Does the Hood have X-Ray vision?

Unknown

In the PvE event there is a Nick Fury essential node versus The Hood.

After I cast Nick Fury's Demolition to drop two trap tiles (I was well aware that Hood had enough Yellow for Twin Pistols), one fell into a match position which I matched for the bonus damage, and the other landed into the top row.

The Hood's Twin Pistols promptly targets the remaining trap as the EXACT CENTRE. As in, the AI thought it wise to only destroy six tiles off the Pistols instead of 9.

Do you take me for a fool? This can't possibly be random.

Unknown

With 2 shots at it, it will happen 3.125% of the time (picking the center). Congrats, you got lucky. More generally, with 6 choices to destroy it and 2 shots it will be destroyed 18% of the time. Not so outlandish a chance.

Edit: I've always thought the AI should pick targets for 3x3 abilities by outlawing the outer row. Even random ones like Judgement. Power says it destroys 9 blocks, let it destroy 9 blocks, and not 4 by picking a corner. Slight defensive boost to Punisher and Hood and MNM. Maybe a couple others.

CrookedKnight

Sometimes the AI matches its own countdown the same turn that it places the thing, sometimes it snipes the hell out of your traps. RNG is a fickle mistress.

Unknown

All abilities that require picking a tile is picked completely at random when used by the AI. They don't care if they picked the bottom left corner and the tile to the right of it for Twin Pistols, or any other combination. Most likely the AI doesn't even know what any of the ability that requires picking a tile does, which is why it selects the tile randomly.

Unknown

Sure, but the code to exclude the outside rim of tiles couldn't be THAT hard, and would see a great improvement on those abillities when used by the AI. It's funny enough when AI Hood overlaps Twin Pistols, but to get 4 AP for one shot is just cruel.

Unknown

Lerysh wrote:

Sure, but the code to exclude the outside rim of tiles couldn't be THAT hard, and would see a great improvement on those abillities when used by the AI. It's funny enough when AI Hood overlaps Twin Pistols, but to get 4 AP for one shot is just cruel.

Problem is that the AI doesn't know what the ability does. While picking the corner would be bad for Twin Pistols, picking the corner would usually be very good for a Star Spangled Avenger. I'm pretty sure the AI literally has no idea what any of the targeted ability is supposed to do. It's probably not very hard to at least put some general guidelines like 'don't pick near corners' (Twin Pistols) or 'stay away from red tiles' (Star Spangled Avenger) but for whatever reason there's no plan to do this yet.

Unknown

Even if all abilities use the same "pick a random spot" any ability with a 3x3 can use a small loop.

Done=false
While (notDone=True)
If (picked tile on outer edge) pick again
else Done=True

There, done. Put that in every 3x3 tile pick ability (in actual code form and not just a logic list) and it will start selecting smarter tiles, albeit at a small delay as the function call goes multiple times, but it seems they are looking for delay anyway so... that's good right?

sidenote: I almost always put my Cap shields on a red tile, one that is far away from other red tiles. By not increasing the amount of red I reduce the chance of it getting matched somehow randomly.

Unknown

The logic is not hard but I don't know if it's laziness or intended but the game's AI is programmed to purposely not understand what any of the moves do. For example if given the choice of:

6 green AP - Prehistoric Bite for large amount of damage
10 green AP - Hypothetical ability that does only damage and less than Prehistoric Bite

The AI will always pick the 10 green AP ability if it has 10 or more green AP, even though a very simple analysis would show that it is never correct to use this hypothetical ability over Prehistoric Bite. And realistically almost no green ability matches up Prehistoric Bite even considering the side effects, especially consider Prehistoric Bite can often KO a character from 100 to 0.

This design premates throughout the game AI and it is not possible to put any meaningful refinement on the AI if it purposely acts as if it has no idea what any of its moves do.

Jathro

If I can play devil's advocate...
It's possible the AI is "smart" enough to pick positions that it would have gained more priority colors by targeting a specific area, even if it meant destroying less tiles.

