PDA

View Full Version : PVE after AI fixes



Sparrow
05-12-2015, 12:03 PM
I've been thinking this morning about what PVE will look like after the AI has been cured of its most egregious problems. Will enemy champion powers be nerfed? Will it be necessary to allow a sideboard? Will rewards be increased? Will players be complaining that it's too hard? Will difficulty settings be primarily used to make the experience easier rather than harder?

I think potentially all of the above could happen and even become necessary, especially the addition of sideboards to Arena. Even if ultimately CZE can't find an elegant solution for the deeper problems in the AI, they could instead approach it on a champion-by-champion and card-by-card basis. That alone would make the AI a lot more challenging.

So, what will the future of PVE look like post AI fixes?

exiledtyrant
05-12-2015, 12:16 PM
When AI gets fixed I'm guessing new players are going to be stuck on grinding first tier for a much longer time. The powers are so strong now that without the current misplays it would be hard to progress. Ruby enchantress and adamant come to mind immediately as problem AI when they get smarter.

nicosharp
05-12-2015, 12:19 PM
New Equipment and New PvP sets will be what offsets a good arena deck from a perfect AI arena deck going forward. This is why the tiers are also scaled with Equipment, not just deck synergy/passive/charge powers.

Khazrakh
05-12-2015, 12:26 PM
Also Arena isn't meant to be an introductory level PvE experience.
By the time the AI will really know what to do we'll have plenty of easier encounters for new players to farm and gear up.

Lafoote
05-12-2015, 01:21 PM
Some arena encounters will be brutal with competent AI. Some will be randomly brutal, and others will still be a free win.

Examples:

Brutal: War Bot. You won't beat him without Robots. He builds charges too fast and can wipe the board about every 3 turns if required.

Randomly Brutal: Ruby Enchantress. Often pulls the Zakiir card that destroys 4 resources. She should win in about 3-4 turns following that play in virtually any game, yet still manages to bungle it most of the time.

Free Win: Construct Foreman. His deck just doesn't produce enough non-trap damage to compete. No AI tweaks will improve him, his deck needs upgrades as well.

Khazrakh
05-12-2015, 01:37 PM
Brutal: War Bot. You won't beat him without Robots. He builds charges too fast and can wipe the board about every 3 turns if required.

Nah. War Bot has no combat tricks, no removal besides his charge power and mostly cheap&weak robots.
He is and will be easily beaten by most control decks, no matter if they run robots or not.
No matter how good his AI will get, a deck facing him has to be able to do two things:

A) Survive his early onslaught.
B) Outlast him later into the match.

Most encounters will profit a lot more from a competent AI than War Bot will.

Lafoote
05-12-2015, 02:36 PM
Nah. War Bot has no combat tricks, no removal besides his charge power and mostly cheap&weak robots.
He is and will be easily beaten by most control decks, no matter if they run robots or not.
No matter how good his AI will get, a deck facing him has to be able to do two things:

A) Survive his early onslaught.
B) Outlast him later into the match.

Most encounters will profit a lot more from a competent AI than War Bot will.

So everyone can only play control decks forever? Primarily many decks that don't match up well are winning anyway by exploiting the AI's inability to recognize single target threats. When the AI plays better, it will destroy a lot more decks that are skating by right now.

Sparrow
05-12-2015, 02:43 PM
Like I mentioned in my OP, I think AI improvements will require the ability to have a sideboard. It would really be a shame if everyone had to play a specific deck archetype to run the Arena -- allowing a sideboard opens up the design space significantly on deck building.

Lafoote
05-12-2015, 02:50 PM
I don't think a sideboard is part of the plan. They want you to run the gauntlet. They just have to ensure the decks aren't too focused. If they are, it may be exceedingly difficult.

At present, it's a little too easy, it will me much harder overall with better AI. Hopefully more PvE cards, a lot more equipment, and those much anticipated mercenaries will balance things out.

hex_colin
05-12-2015, 02:53 PM
You're all forgetting that we'll eventually being able to swap in Mercenaries for encounters a given deck doesn't match up well against. They are the ultimate sideboard!

Xenavire
05-12-2015, 02:58 PM
You're all forgetting that we'll eventually being able to swap in Mercenaries for encounters a given deck doesn't match up well against. They are the ultimate sideboard!

I do hope that doesn't just apply for bosses in the arena, and I also hope we can do it for more than 1 encounter! I like the idea of freely swapping (because really, I sometimes get bored, but don't like to go in without my 'powerhouse' deck. Having a merc would allow me to just build two completely different fun strats that don't leave me vulnerable when I just want to kick around. Uruunaz really brings out the worst of my paranoid reasoning :p)

Sparrow
05-12-2015, 03:08 PM
Well, yeah, swapping in a mercenary would be great. Though, I don't think it's written in stone anywhere that Arena won't eventually have a sideboarding option.

