I'm about done with Siege until Hex makes some AI improvements

    • I just spend more than hour attacking and honestly the idea even in current implementation is not that bad. I have an idea what players will put up absolutely deadly sieges, so I did not go for those, but the medium ones were absolutely reasonable. You spend 1-3 tries attacking the siege to understand the decks that are against you, you tune your deck a little bit or bring a different one and then get some rewards. I met old friend like Killipede or Reese and was pretty amused with creativity some of the players put up.

      As written above the incentives currently are really badly set and this is valid especially for defenders (I would say attackers are pretty fine if you take basic gold amounts committed). Purely monoruby sieges just have to stop, because that's not fun to play against and if the system does not guarantee it, it needs to be changed. If HEX is willing to award 10k+ gold per hour for PvE content, it should also do so for sieges. Honestly, nobody minded AI in PvE that much, it's the whole incentive system that is putting up posts like this. I wish there was official post saying, we are aware and we are thinking about ways to fix it, oh well...
    • Sethanon wrote:

      If HEX is willing to award 10k+ gold per hour for PvE content, it should also do so for sieges. Honestly, nobody minded AI in PvE that much, it's the whole incentive system that is putting up posts like this. I wish there was official post saying, we are aware and we are thinking about ways to fix it, oh well...

      Oh, absolutely. Almost nobody would care at all if this mode were a gold GENERATOR (as it should have been) instead of a gold and platinum sink. Then the bad AI would just be funny and cute and we'd all laugh at how our silly defenses got beaten for dumb reasons. It's the fact that it costs money AND is bad combined that makes it all inexcusable and greedy and stupid.

      This mode should have been a fun, casual, entertaining, minorly profitable source of dynamic content that the community could have used to keep ourselves occupied practically indefinitely while HexEnt got their asses in gear and finished the rest of all the missing pieces of this game and cleaned up all the existing pieces of it that either suck or look like shit. Instead, it's yet another insulting source of frustration for everyone because of what could have been, while the company greedily sucks every bit of cash they can out of those people still stubbornly loyal to them after all these years of disappointment, and unfulfilled promises and wasted potential.

      I have been saying this for months now. HexEnt needs to stop putting the cart before the horse. Stop trying to charge money for modes that AREN'T FUN because you are forcing them to be competitive in predatory and manipulative ways. Instead, make fun, casual, FREE modes that are ACTUALLY FUN TO PLAY. Then people will willingly pay you so they can collect all the cards and accessories. Clean up the interface, new user experience, and onboarding process while you're at it, and the game can actually start growing again instead of festering and dying out.

      You're doing it backwards, you refuse to see or admit that, and you're killing your own product in the process.
      --ossuary

      "Love all, trust a few, do wrong to none."
      - Shakespeare, All's Well That Ends Well
    • Ossuary wrote:


      I have been saying this for months now. HexEnt needs to stop putting the cart before the horse. Stop trying to charge money for modes that AREN'T FUN because you are forcing them to be competitive in predatory and manipulative ways. Instead, make fun, casual, FREE modes that are ACTUALLY FUN TO PLAY. Then people will willingly pay you so they can collect all the cards and accessories. Clean up the interface, new user experience, and onboarding process while you're at it, and the game can actually start growing again instead of festering and dying out.

      You're doing it backwards, you refuse to see or admit that, and you're killing your own product in the process.
      These modes are a way to get income in between set releases without having to expend effort that is anywhere close to an actual set release. For that purpose, it is clearly working, otherwise they would have stopped after the first melee. In a way this is very much like the idea they floated for QST, people will pay for 'mystery' games the developers have full control to make in any way they want. Gotta give them credit for advancing creative direction while not forgetting to collect payment, exactly what a good business should be doing.
      "Winning with terrormill is not fun? Try losing with it"
    • We have been in this spot and holding for a very long time. The AI fully understands the rules of Hex, but has no idea what the objective is, or how anything actually works.

      It can't read what the board will do, only what the board is. "This troop gains flight on attack and will be unblocked? Slow down there, we don't want to be productive!"
      It can't identify which shards are required. "All Ruby cards and all Wild Shards? PERFECT, that's a keeper!"
      It just randoms on Fateweave and Verdict. "Oh? You've got 6 shards and no play cards in hand? Lemme just get you another shard."

      Most agonizing is I specifically built a clever deck to smother all the degenerate pump up stuff people are playing. Problem is, the AI won't play the damn Filk Ape. Just saving it I guess, in case there's a princess that needs to be carried to the top of a construction site. Equally as horrifying. the attacker has a troop with like, Momentum 6, and a troop with no effects at all. AI plays Gorilla Pounder (with equipment) and reverts the troop with no effects. WTF. Really? Well played! Why not just print cards that say 'give your opponenet 400 gold' and save us all some time.

