should be easy enough to take care of. I assume that Solemn Notice has a blacklist. I'd suggest adding the Graveyard effect of PSY-Framelord Omega to that blacklist.



    I am currently preparing a general overhaul of negation effects. Basically, I want to group them into categorys and distribute negation priorities based on the properties of the activated effect. No idea, how this will turn out, since the information I can get on the effects is kinda limited, but I will probably factor in stuff like the location of the activation, the targets, the category of effect, if available, and the specific situation, to a degree.

    For example, activations on the field are usually more dangerous, and many negation cards also destroy the card on the field, so prefer to negate those over effects, that activate in grave or banished zone. Draw and search are not as dangerous as a summon. Removal or mass removal are even worse. If the AI is currently in the process of attacking for game, it should probably negate more stuff. If the opponent is low on cards, probably consider giving search and draw effects a higher negation priority. Also, Breakthrough Skill is not as valuable as Solemn Notice, so BTS should be used for less dangerous effects as well, especially the graveyard effect. Then we add in banlists or explicit priorities for specific effects, that would be mishandled by these rules.

    Kinda arbitrary, but we'll see, how it works out.

    Similar Topics
    Users browsing this topic