Edit: I guess it depends how often a mercenary switch can happen. Sideboards may not be necessary at all.

tolroe
05-12-2015, 04:06 PM
I'd love to freely swap in Mercs, but maybe there is some middle ground to encourage single deck runs? Perhaps swapping in a Merc takes a percentage of your potential gold winnings for that particular encounter (most mercenaries like to be paid)? Perhaps the percentage starts out small and increases for subsequent swaps in a single run?

Lafoote
05-12-2015, 04:17 PM
I think the design is intended to let you merc 1 or 2 encounters total. There has also been discussion suggesting running a Merc as your champion will be allowed at some point. The intent is geared toward replay ability and fun, especially when grinding encounters over many, many runs. Not sure if that will still allow for swap outs in the same fashion.

Sparrow
05-12-2015, 04:17 PM
I like the idea of having to pay the mercenary. I think I'd prefer it to be just be a flat fee subtracting from a player's gold horde each time a mercenary is used. While there's nothing wrong with the mercenary taking a percentage of a dungeon/arena run's gold, it seems overly complicated.

sukebe
05-12-2015, 04:18 PM
So everyone can only play control decks forever? Primarily many decks that don't match up well are winning anyway by exploiting the AI's inability to recognize single target threats. When the AI plays better, it will destroy a lot more decks that are skating by right now.

actually, it doesn't even need a control deck. if you play an aggro deck right you will still beat the war bot almost every time. My sly huntress deck has yet to lose to warbot even once because I just play around his ability. Even if he gets better at using his charge power I don't see him getting that much more difficult. The only aggro decks that should have trouble against him are decks without troop removal that can hit artifacts.

Yewstance
05-12-2015, 04:33 PM
Did KG lie to me? I was under impression from discussions in-client chat that mercenaries were not intended to be usable in Arenas, or at least the Frost Ring, and they were for Dungeons/etc. I'd be overjoyed to see mercenaries usable in Arena, though, just as I'd be excited to see sideboards/reserves allowed - the latter I feel is particularly important when Spectral Lotuses become usable, lest you pack your deck with a set and then spend the majority of your games treating them as dead draws interfering with your deck because you want to save them for a particularly hard encounter later (such as Xarlox). Sideboarding solves that issue nicely.

Otherwise, in regards to the AI difference in difficulty, I think it's safe to state that almost every AI will skyrocket in difficulty, INCLUDING Construct Foreman. When using cards that enable you to see the opponents hand, it becomes quickly clear how disastrous the current AI handles its threshold generation, the timing of its plays and - perhaps most notably - it's reluctance to use removal, or its bad prioritizing/targeting when it chooses to do so (Heat Wave on an empty board, anyone?). Even Construct Foreman has a nice removal suite and a few potential win conditions, albeit he's still going to be one of the weakest AI (now give him blue threshold and Ushul with the shard=mill equipment like one of my decks, on the other hand...).

Zoltog seems, off the top of my head, one of the champions least reliant on strong AI, but even teaching him what "This creature can't block" means as well as the implication of "Must Attack" would be nice.

I do agree with previous posts, though. More control-oriented champions such as Ruby Enchantress (and possibly Blood Sphinx) are going to be scary with their sudden ability to handle their removal - backed up by their champ powers giving superior card advantage - with some degree of skill and prioritization. The former with its random power cards, the latter with its recurring field wipes.

Maybe King Gabriel will become a challenge once he works out how to judge an opening hand based on colors available, and be able to handle his threshold generation to actually curve out and survive long enough to end up with a Boldheart given spellshield via Mirror Magi or something to that effect.

Xenavire
05-12-2015, 04:37 PM
Did KG lie to me?

There has been a lot of conflicting information (especially when the design of the mercs changed) so we aren't sure of anything. Last word, as far as I am aware, is that mercs are planned to land in arena at some point. Can't say anything with 100% certainty however.

Gwaer
05-12-2015, 05:21 PM
Yes, KG probably lied to you.

Oroniss
05-12-2015, 07:02 PM
I always thought that the sideboard would be a really good account levelling feature. Start off small, like 5 card sideboard useable twice per dungeon/arena say, and work up to some larger size plus 1 or more mercs, useable before each fight.