      I am a PvE player. I spend heaps of money on cards because I enjoy variety and experimentation. I just want the stuff from the packs then I'll abandon this mode like Chernobyl. It's unfortunate, but aspects of this game feel like they are much more frustrating of late, than fun.

      The post was edited 1 time, last by ShinGuard ().

    • KozHex wrote:

      Even if the AI doesn't understand that LV hits the face, Edswor still has a full board of troops that would die to this.
      For the sake of figuring out what the reason is for the AI not playing it in your situation; AI does know that LV hits the face.



      The one difference I can see is that LV by itself wasn't lethal in your case and it was in mine. I'm assuming that may be relevant because it played LV in my situation even though it couldn't kill any troops with it.
      "Ignorant beliefs are stains upon the mind."
    • Oh, I've seen it play LV before too, both in my decks and against me. The problem was that the AI didn't understand that playing two of the three one cost LV's in its hand was lethal.

      What's really stupid is that how many times in the FRA have we seen the AI Heat Wave an empty board? Or play any number of cards that have zero effect on the board state? Yet it sat on three 1-cost LV's that it had the resources to play? Makes no sense
    • So what we can conclude from this is that the AI does not realize it has lethal if that lethal is achieved by playing more than 1 card from the hand. Additionally it doesn't play the first LV because it doesn't think killing your troops and doing non lethal damage is worth it. (Regardless of what is in it's hand. I'm sure if it did play the first for damage, it would then play the second for lethal.)

      I can understand it not realizing lethal with multiple cards played from the hand. That's just something that seems to be missing from the code.

      I do not understand why it doesn't use the first LV to take care of the troops.

      It might both be improved upon, hopefully.
      "Ignorant beliefs are stains upon the mind."
    • Transience wrote:

      So what we can conclude from this is that the AI does not realize it has lethal if that lethal is achieved by playing more than 1 card from the hand. Additionally it doesn't play the first LV because it doesn't think killing your troops and doing non lethal damage is worth it. (Regardless of what is in it's hand. I'm sure if it did play the first for damage, it would then play the second for lethal.)

      I can understand it not realizing lethal with multiple cards played from the hand. That's just something that seems to be missing from the code.

      I do not understand why it doesn't use the first LV to take care of the troops.

      It might both be improved upon, hopefully.
      If i remember correctly there was some AI code in the developer stream with Chris that basically created a "dictionary" (was this the word?) of cards. If it doesn't update this dictionary whenever a card in your hand changes cost / triggers assault it will never see the LV as playable.
      Beware of he who would deny you access to information, for in his heart he dreams himself your master.
      • Commisioner Pravin Lal - Alpha Centauri
    • SlayerStronghold wrote:

      Transience wrote:

      So what we can conclude from this is that the AI does not realize it has lethal if that lethal is achieved by playing more than 1 card from the hand. Additionally it doesn't play the first LV because it doesn't think killing your troops and doing non lethal damage is worth it. (Regardless of what is in it's hand. I'm sure if it did play the first for damage, it would then play the second for lethal.)

      I can understand it not realizing lethal with multiple cards played from the hand. That's just something that seems to be missing from the code.

      I do not understand why it doesn't use the first LV to take care of the troops.

      It might both be improved upon, hopefully.
      If i remember correctly there was some AI code in the developer stream with Chris that basically created a "dictionary" (was this the word?) of cards. If it doesn't update this dictionary whenever a card in your hand changes cost / triggers assault it will never see the LV as playable.
      Simply not true i have seen it cast LV multiple times at bellow max cost.
    • Wolzarg wrote:

      SlayerStronghold wrote:

      Transience wrote:

      So what we can conclude from this is that the AI does not realize it has lethal if that lethal is achieved by playing more than 1 card from the hand. Additionally it doesn't play the first LV because it doesn't think killing your troops and doing non lethal damage is worth it. (Regardless of what is in it's hand. I'm sure if it did play the first for damage, it would then play the second for lethal.)

      I can understand it not realizing lethal with multiple cards played from the hand. That's just something that seems to be missing from the code.

      I do not understand why it doesn't use the first LV to take care of the troops.

      It might both be improved upon, hopefully.
      If i remember correctly there was some AI code in the developer stream with Chris that basically created a "dictionary" (was this the word?) of cards. If it doesn't update this dictionary whenever a card in your hand changes cost / triggers assault it will never see the LV as playable.
      Simply not true i have seen it cast LV multiple times at bellow max cost.
      Well, have you considered that some event caused the dictionary to update?
      Beware of he who would deny you access to information, for in his heart he dreams himself your master.
      • Commisioner Pravin Lal - Alpha Centauri
    • Transience wrote:

      So what we can conclude from this is that the AI does not realize it has lethal if that lethal is achieved by playing more than 1 card from the hand. Additionally it doesn't play the first LV because it doesn't think killing your troops and doing non lethal damage is worth it. (Regardless of what is in it's hand. I'm sure if it did play the first for damage, it would then play the second for lethal.)