Unlikely, but possible.

scottee

Oh, they could have programmed it that way. It's intentional. They've stated previously that smarter AI made players lose too much. Which equals quitting. So they made the AI dumber on purpose.

Unknown

^ yeah i think they mentioned it about how the AI was always taking advantage of connecting 5 tiles together whenever possible, which naturally caused a lot of cascading and damage... which is why the AI sometimes misses connecting Critical tiles or 5 tiles together when its not in a straight line...

Unknown

scottee wrote:

Oh, they could have programmed it that way. It's intentional. They've stated previously that smarter AI made players lose too much. Which equals quitting. So they made the AI dumber on purpose.

This just makes me very sad. Maybe it's the Battletoads veteran in me but harder does not equal worse.

Also, didn't they recently up the scaling limit to 400 because the game was to easy? Once again pulling in opposite directions.

Unknown

Lerysh wrote:

scottee wrote:

Oh, they could have programmed it that way. It's intentional. They've stated previously that smarter AI made players lose too much. Which equals quitting. So they made the AI dumber on purpose.

This just makes me very sad. Maybe it's the Battletoads veteran in me but harder does not equal worse.

Also, didn't they recently up the scaling limit to 400 because the game was to easy? Once again pulling in opposite directions.

The scaling was more like a response to Spiderman. I haven't seen any standard encounter to higher than level 200 not counting the all goon fights (those tend to be higher level) or fights where there aren't 3 opponents (1 guy at level 390 = 3 guys at level 130 in MPQ logic), and the level 200 guys are often putting up a good fight against even Magneto. If Magneto is also nerfed, you wouldn't need to scale anything past level 200 to wear down players.

Unknown

I've been thinking about the AI and it'd be very problematic if the game was balanced. The current AI would likely be too good for the average player, and while scaling can take care some of that (by giving even lower level enemies), there's a fundamental problem in that the AI's strategy is overwhelmingly dominant if the game goes on long enough. The AI, with its iron man endurance, has a rather simple strategy of trying to quasi-optimally maximize AP gain. For a short period of time it is possible to keep up, but if you even superficially keep track of the amount of AP the AI is gaining versus yours on an equal footing (and factor out stuff the AI is programmed to be unable to use, like any Magnetic Field type effect) you'll notice the AI gradually catches up and overtakes you if the game goes on long enough. Now currently with how broken all the offense is you can easily put the game away before the AI's AP advantage comes into play, but without the broken offense things will get progressively worse for the player over time. At some point you're going to hit a state of "I'm waiting for my sub award on the sub that never happened" and the AI will start slapping you silly. So for things to be even, you've to have the AI gradually scale back on its precision as the fight drags on because it has to factor that human players get tired/lose focus over time while it does not, and that'd be quite complicated to figure out.

SunCrusher

Was it you, Phantron, or someone else in the forum who said the equivalent of the AI is ALWAYS guaranteed to be better if only because it can never really make a mistake?

Either way, I agree with your thought; a long drawn out battle with a smarter AI is going to inevitably lead to mistakes... from the player end. The computer never tires but people do and that's important to remember.

To illustrate, there's a game called Cruel Jewels on iOS. It's a match-3 that's closer to the Bejeweled genre... but with a twist as there's AI intelligence scaling in effect in the you versus AI modes. In the potentially neverending you versus AI mode where you go through a gauntlet of match-3 exercises before ending the series with a real time 'battle' against the AI (sort of like how MPQ works), the further you get along in those gauntlets the stronger and smarter the AI becomes. It is opportunistic (will always pick THE best option after you go first; never any room for mistakes so don't leave any 5 matches behind!) and if you're even remotely tired, the AI WILL end up kicking your ****. If you play Fisticuffs mode, you'll get to play consecutive rounds of you versus AI where the AI gets smarter and smarter and smarter until you're pretty much guaranteed to die.

If anyone's a fan of Pokemon HeartGold or SoulSilver, the same thing with a 'smarter' AI is applied to the game of Voltorb flip where the further along in points you get, the dicier and dicier each move becomes with the AI very visibly taking on a more almost 'human' and 'psychological' approach in regards to the placement of the cards. The card that has the highest probability of having a Voltorb behind it JUST MIGHT also be the 'safe' card. Are you willing to risk it, are you?!