I think this would dramatically increase the number of useful cards for PvE. At the moment you can't really tailor much to any particular fight so the arena feels less like a 'puzzle' to be solved and more like a case of build a standard deck and hope RNG doesn't match you up with too many opponents your deck struggles against.

Just my 2c.

Patrigan
05-12-2015, 10:13 PM
I want to point out that the AI is never intended to play perfect. IN fact, it will get flaws and some major ones to boot. It's called "personalities" and it'll be part of the way to beat him.

The example they gave was a champion with a hate for dwarves. If he sees a dwarf, he'll throw everything he has at him to destroy him, even if that's a bad play (for example, he could've killed your awesome 20/1, but instead he chose to kill the measly 1/1 dwarf instead). Once these personalities are getting known, they'll clearly also start influencing our deckbuilding.

Sparrow
05-12-2015, 10:37 PM
No one is asking for or anticipating a perfect AI, as far as I can tell. The way the AI behaves now isn't a case of personality -- unless you consider brain damaged a personality trait.

Gwaer
05-12-2015, 11:00 PM
It might be neat to have a very small sideboard that you could access mid game in pve. Kind of like being able to use special potions or something in other games, 1 or 2 cards that you could put into the top X cards of your deck, or maybe pay for them at any time but they have some sort of delay before activating. Needs some kind of downside to be a usable mechanic, pretty much impossible to use anywhere but pve unless you had a very limited list like camps or gems that were balanced to be used that way. Just trying to spitball a middle ground between losing and switching out some cards/changing decks entirely with a merc. Maybe you only get one or two card slots you can bring in at almost any time but they don't refresh between encounters. Can only use them once for an entire dungeon.

Needs thought.

poizonous
05-12-2015, 11:24 PM
Personality is fine, but Horrific plays should never enter the game, personality or not, an AI's personality should not cost him the game because of a hatred of dwarves. I can picture the personality now "Oh no, its a pesky worker bot, better kill that even though the ash harpy is about to kill me next turn" Sorry but thats unacceptable and ruins any challenge the game can present to us

Clawdius
05-13-2015, 01:06 AM
Personality is fine, but for lack of better term "Retarded" plays should never enter the game, personality or not, an AI's personality should not cost him the game because of a hatred of dwarves. I can picture the personality now "Oh no, its a pesky worker bot, better kill that even though the ash harpy is about to kill me next turn" Sorry but thats unacceptable and ruins any challenge the game can present to us
Wait you mean my AI opponents shouldn't routinely play Gore Feast with no creatures on the field? But what about style, panache, machismo? Nah, you're probably right, the AI might ought to stop playing Shamed Gladiator when it has 2 health left.

Khazrakh
05-13-2015, 02:28 AM
Readying Droo's Colossal Walker with 5 health left is fine though isn't it :)?

wolzarg
05-13-2015, 02:30 AM
Personality is fine, but for lack of better term "Retarded" plays should never enter the game, personality or not, an AI's personality should not cost him the game because of a hatred of dwarves. I can picture the personality now "Oh no, its a pesky worker bot, better kill that even though the ash harpy is about to kill me next turn" Sorry but thats unacceptable and ruins any challenge the game can present to us

Poor, suboptimal, terrible, horrific, disastrous, flat out bad, not good, bad.
No need to thank me just pick from the list or Google thesaurus then edit your post.

poizonous
05-13-2015, 02:42 AM
Poor, suboptimal, terrible, horrific, disastrous, flat out bad, not good, bad.
No need to thank me just pick from the list or Google thesaurus then edit your post.

But none of those words stand out as strong as the one I chose, LOL j/k

Mourad13
05-13-2015, 02:55 AM
When the AI gets fixed i think we might see less 20-0(or 15-0) if arena stays the same because some champions just making the right play(pump spell at the right time, attack/defense logical, removal played at the right time, damn storm cloud getting at last some counters(i would have lost so many games if it was the case^^),...) will suffice to make arena challenging.
Maybe resulting less speed run and more aggro/control decks according to me
My 2 plats

wolzarg
05-13-2015, 03:21 AM
Good choice fine sir.

Zophie
05-13-2015, 08:57 AM
But none of those words stand out as strong as the one I chose, LOL j/k

I would have went with Derpy (http://www.urbandictionary.com/define.php?term=Derpy), that's popular with the kids lately.

In response to your concerns with lack of challenge though, remember that these "personalities" would be intended to be a part of the strategy in how you defeat these encounters. A dungeon will have many different encounters, and maybe a couple of them you know are affected by certain types of cards more than others, so you can build your deck around that to help you through them, or maybe just build a special merc deck to swap in when you get to that encounter so you can use whatever normal deck you have on the rest of the encounters. I don't think these personalities will simply trivialize the encounters, but I think of it rather as a way to reward preparedness by giving you ways to optimize your deck to leverage the personalities to your favor.

