So, I've Noticed This Recently ...

chamber44

I know this is just confirmation bias, but humor me for just a minute:

Is it just me, or does it seem like the AI has suddenly gotten much better at the order with which it casts particular powers?

Like suddenly, the AI is using batteries properly (except for GSBW's green) . . . like casting 3Clop's yellow, then red, then black.

I know it's not a conspiracy. I know the AI isn't that clever. But it's almost more frustrating than a never-ending cascade.

Almost.

mbaren

I've noticed that too. Agreed, might just be random and/or bias on my part, but I've also gotten the sense that the AI is making slightly smarter decisions about when to trip certain powers.

MarvelMan

Its not playing better. Your expectations for its quality of play have just finally sunk below its actual "skill" level.

Ebolamonkey84

As far as I know, the AI will cast powers in the order the AP is listed: Yellow --> Red --> Blue --> Purple --> Green --> Black

Your cyclops example actually follows this pattern. I think it still works this way, as I have still never seen an AI Hulkbuster do an Overdrive into Repulsor Punch on the same turn.

Vhailorx

It's not writing better code that prevents the devs from altering the ai. I think someone from demiurge has already written a better script for the ai (as per a dev blog last summer) so much of the work is already done.

The problem is that rebalancing the game to our satisfaction is more trouble than it's worth. The game currently used over-leveled opponents to ramp up challenge, so all of the scaling mechanisms would have to be changed. And even then, we players would HATE a smart ai in the current system, since our winning rate would drop sharply and we already freak out when we don't win 80+% of the time. PvP especially rewards winning very reliably (every loss in high level PvP play is devastating), so drastic scaling changes are sure to puss everyone off.

JVReal

Ebolamonkey84 wrote:

As far as I know, the AI will cast powers in the order the AP is listed: Yellow --> Red --> Blue --> Purple --> Green --> Black

Your cyclops example actually follows this pattern. I think it still works this way, as I have still never seen an AI Hulkbuster do an Overdrive into Repulsor Punch on the same turn.

I'm pretty sure I have been blasted by Overdrive into Repulsor punch when the AI does not have enough AP for Repulsor Punch, until AFTER Overdrive generates red AP and enough to trigger Repulsor Punch.

Vhailorx

I am sure lots of players think they want to play against a smarter ai (including me). I wonder how many of us would still feel that way after trying it. . .

cyineedsn

Vhailorx wrote:

I am sure lots of players think they want to play against a smarter ai (including me). I wonder how many of us would still feel that way after trying it. . .

Do I ever ask myself, I wonder if this game would be more fun if I had to spend even more time in PVE clears due to wiping against a smarter AI, and even more engaging if I spent more time stopping to shield in PVP after some bad wipes? No, not really.

GrumpySmurf1002

JVReal wrote:

Ebolamonkey84 wrote:

As far as I know, the AI will cast powers in the order the AP is listed: Yellow --> Red --> Blue --> Purple --> Green --> Black

Your cyclops example actually follows this pattern. I think it still works this way, as I have still never seen an AI Hulkbuster do an Overdrive into Repulsor Punch on the same turn.

I'm pretty sure I have been blasted by Overdrive into Repulsor punch when the AI does not have enough AP for Repulsor Punch, until AFTER Overdrive generates red AP and enough to trigger Repulsor Punch.

I'm fairly sure when people see this it's because of a lost turn to stun or some similar action, where the AI is actually just starting the cycle again at the beginning of their next turn.

GrumpySmurf1002

OJSP wrote:

The AI always process each colour in the sequence of (just like the AP list on the match screen) .. So as you can see, some characters benefit from having their accelerating powers being processed before their other powers:

Cyclops: Mutant Revolutionary, Optic Blasts, Full Blasts
Mystique: Infiltration, Shapeshift, Masterstroke
Ares: Sunder, Rampage, Onslaught
Thor: Thunder Strike, Call the Storm

All the above will be processed in the same turn if the AI have enough AP.

That's why the AI won't ever be able to do specific combo in the same turn like Hulkbuster's Overdrive to Repulsor Punch, GSBW's Deceptive Tactics to Sniper Rifle, or KK's Inhuman Stretchiness to Embigenned Bash.

This is also why it's so annoying playing against the Dark Avengers, especially with Venom who can stun the whole team with Symbiote Snare. If paired with Juggernaut and Daken, that extra turn they have to make another match can practically restart the whole cycle of Chemical Reaction, Unstoppable Crash, Symbiote Snare, Headbutt, Chemical Reaction, Unstoppable Crash, and so on and so forth.

Edit: beaten by someone else by one minute.. why do I type long posts..?

This is 100% correct. Also makes something like Fury's yellow have more utility in the AI's hands, as he won't drain the other colors first. Ditto 4Cyc's red (if 10 TUAP). Also why you don't want to give the AI an active yellow (that costs <12) with OML if you can avoid it.

JVReal

GrumpySmurf1002 wrote:

JVReal wrote:

Ebolamonkey84 wrote:

As far as I know, the AI will cast powers in the order the AP is listed: Yellow --> Red --> Blue --> Purple --> Green --> Black

Your cyclops example actually follows this pattern. I think it still works this way, as I have still never seen an AI Hulkbuster do an Overdrive into Repulsor Punch on the same turn.