Both of these games are fun as all hell, but unlike MPQ, 'dying' isn't any kind of a huge deal (unless you're the sort to ragequit over missing surpassing your previous best score - I'm not). It isn't competitive as they're both single-player only and dying doesn't do anything save for make you start all over from the beginning again.

In MPQ, people (including me!) complain about characters dying from unlucky cascades and firing abilities all the time. Dying means either waiting (people get horribly impatient) or spending health packs (people don't like spending health packs if at all possible). Dying means the difference between one tier of rewards and another tier of rewards.

Smarter AI that's closer to adult human level means LOTS more dying and potentially lots more of what we would call pre-mature dying.

Personally, I've seen plenty of instances where the AI has been 'super smart' and plenty of instances where the AI was 'super dumb' and I've had my own gripes when the AI manages to string together a gigantic cascade from the top row or straight down vertical so... no thank you. My Hood I went against was at least smart enough to target two different spots (couldn't knock him out fast enough or steal enough yellow because I had a hugely yellow-dominant board that time somehow) and managed to collect a ton of AP from both.

If people are STILL griping over getting unlucky cascades and the AI getting too many lucky cascades despite all the reassurances that both AI AND player get their share of luck and lack of luck, increasing the AI's intelligence is only going to make it worse. This time, however, it really might just zap what's left of the fun from this game.

Just my two cents, though.

Unknown

See, to me, the "dying is a problem in MPQ" is the root problem, and not "The AI needs to be stupid or else it would murder the average player". Dying in games hasn't been a problem really since the days of Mario. You just start again. Like it never happened. That's what MPQ needs, in my opinion. Dying costs way too much, and it's inevitable, even in perfect play you lose HP to match damage. A "flawless" victory is one that ends in maybe 5-7 turns where you only take 200ish match damage over those turns, 1200 average damage over about 6 matches and SOMEONE is dead.

Why does the system punish people who actually play it?

Smarter AI leads to smarter players. The "average" would go up. Those who can't beat the AI settle into the MMR pattern of their competence level. It's the same as it is now, just with larger extremes.

Unknown

Lerysh wrote:

See, to me, the "dying is a problem in MPQ" is the root problem, and not "The AI needs to be stupid or else it would murder the average player". Dying in games hasn't been a problem really since the days of Mario. You just start again. Like it never happened. That's what MPQ needs, in my opinion. Dying costs way too much, and it's inevitable, even in perfect play you lose HP to match damage. A "flawless" victory is one that ends in maybe 5-7 turns where you only take 200ish match damage over those turns, 1200 average damage over about 6 matches and SOMEONE is dead.

Why does the system punish people who actually play it?

Smarter AI leads to smarter players. The "average" would go up. Those who can't beat the AI settle into the MMR pattern of their competence level. It's the same as it is now, just with larger extremes.

If dying doesn't slow you down that usually implies all your characters heal to full every game. PvE would be even more of an endurance match than it is now, and PvP would be a complete exercise in futility for those who do not shield since nobody would ever be afraid of anyone else if their teams are remotely on the same tier when all your characters heal to full every match. Of course there's also monetization issues for having health packs, but even if D3 doesn't care about money, the game would still be a lot worse than it is now if dying is no longer a problem. You'd need to redesign all PvP and PvE to accomodate for that. I'm not saying that's a bad thing, but it's going to be a lot of work to figure out how to make it work.

SunCrusher

Lerysh wrote:

See, to me, the "dying is a problem in MPQ" is the root problem, and not "The AI needs to be stupid or else it would murder the average player". Dying in games hasn't been a problem really since the days of Mario. You just start again. Like it never happened. That's what MPQ needs, in my opinion. Dying costs way too much, and it's inevitable, even in perfect play you lose HP to match damage. A "flawless" victory is one that ends in maybe 5-7 turns where you only take 200ish match damage over those turns, 1200 average damage over about 6 matches and SOMEONE is dead.