Sparrow
05-13-2015, 03:45 PM
Rather than starting a new thread, I thought I'd add a thought I had about fixing the AI here.

Much like the brute force a chess AI uses, the AI should try different combinations, play the turn to itself, and then compare the possible boards one turn into the future. Clearly it doesn't do anything resembling that now and this would take care of a large chunk of the problems with it. Depending on difficulty setting or how much time it takes, it could even look more turns into the future.

The result? A huge reduction in bad targeting, no more Heat Waves on an empty board, no more suicide plays, no more rezzing a creature from the graveyard that can't be used, more logical decisions on which units to attack with, etc., etc., etc.

If the AI takes too long to search, game clocks could be applied and the AI would have to prioritize its search order in some way. This would be much easier to program and be much more successful than having the AI try to logically choose what to do, one move at a time.

Zophie
05-13-2015, 04:01 PM
Rather than starting a new thread, I thought I'd add a thought I had about fixing the AI here.

Much like the brute force a chess AI uses, the AI should try different combinations, play the turn to itself, and then compare the possible boards one turn into the future. Clearly it doesn't do anything resembling that now and this would take care of a large chunk of the problems with it. Depending on difficulty setting or how much time it takes, it could even look more turns into the future.

The result? A huge reduction in bad targeting, no more Heat Waves on an empty board, no more suicide plays, no more rezzing a creature from the graveyard that can't be used, more logical decisions on which units to attack with, etc., etc., etc.

If the AI takes too long to search, game clocks could be applied and the AI would have to prioritize its search order in some way. This would be much easier to program and be much more successful than having the AI try to logically choose what to do, one move at a time.

As research in quantum computing gets more advanced these kinds of computations will become a lot easier to process in lots of different AI applications. I really look forward to more breakthroughs to be made in this field and it's really exciting to imagine if some day in the future Hex used quantum computing in their AI. :)

Sparrow
05-13-2015, 04:57 PM
I heartily endorse a marriage of Hex and quantum computers, though I'm not sure of the age of consent for a digital card game.

In the meantime, though, considering the power of brute force chess AI on PC's, I think the hardware already exists to make the Hex AI as powerful as we'd like.

Mejis
05-13-2015, 08:12 PM
Rather than starting a new thread, I thought I'd add a thought I had about fixing the AI here.

Much like the brute force a chess AI uses, the AI should try different combinations, play the turn to itself, and then compare the possible boards one turn into the future. Clearly it doesn't do anything resembling that now and this would take care of a large chunk of the problems with it. Depending on difficulty setting or how much time it takes, it could even look more turns into the future.

The result? A huge reduction in bad targeting, no more Heat Waves on an empty board, no more suicide plays, no more rezzing a creature from the graveyard that can't be used, more logical decisions on which units to attack with, etc., etc., etc.

If the AI takes too long to search, game clocks could be applied and the AI would have to prioritize its search order in some way. This would be much easier to program and be much more successful than having the AI try to logically choose what to do, one move at a time.

This type of AI decision making already exists, no? E.g. the dotp AI. This is from the wiki:
"The game's artificial intelligence (AI) is based on creating a multistep "lookahead" decision tree from the current state of the board, evaluating possible known moves by its opponents, scoring the current and resulting situations both positively and negatively, and then selecting the option with the highest score. The artificial intelligence utilizes threading to run the main intelligence algorithm alongside three "sub-contractors" that evaluate the possible future states and report back to the main algorithm. Each of these instances of the AI uses a special engine that both implements the rules of Magic, while also providing "undo" actions so that they may explore up and down the decision tree. The game's intelligence algorithm runs on a separate CPU core than the main game to avoid creating framerate issues with its display while the player is considering their actions.[4] When the player makes an action, the AI threads are interrupted and brought to the same state as the game, then continue processing, in some cases, reusing existing branches on the decision tree that match with the player's selection.[4] The AI itself does not employ any strategy in terms of strong card combinations, but instead, when such combinations improve the computer's situation, the chances of playing these combinations will "ripple" back through the decision tree. In this manner, the computer AI will play one card of such a combination should it be the best choice at that time, and follow up with the second card of the combination should that option still remain the best after any other player actions."

Sparrow
05-13-2015, 09:40 PM
I've never played Duels, but I'm glad to see this has already been tried with a TCG.

Elwinz
05-14-2015, 12:16 PM
I dont wanna imagine Blood sphinx with proper AI.