What if higher MMR had effects on the level of AI difficulty

Meto5000

Recently I've been thinking about how this game could move to a system that gave benefits to people with higher MMRs. In PvP, one way would be to improve the skill of the AI when it comes to playing your team. Yesterday I faced so many OBW + mid level Loki teams where the AI consistently used Illusions instead of waiting for 2 more purple AP to cast Aggressive Recon. All these people would have been better served using the loaner Loki with only the black skill. Imagine if the AI knew how to play a MN Magneto + C.Storm team? Right now, those teams are fat juicy targets with almost no shot of winning outside of miracle cascades.

On the other hand, what if PvE enemy intelligence scaled up instead of just level. Would a smart Spidey be better or worse than 3x 395 goons?

Scaling AI probably represents an unrealistic amount of work for the devs but I'm wondering what other people think about this.

Unknown

What if your pvp mmr raised the levels of your opponents team, so you could fight 395 juggernaut OBW spidey in pvp too.

I'd rather play any other game than this one at that point.

Unknown

High MMR should be a *benefit* rather than a *penalty* it is now and sound being even more so with your proposal.

But your idea could be salvaged: having higher MMR I could gain AI boosts for my defensive team. Say ability to switch off character powers or tune their preference, tune color pick preference, adjust effective level, adjust covers, set fixed defender team instead of using the offensive one....

Unknown

While you can argue the merits of your idea it's kind of irrelevant. As long as the game is designed to run on phones it will have **** AI. People wouldn't want to wait the length of time between turns it would take for a phone with anything other than fairly rudimentary AI to run.

Incidentally I would prefer better PvE ai to lvl 395 goons. It would be a much better way to do things but won't happen.

Unknown

bonfire01 wrote:

While you can argue the merits of your idea it's kind of irrelevant. As long as the game is designed to run on phones it will have **** AI. People wouldn't want to wait the length of time between turns it would take for a phone with anything other than fairly rudimentary AI to run.

A plenty of AI improvements would take no measurable time difference and not so hard to implement either. The AI remains **** for design considerations mostly.

Unknown

The hardest part of making an AI is to make it dumb, not smart. It's generally very trivial for the AI to totally slap any human player silly without the need to cheating (and they can cheat if they wanted to too). It also won't be very fun to play. I continue to see people make the argument like "The AI should totally defend the points I plundered from an inferior AI", without realizing that if the AI was actually strong, you would never have all those point in the first place because you'd be fighting against a much stronger opponent along the way.

Unknown

We've had this discussion before. There are some changes they could make that would be overkill (imagine if it scanned to match 5 with 3* Magneto's blue every chance it popped up), but the biggest QOL ones would be fairly simple things that would turn some pretty trash 3* heroes (purely because they're terrible at defense) into actual, decent characters. Imagine if IM40 only used Recharge when no yellow matches were available? What if GSBW used her purple to actually generate big fat green cascades? What if Punisher used his green before his black every time? Really obvious, pants-on-head things for players, that the AI is too lazily coded (yes, lazily, these aren't hard from any point of view - well okay GSBW's might take more than three lines of code) to put up any threat on defense with many characters. I once had a Captain America hit me 4 times with his shield, and every time he put it near the bottom of the screen. He killed everyone but my last character and I barely squeaked out a win. Can you imagine if that wasn't just pure bad luck and he could actually do that reliably? I'd just like to see the AI not do stupid things, like IM40 casting recharge, 2 of the 3 tiles are in the same yellow match... and then making that yellow match.

Unknown

bonfire01 wrote:

While you can argue the merits of your idea it's kind of irrelevant. As long as the game is designed to run on phones it will have **** AI. People wouldn't want to wait the length of time between turns it would take for a phone with anything other than fairly rudimentary AI to run.

A lot of the people discussing ways where you can select move priority on defense got me thinking about it and it would probably be less computationally expensive than the current way

Right now we know there's some kind of random factor in determining if an ability is used, and what order abilities are used in.

If a player could open a menu when going into the battle that lists all 9 character abilities and for each you can select
"Use this ability as soon as possible"
"Never use this ability"
"Use this ability only after X other ability has been used at least once"
"Use this ability only immediately after X other ability"

it would make building defensive teams a lot more interesting and wouldn't slow the game down

Unknown

Currently the AI only goes up to single-digit 1*

Currently the AI doesn't know how to make matches with the color-changing abilities nor does it know how to make 3 by 3 matches.

The AI doesn't know that a lvl 15 guy with the same color as a lvl 141 would still uses the **** lvl 15 guys abilities

Would be a great change if the AI had a clue. And on that note, would be nice if we could set skill prioritization for our defensive teams.

(IE Rank order of skill usage for each color)

(I'm happy the spiderman AI sucks or we'd all quit the game)

Unknown

bonfire01 wrote:

While you can argue the merits of your idea it's kind of irrelevant. As long as the game is designed to run on phones it will have **** AI. People wouldn't want to wait the length of time between turns it would take for a phone with anything other than fairly rudimentary AI to run.