Why does the system punish people who actually play it?

Smarter AI leads to smarter players. The "average" would go up. Those who can't beat the AI settle into the MMR pattern of their competence level. It's the same as it is now, just with larger extremes.

I very much agree and I DO enjoy the challenge otherwise I wouldn't be playing something like Cruel Jewels on Cruel mode in Fisticuffs.

To add on to what you just said, one of the struggles with MPQ as it is right now is that unlike a game like Super Smash (Game Cube version, mind) where you have a game cart and I have a game cart and we're playing versus and dying is just inevitable for either of us and it's your skill versus my skill and that's kind of it, in MPQ, people with more money and resources can push their way PAST that 'dying' point or, in some cases, help prevent the 'dying' problem from even happening in the first place.

Like you said, dying currently costs a lot, but it doesn't cost players equally and dying, in this case, doesn't necessarily end up being about skill.

If we're talking about skill versus skill alone: Those who can afford Health Packs pick themselves up and run again after being beat up; those who cannot get to limp behind. (What you talked about re: the mechanics of dying need to be revisited touches on this exact issue.)

If we're talking about other ways to buff gameplay so that you have a better chance to avoid dying: Those in super strong Alliances can get/request strong loaner characters, the people who can climb PVP and can afford to shield-hop can collect better TU characters, the people who can afford all the best characters are, in general, going to hold out better and also have the best point nodes in PvE, period.

IF the dying system changed, then obviously some of this wouldn't be such a disparity such as the Health Packs issue.

But I personally sincerely doubt D3 is going to change their current 'dying' mechanic and in which case, my above post stands. A more difficult AI coupled with the current 'dying' mechanic we have now is just not something I would look forward to and I've actually spent money on this game.

The fact that D3 went and removed Hot Dog stands from the Prologue still baffles me. Was 50 HP THAT disruptive to gameplay? REALLY?

over_clocked

Hood did, in fact, have X-Ray vision when under Dormammu's possession. He was able to recognize disguised Skrulls when Wolverine, Spidey sense, advanced Stark and Reed Richards tech and Dr Strange's magic could not. (Villain Sue! Yay!)

As for the AI, I would love it to be smarter in PvE and for PvE to be less of a grind and more of a challenge. I would like to see featured-related mini-challenges that award their separate bonus points and that you are guaranteed a 3* cover if you get the required bonus point score.
E.g. currently HT, Fury and Psylocke are buffed/required. Why not add a mini-game of 'Use HT's red to remove 4 enemy strike tiles'; 'Try not to let the enemy match a single Psi-Katana away', 'Win a game immediately after using Bewilder', 'Successfully trigger 3 Fury traps', 'Do more than x damage with Fury alone', 'Get Psylocke/HT do x damage while stunned' (attack tiles work through stun), etc, promoting the required characters this way. Of course this might be too difficult to implement right now, although I don't imagine it too difficult at all. PvE should be adequately hard without the scaling switch, and not reward grinding but skill and cunning instead. I miss the old Prologue days when the player had to learn the ways characters cooperate and could be crushed by the likes of Yelena and Venom (not ashamed to admit that).

Unknown

Lerysh wrote:

Smarter AI leads to smarter players. The "average" would go up. Those who can't beat the AI settle into the MMR pattern of their competence level. It's the same as it is now, just with larger extremes.

I think you'd be surprised by the sheer number of people driven away by a smarter AI. You have to remember the point of entry: if the AI is too 'clever' when I first pick up this game, then I'm not going to learn from it, I'm going to chuck it. The average would, indeed, go up - by virtue of pushing out the 'dumber' players.

gambl0r312

Bag-man might have it too! Scary i know

I had an instance in the just finished LRs where the AI used the Bag-man Team Up 'Webslinger' to "lock down" my only Attack tile from my Psylocke's Black skill.

Out of the entire board to pick, i am not sure of the probability %, to randomly pick the same single tile as my Attack tile....


I'll file it next to the "confirmation bias" cascade memes