      I can understand it not realizing lethal with multiple cards played from the hand. That's just something that seems to be missing from the code.

      I do not understand why it doesn't use the first LV to take care of the troops.

      It might both be improved upon, hopefully.
      Quoting myself here but I just saw AI use LV to take care of the opp's troops and then attacking for lethal afterwards. Like it should have done in KozHex's situation.


      So, i'm not sure what the reasons are for sometimes playing correctly and sometimes not playing correctly.
      "Ignorant beliefs are stains upon the mind."
    • Yeah, the AI's behavior with this card is very inconsistent. Recently I had it use it correctly to win the game, but last night it once again refused to play it.

      The situation was this: the opponent had one troop left on the board, an Annihilix, and 9 health. The LV had ticked down to 5 and the AI had five (unused) resources and nothing else in its hand except a shard. It ended its turn without playing the LV, which of course ended up costing it the game as the Annihilix forced it to discard it the following turn and just snowballed from there

      Super annoying
    • Seriously, you guys. Are you ever going to address this shit?

      AI: has 7 resources, and burn to the ground in hand. Opponent: has 4 health. AI: plays BttG on a random troop; dies to attack next turn.

      Fucking pathetic. My 13 year old programs better than this.
      --ossuary

      "Love all, trust a few, do wrong to none."
      - Shakespeare, All's Well That Ends Well
    • Perhaps there's code in the AI to find what it thinks the top 2-3 plays would be, and then executes #1 80% of the time, #2 15% of the time, and #3 5% of the time.

      That would result in the AI not being entirely predictable, have the ability to 'bluff', and sometimes just make misplays. In the past they've stated that they don't want the AI to just stomp players all day, so having imperfections in play could be intentional. Not to say they shouldn't take that out for siege, where the AI is a substitute for the player, but it's a possibility that explains how you can see the AI make the right play some times, and the wrong play at other times.

      Edit: And if it is something along these lines, I would suggest removing 'no play' as a possible line of play. It'd explain how the AI passes instead of playing cards on early turns, despite clearly having plays to make.
      Old username: Aradon | Collector backer | Starting a guild for Newbies -- "The Cerulean Acadamy" -- Taking applications once guilds are implemented
    • Ossuary wrote:

      Seriously, you guys. Are you ever going to address this shit?

      AI: has 7 resources, and burn to the ground in hand. Opponent: has 4 health. AI: plays BttG on a random troop; dies to attack next turn.

      Fucking pathetic. My 13 year old programs better than this.
      Also, still, AI considers Annihilix unblockable, both on attack and on defence.
      Played FRA, AI doesn't block him with stronger PRISMATIC troops.
      Watched a Siege replay from a deck I was testing.. AI has in hand Morphology and Time Ripple and Annihilix on board.Opponent has Lilygrove Warden and a Candlekin.
      I am like "cool, will remove the two blockers, them attack". What AI did? Just attack. And then die of course. Along with my dreams. :P
      The Wardens of Entrath is a large, open to public community of Hex enthusiasts with a mission to help and support Hex players and content creators alike!

      Find us on our Discord server or website!
    • Obsidian wrote:

      Perhaps there's code in the AI to find what it thinks the top 2-3 plays would be, and then executes #1 80% of the time, #2 15% of the time, and #3 5% of the time.

      That would result in the AI not being entirely predictable, have the ability to 'bluff', and sometimes just make misplays. In the past they've stated that they don't want the AI to just stomp players all day, so having imperfections in play could be intentional. Not to say they shouldn't take that out for siege, where the AI is a substitute for the player, but it's a possibility that explains how you can see the AI make the right play some times, and the wrong play at other times.

      Edit: And if it is something along these lines, I would suggest removing 'no play' as a possible line of play. It'd explain how the AI passes instead of playing cards on early turns, despite clearly having plays to make.
      Yeah that is a lie they told back in AZ days to explain the poor AI after some people who had cards that revealed opposing hand highlighted some strange plays. If they really had a super amazing AI they would have put it into siege instead of trying to fix the current one.

      Edit: @Chris use to brag about a super amazing AI that was too good for AZ/FRA, so they didnt use it. Seems like siege is a logical place for it since money is on the line. Cory keeps saying they have limited resources, so why keep spending them on the "bad AI" if they have a better one they can just use for siege? Simple answer all that bragging was a lie.

      The post was edited 1 time, last by NeroJinous ().

    • Why won't Hex interact with the community? I know it's been a long time that people have complained about communication. But they must see the state of the community. Like, how does an internal conversation about communicating go? Do they just hope it will go away if they say nothing?