Incidentally I would prefer better PvE ai to lvl 395 goons. It would be a much better way to do things but won't happen.

Although I kind of agree with the patience angle, I must say I have way more patience using my phone than I do sitting at a computer. My phone use is usually somewhere much more comfortable, frequently outdoors, in a park or even home on the couch. Still envious of the computer layout as not having to swipe to see enemy color pools is so needed on the phone/ipad

Unknown

The AI purposely does not evaluate the strength of its AP consuming moves which is why you can have Devil Dino as an opponent and still beatable without infinite combo, because it doesn't prioritize on his green that pretty much is guaranteed to kill anyone in two matches.

It might be a good idea if the AI has preference on certain moves, like Star Spangled prefers an enemy special tile that isn't near red, and Deceptive Tactics prefers to use the tiles near green. Of course, the net effect is still basically 0 because if this indeed makes someone like GSBW better then you'd also expect more resistance from the AI when you fight GSBW.

Meto5000

gamar wrote:

Right now we know there's some kind of random factor in determining if an ability is used, and what order abilities are used in.

If a player could open a menu when going into the battle that lists all 9 character abilities and for each you can select
"Use this ability as soon as possible"
"Never use this ability"
"Use this ability only after X other ability has been used at least once"
"Use this ability only immediately after X other ability"

it would make building defensive teams a lot more interesting and wouldn't slow the game down

I really love this idea and think it could be translated to the mobile and steam UI platform. I think at it's core just a simple slider that assigns priority would be great. You assign a 1-5 number to your skill and if anyone on your team has an ability that shares a color it checks the priority for which it should use. I think the ability to say, only use Molotov Cocktail after Judgement, while nice, would be expecting too much.

Meto5000

Phantron wrote:

The AI purposely does not evaluate the strength of its AP consuming moves which is why you can have Devil Dino as an opponent and still beatable without infinite combo, because it doesn't prioritize on his green that pretty much is guaranteed to kill anyone in two matches.

It might be a good idea if the AI has preference on certain moves, like Star Spangled prefers an enemy special tile that isn't near red, and Deceptive Tactics prefers to use the tiles near green. Of course, the net effect is still basically 0 because if this indeed makes someone like GSBW better then you'd also expect more resistance from the AI when you fight GSBW.

Making the AI do things in a certain way, eg. creating matches with GSBW green/Mag's Blue vs just making certain skills have respective priority comparisons represent two different challenges development wise. The first approach has a lot of tie in to board state and would likely require a lot of new code. The second approach will make it so you can have Thor and Cap'n on the same team, but make Cap's red always have priority over Thor's. The game already does some logic to determine which skill to use, so adding in the priority check would hopefully be somewhat simple to do.

Unknown

Meto5000 wrote:

Making the AI do things in a certain way, eg. creating matches with GSBW green/Mag's Blue vs just making certain skills have respective priority comparisons represent two different challenges development wise. The first approach has a lot of tie in to board state and would likely require a lot of new code. The second approach will make it so you can have Thor and Cap'n on the same team, but make Cap's red always have priority over Thor's. The game already does some logic to determine which skill to use, so adding in the priority check would hopefully be somewhat simple to do.

All things related to AI are actually pretty trivial. For example here's a very simple description to use Star Spangled Avenger more effectively:

Target enemy special tiles that are not directly next to a red tile.
If this isn't possible, pick a tile at random that's at least 1 tile away from a red tile
If this still isn't possible, pick a tile at random.

Since the AI obviously recognizes any match 4s on the screen, it must have an algorithm that knows what tiles forms match 4 so you'd just apply the algorithm that detects match 4 for placing new tiles on abilities like Magnetic Field/Deceptive Tactics.

The question is do you really want the AI to always despoit a Star Spangled Avenger tile in somewhere you can't reach or always form awesome matches with Deceptive Tactics?

For ability priorities, it'd create problem like making Devil Dino way harder in PvE since Prehistoric Bite should have priority over virtually everything. A lot of the PvE battles only work because the AI doesn't behave in a logical way when there are two abilities on the same color.

But even if we ignore that, having behaviors like 'wait for X to happen before using Y' is problematic with the computer, given it does not get to pick who is taking the damage so it can end up losing someone before the condition is fulfilled. If the behavior is probabilistic as opposed to strictly a priority queue, the current behavior for multiple abilities on the same color creates enough randomness already.

Unknown

Phantron wrote:

But even if we ignore that, having behaviors like 'wait for X to happen before using Y' is problematic with the computer, given it does not get to pick who is taking the damage so it can end up losing someone before the condition is fulfilled. If the behavior is probabilistic as opposed to strictly a priority queue, the current behavior for multiple abilities on the same color creates enough randomness already.

