dkates
  • dkates
  • Advanced Member Topic Starter
2016-09-01T13:10:40Z
It's been a while since I did this, and I had some material, so here we go. Not really organizing this in any particular way.

  • Any plans to implement Vatlamyus for the tellarknight deck?
  • The AI will frequently attack face-up Defense Position monsters with monsters that have the same ATK as the target's DEF. While this makes sense for the Gladiator Beasts, for any other deck, it's a waste of an attack and unnecessarily gives the player opportunities to activate cards that respond to attacks.
  • I've reported this before, but the AI will still frequently attack into monsters that have the effect to prevent the AI from activating effects during the battle (the prime example being Utopia the Lightning), while clearly having the intent to activate one or more effects during that battle. If the AI is able to recognize that monsters have that effect, then this should be possible to code generically. If (as I suspect is the case) the AI cannot see this, it should at least be able to check for the most commonly played monsters that have this effect.
  • In the Blackwings at least, the AI will still frequently waste its Instant Fusion and Norden by not doing anything with Norden or the monster it brings out, despite having the ability to do so. It may be possible to code this one generically -- basically, if the AI uses Instant Fusion and brings out Norden, have it store the fact that it did that and the monster it brought back with Norden (if any), and have it avoid leaving Main Phase 2 before using the Norden and/or the monster it brought back for something, unless it is unable to do so. For the Xyz-based decks, it usually uses Norden just fine, it just doesn't seem to be as good about it in the decks that are more Synchro-based.
  • On my end, the Burning Abyss deck has only 38 cards. I don't know if this is specific to me, or if the deck file got changed in general and was left unfinished. It needs to be updated for the new banlist anyway, though since it only loses 1 Cir and 1 Beatrice, it shouldn't be too difficult to update. I doubt that the script would need any particular updates. (EDIT: Resolved the 38-cards thing by re-downloading the AI decks. The copy of the file I had was missing the Painful Escapes.)
  • In general, if the AI's Mystical Space Typhoon is about to be hit with a removal effect, Mystical Space Typhoon will be activated (as it should), but it needs a bit of tweaking to its targeting for these instances. Usually, if whatever is before Mystical Space Typhoon in the Chain is something that can be targeted with that Mystical Space Typhoon, the AI will do so, even if that means that it is targeting its own card, or a card the player controls that would not have stayed on the field anyway (such as the player's own Mystical Space Typhoon or Twin Twisters), and the player has S/T that would be more appropriate targets. I don't recall if this also applies to Artifact Ignition.
  • Speaking of Twin Twisters, any plans to implement Twin Twisters for any of the AI decks? Some of them (such as the Majespecters) might be better off sticking with Mystical Space Typhoon, but a lot of them would probably benefit from the switch.
  • The Kozmo deck needs to be updated for the new banlist (it loses 2 Dark Destroyer and 1 Emergency Teleport). Unlike with the Burning Abyss, this one might require some updates to the script, to account for the fact that Dark Destroyer is now Limited. (I don't expect that there would be any need to adjust the script to account for the lost E-Tele.)
  • The Masked Heroes deck will chain Mask Change to effects that would remove a HERO from the field, which is fine... except that it will do this even if the effect is chaining to would instead end up just killing whatever it brings out. It should avoid chaining Mask Change to effects such as Dark Hole, Raigeki, and Torrential Tribute, at least if it would be targeting a Shadow Mist or Neos Alius; if it would be targeting a Bubbleman, and the player has S/T that Acid would usefully destroy, it might still want to use Mask Change in those cases. In the case of one of the Mirror Force variants, if it does chain Mask Change, it should bring out the new monster in defense.
  • The Mega Monarchs lose a lot of cards in the new update, and will probably need a script update to account for some of those lost cards. The lost Ehthers are probably not going to cause any particular need for script updates, but the lost Spells might.
  • In the Qliphort deck, the AI will frequently attack with a Qli that's currently at 1800 ATK, clearly intending to use Forbidden Chalice so that the Qli will win the battle, but not activate the Forbidden Chalice at the appropriate time even though it still has the ability to do so. This seems to mostly happen if the attack target is in defense. Also, does the AI have the ability to "Chalice for game"? (That is, use Forbidden Chalice on a Qli in cases where the current ATK is not enough for lethal, but the updated ATK would be. The Blackwing deck already does this for Kalut, though I don't think it checks if it would have had lethal even without the Kalut.)
  • I've reported this before, but the Frognarchs deck will frequently go into an Extra Deck monster, then Tribute the Extra Deck monster for a Tribute Summon, when it would have made more sense to just Tribute one of the things it ended up using as Material. Also, that deck loses two copies of The Monarchs Stormforth with the new banlist, so it will need some changes to its Spell lineup. I may just do this myself on my end, but still. With Thousand-Eyes Restrict no longer Limited, and the deck now having some slots it will need to fill, Instant Fusion into Thousand-Eyes might be worth implementing for this. (I don't know if it would already use it correctly; I know that at least one AI deck does have both Instant Fusion and Thousand-Eyes, but I'm assuming the AI wouldn't know to then use Thousand-Eyes for a Tribute Summon or Extra Deck Summon if possible.)
  • You've mentioned before that you intend to implement Gold Gadget and Silver Gadget for the Gadgets deck, and update the decklist accordingly. Any progress on that? You mentioned that you would probably either switch over to pure Gadgets, or keep it as a Machina/Gadget. Personally, I'd suggest keeping it as Machina/Gadget. I don't expect the new Gadgets to need too complex of a script for their use, mostly just priority lists for the effects. (Extra credit if it can be made to adjust its priority on the colored Gadgets based on whether it has the appropriate Gadget in the Deck to be searched with the one it Summons.)
  • Nothing that requires a change, but I do find it a bit amusing that all of the AI decks that were using Maxx "C" had two copies of it, so in that sense, they're still legal with the new list.
  • You've mentioned before that the AI chooses which Position it Summons a monster in based on a table. For monsters where it defaults to Attack Position, but Defense Position is a legal option, would it be possible to add a check to see if bringing out the monster in Attack Position would give the player lethal, and if so, have the monster come out in Defense Position instead? Obviously, even this wouldn't be a perfect solution, but it should help some.
  • Would it be possible for the AI to be able to see when using its cards would come with a penalty that is normally not present? For example, if the player has Red-Eyes Flare Metal Dragon with Xyz Material, have the AI account for the fact that it will take 500 damage for each card/effect activation that doesn't get that Red-Eyes Flare off the field. Based on what has been said about the AI before, I'd expect this one to be difficult at best, but it would be nice to have if it can be done.
  • The AI seems to always assume that the player's monsters are all valid attack targets. Would it be possible to add something to the attack logic that has the AI check which of the player's monsters actually are valid attack targets, and if none of those are ones it wants to attack, refrain from attacking (when possible)? As is, the AI will often attack into the player's monsters that it shouldn't want to attack, because the monster it wanted to attack is not a valid attack target.
Snarky
2016-09-10T13:22:28Z
Obviously, pretty much all of these are valid points I'll need to work on, so instead of addressing them all one-by-one, let me answer a little broader: Many of these issues are already attempted to be handled by the system, but a lot of the time, those solutions don't work correctly, or just don't cover all relevant situations.