I'm pretty sure I have been blasted by Overdrive into Repulsor punch when the AI does not have enough AP for Repulsor Punch, until AFTER Overdrive generates red AP and enough to trigger Repulsor Punch.

I'm fairly sure when people see this it's because of a lost turn to stun or some similar action, where the AI is actually just starting the cycle again at the beginning of their next turn.

I would normally believe that, if we did not repeatedly get pummeled by Juggs Green, giving him AP to do his Headbutt... which fall outside of AP order on the screen... right?

I'll watch closer next time.

GrumpySmurf1002

JVReal wrote:

I would normally believe that, if we did not repeatedly get pummeled by Juggs Green, giving him AP to do his Headbutt... which fall outside of AP order on the screen... right?

I'll watch closer next time.

Assuredly it's either Headbutt->Crash -> (Next Turn) Headbutt or Crash -> Venom Snare -> (Next Turn) Headbutt.

The AI is pretty dumb, it doesn't even start the cycle over after a match-5. The trigger to the power cycle (presumably) is only the end of your turn.

IamTheDanger

I remember that the devs once posted that they intentionally leave the A.I. intelligence at a low level. And personaly, I hope they leave it like that. Like with Business Partners node in TaT, (one of my 3 least favorite nodes in the entire game), just imagine if a level 326 C. Mag waited until there were a Lot of team up tiles on the board before casting Polarizing Force. Playing this node the other day, Mags cast PF when there were only 4 tu tiles on the board.

I was using Steve Rogers. I had both a red and blue shield count down on the board, with a few protect tiles out. Hood used his Twin Pistols and missed all of them except one protect tile. Personally, I'd want it to stay that way. I don't want the A.I. targeting my cd/attack/protect/strike tiles.

JJ

Calnexin

aesthetocyst wrote:

As long as they knew what they were getting into, I believe players would love squaring off against smarter AI rather than silly-scaled opponents....

I believe the opposite. A dumb AI is exploitable. Challenge from overleveling can be accommodated with your roster selection and matching strategy - you know going in what colors will kill you, so you deny them while building your own pool as best you can. If the AI suddenly gets smart, player advantage is gone. You have to play as though you were facing a person, which is a much tougher nut to crack. If they ever went that way, I'd hope they'd seriously ramp up the rewards per match and decrease the progression tiers. Each match would last longer and take a much bigger toll on you team's health.

It would make the game more dynamic, but given the outcry over the DDPQ randomization, it's really not what players want right now. Personally, my roster development is slow enough. Slogging through longer matches and being unable to play because I've burned through all my healthpacks sounds atrocious.

The Bob The

It also depends on the costs of those powers that work best in sequence. I found myself thinking exactly like you last week when facing multiple boosted Beasts and Cyclopses who seemed to have this sequencing thing figured OUT. Thing is, their accelerating powers (blue Hank, yellow Scott) are cheaper than the powers that benefit (green Hank (or yellow), red Scott), so it's more likely they'll have the AP to fire the first before the second. Not always, mind - they're still likely to fire whatever they've got the AP for - but the annoying nature of the matches made a lot more sense after that.

jackstar0

Just watched the AI cast Wings of Liberty and then Overdrive. The AI is not smarter.

Vhailorx

You are both right and wrong.

The challenge of playing a smart opponent would be interesting. As would the feeling of accomplishment when you won.

But this is a game that imposes several very harsh penalties for failure. In PVE, a loss means losting 30% of your healthpack reserve, or losing the use of 3 characters from your roster. That's brutal when the scoring/reward mechanism in PVE require players to beat the same nodes several times every day. In PVP, the price of failure is even higher at the high end of play. You lose healthpacks and can't just sub out the featured character, you lose points, and you also waste a precious shield hop, permanently crippling to max available score.

So while I like a smart opponent in the abstract, I think I would hate the idea in practice given the games other mechanics. And since I don't think that demiurge is about to remove all of the various play-gating aspects that make them money, I don't want to see a more challenging AI any time soon. . .

The Bob The

aesthetocyst wrote:

Way back when I developed a slew of simple games, and the best part of doing so was developing and refining the AI. Not claiming any amazing expertise here, I'm talking about simple 8bit graphical actioners and text-based strategy games. The game's brain is like a baby, to nurture and play with, and, eventually, teach to trash talk (All my games used their situational analysis engine to generate wisecrackin' "conversation").

It kills me to see such missed opportunity in MPQ!

I dunno, I think the game does do a little trash talk. For example, every time I fumble a match a few time, it gives me the "perhaps you've forgotten how to play a match-3" message. Classic!

Bulls

Vhailorx wrote:

The challenge of playing a smart opponent would be interesting. As would the feeling of accomplishment when you won.

It wouldn't. As in case of really "smart" opponent - ie properly written code to make best of each scenerio player would be ALWAYS in disadvantage - you're likely to miss that 4 or 5 match or didn't see that 5 that will be set just after your move - well scripted ai wouldn't have that problem. Only way to win is either by overwhelmingly better characters put into fight or better luck with cascades. Just think about one small change would make a big of a difference now - AI able to do match 5 in L shape and then thank devs it is at it is