I don't get how that would be hard? If the X guy is dead then you obviously just use the Y ability regardless.

Unknown

sms4002 wrote:

Phantron wrote:

But even if we ignore that, having behaviors like 'wait for X to happen before using Y' is problematic with the computer, given it does not get to pick who is taking the damage so it can end up losing someone before the condition is fulfilled. If the behavior is probabilistic as opposed to strictly a priority queue, the current behavior for multiple abilities on the same color creates enough randomness already.

I don't get how that would be hard? If the X guy is dead then you obviously just use the Y ability regardless.

But if you're forced to use an ability when X is close to death then you could've easily ended up using it in a worse time than using it as soon as possible, and since you can't calculate something like this in general, you might as well just use it as soon as possible.

The player has the luxury since you usually have control on who is taking the damage so you usually have enough time for the optimal conditions. The computer cannot do this at all unless they gain the ability to prevent you from targetting a certain guy globally.

Unknown

"They could greatly improve the game by doing this"
"That won't work because it wouldn't make everything perfect all the time"

-every conversation with Phantron

Unknown

gamar wrote:

"They could greatly improve the game by doing this"
"That won't work because it wouldn't make everything perfect all the time"

-every conversation with Phantron

In his defense, he doesn't really say it won't work, he just points out all the side effects that people tend to miss. And he's really good at it. Phantron, you don't happen to be a developer, right?

Unknown

Moghwyn wrote:

gamar wrote:

"They could greatly improve the game by doing this"
"That won't work because it wouldn't make everything perfect all the time"

-every conversation with Phantron

In his defense, he doesn't really say it won't work, he just points out all the side effects that people tend to miss. And he's really good at it. Phantron, you don't happen to be a developer, right?

Actually the way it is put is pretty misleading as half of the picture is simply ignored. Yes, if the AI was better you could not beat the lvl240 dino with your 85 company or the 141 one. But then so what? The dino company need not be lvl 240, it can be 200 or 140 or 80 or whatever that *can* be beaten with good play.

Also the AI abilities are flexible and can be turned on/off on a whim (or better on good design so it could play at different difficulty level in PVE and PVP. Aslo could play differently in easy and hard version of the PVE brackets -- like now we have boring repetition in the pair with just little level difference, instead it would be a completely different play.

As I mentioned before, AI play level could be used to compensate for high MMR the player faces, his defensive team could play with better AI if he was beating harder teams.

The motivation to keep with a dumb AI is probably the aim for stupid public -- but that is questionable for a strategy game. Sure if the AI was put to play at genius level many people would not get hooked even for a week to buy some slots -- but we have no good argument it would also happen with slight improvements. And especially if there were different level arenas. As AI could increase in difficulty as prologue is played, gradually.


TL;DR: there are some fair arguments for dumb AI but they are not killer and not convincing that alternatives would be actually worse in a coherent changeset

Unknown

Moghwyn wrote:

gamar wrote:

"They could greatly improve the game by doing this"
"That won't work because it wouldn't make everything perfect all the time"

-every conversation with Phantron

In his defense, he doesn't really say it won't work, he just points out all the side effects that people tend to miss. And he's really good at it. Phantron, you don't happen to be a developer, right?

No, but I probably think like one.

There's a rather delicate balance between my view which is like 'why bother if people are going to break it anyway' and 'don't worry nobody's going to try to break this feature!' I do understand their plight because if you do nothing people get mad at you, but if you hope people won't be breaking your feature, well... people always do that too.

Unknown

pasa_ wrote:

Actually the way it is put is pretty misleading as half of the picture is simply ignored. Yes, if the AI was better you could not beat the lvl240 dino with your 85 company or the 141 one. But then so what? The dino company need not be lvl 240, it can be 200 or 140 or 80 or whatever that *can* be beaten with good play.

Devil Dino's abilities are designed with the dumb AI in mind. Prehistoric Bite is just flat out overpowering because it's expected he won't use his green all the time due to the AI's flaw. Even with scaling it's hard to see his level drop to something you can manage. The AI's weakness is built into his inherent strength. So if you made the AI smarter, you'd have to rebalance Prehistoric Bite or you might as well just skip all nodes with Devil Dino if stunlock doesn't look like a viable option. Yes this is a relatively minor part of the game but there's a lot of PvE stuff where you can see the weakness of the AI is factored in so there would need to be a considerable rebalancing of encounters.

An unrelated but similar example would be Venom + 2 purple AP generator in Heroic Venom. Why would you design an encounter where the computer takes 3 turns for every turn you get after turn 3? How can that possibly be fair? Probably because it's a stock template for 'anti Spiderman/Magneto' encounter, despite the fact that they cannot be used in that event. The game's various balance issues heavily influences all the PvE encounters themselves, so even fairly minor changes can have a rather large impact in the overall balance and thus is not as easy as it looks.