Priority for now is modifying the decks for the newest banlist. I might change some decks a little more than necessary, for example I am thinking about switching the BA deck to the popular PK Fire build, utilizing Phantom Knights and probably the Speedroid engine. Not sure, what I will do with the Monarch variants yet, I'll look for some options first.

Of course, I did add these issues to my ever-growing todo list, and hopefully I can address all these points at some point 🙂
dkates
  • dkates
  • Advanced Member Topic Starter
2016-09-10T20:37:44Z
Sounds good. If there's anything I can do to help, please feel free to let me know.

On the Burning Abyss, I'm honestly not sure whether to recommend keeping it as mostly-pure Burning Abyss, going the PK Fire route, or splitting the PK Fire off into its own AI deck but still updating the Burning Abyss build as well. Obviously, that last option would be the most work for you, whereas simply updating the Burning Abyss without adding the Phantom Knight and/or Speedroid engines would be the least, but it's ultimately up to you.
Snarky
2016-09-11T06:44:02Z
Originally Posted by: dkates 

Sounds good. If there's anything I can do to help, please feel free to let me know.


Alright, you fix the first half of the issues, I'll do the second half :P
Quote:


On the Burning Abyss, I'm honestly not sure whether to recommend keeping it as mostly-pure Burning Abyss, going the PK Fire route, or splitting the PK Fire off into its own AI deck but still updating the Burning Abyss build as well. Obviously, that last option would be the most work for you, whereas simply updating the Burning Abyss without adding the Phantom Knight and/or Speedroid engines would be the least, but it's ultimately up to you.


I would probably go with swapping to PK Fire, and keeping the other variant in the "old decks" archive, so if anyone really wants to play that, he can get it there. I feel like PK Fire and BA would be too similar to include both of them by defaut. Granted, they are quite different... not sure yet.
dkates
  • dkates
  • Advanced Member Topic Starter
2016-09-19T17:01:57Z
Originally Posted by: Snarky 

Originally Posted by: dkates 

Sounds good. If there's anything I can do to help, please feel free to let me know.


Alright, you fix the first half of the issues, I'll do the second half :P



If only I could. :P

In all seriousness, though, I'm grateful for all the work you do on the AI scripting, and I'm glad that you apparently have taken all of my reports and suggestions in the light they were intended, as an honest effort to help the AI improve.

I can't say for sure unless I learn to understand the AI script and look at it in detail, of course, but I wouldn't be surprised if adding the Phantom Knight engine to the Burning Abyss (AKA making it into PK Fire) basically just required figuring out the proper decklist and scripting the Phantom Knight plays (which, luckily, tend to be pretty straightforward).
Snarky
2016-09-20T07:09:39Z
Originally Posted by: dkates 


I can't say for sure unless I learn to understand the AI script and look at it in detail, of course, but I wouldn't be surprised if adding the Phantom Knight engine to the Burning Abyss (AKA making it into PK Fire) basically just required figuring out the proper decklist and scripting the Phantom Knight plays (which, luckily, tend to be pretty straightforward).


I did work on PK Fire a little already, and indeed you are correct for the most part. The Speedroid engine is pretty standalone and easy to use for the AI (well, it took me forever to implement, but only because I am stupid and messed up a lot). The PK plays are pretty seperate as well, basically the only challenging part seems to be integrating it into already existing BA plays. Essentially the question, when is it okay to trigger BAs selfdestruct on the field, and when it is not. For now the AI is very reluctant to do it, and might miss out on good plays because of it. A smart player might notice and abuse this behavior.

I'll have to confess, I got a little bored with fixing all these issues and updating the decklists, though, and I might have accidentally started a new deck instead [:doh:]. Shame on me, right?

Similar Topics
Users browsing this topic