Good Difficulty vs. Bad Difficulty

Dormammu
Dormammu Posts: 3,531 Chairperson of the Boards
There are a lot of threads out there right now complaining about the difficulty in PvE events. I'm going to try and be more constructive and not sound like I'm complaining. Not sure how it will turn out.

There is good difficulty in games and bad difficulty. Good difficulty challenges the player to be better - to increase in skill. Bad difficulty simply makes things more annoying. Unfortunately, MPQ employs bad difficulty.

Let's use a first-person shooter as an example of what I am talking about. A good method of increasing difficulty in a game of this type is to buff the intelligence of the AI; have opponents utilize cover better and play smarter with their fellow bad guys. Tactical stuff. A bad method of difficulty is doubling their health and making their attacks do twice as much damage. If you want to make a crossword puzzle more challenging you don't double the words required to complete it; that doesn't make the puzzle any more difficult, it simply means it takes longer to complete. Bad difficulty does not make the task more rewarding, it makes it more exhausting and tedious. Tripling the size of the puzzle is going to chase people away. This is the path MPQ has chosen to increase their difficulty - a poor choice. Really, a lazy choice.

The players aren't having any trouble defeating level 150 opponents? Let's make them 200th level. Oh, they can defeat those too? Let's increase enemy levels to 250. To 300... wait, they are having some trouble defeating enemies with over 3x the health and damage they can do? Let's give them boosts to even things out. Wait, they can defeat level 300 enemies with the boosts? Let's increase the level max to 400.

THIS IS A POOR DEVELOPMENT CIRCLE OF HELL THAT WILL NEVER END.

Instead of adding bad difficulty, add good difficulty. Make opponents smarter. Why does the AI always go for the match-4 when a match-5 is so obvious? Why is it that the AI can't use Black Widow's ability to place green tiles next to other green tiles? Why does the AI seem to match away it's own attack/defense/strike tiles as though it can't see them? Why does the AI place priority on environment tiles that will only get them a useless hot dog stand? Why does Ares kill himself with his yellow ability?

Players like to complain about other players using OBW and Spider-Man to grind and heal PvE, driving up the scaling. It's not the players' fault that they are using the tools given to them within the allowable parameters of the game - it's a poor system for increasing difficulty. It's bad difficulty.

Someone needs to start examining ways of implementing good difficulty.

Comments

  • Dormammu wrote:
    Players like to complain about other players using OBW and Spider-Man to grind and heal PvE, driving up the scaling. It's not the players' fault that they are using the tools given to them within the allowable parameters of the game - it's a poor system for increasing difficulty. It's bad difficulty.

    It's not the player's fault for using the tools they're given. Although there seems to be a very good correlation between those using the broken characters and those getting ridiculous enemies to fight.

    Demiurge urgently needs to fix the broken characters. Setting the game to ludicrous difficulty was a half-hearted attempt at patching a bad situation. Unless they do that, "good difficulty" will turn out to be even more frustrating than "bad difficulty" as players get into permanent stun-lock from Spidey and endless cascades from Mags.
  • Dormammu
    Dormammu Posts: 3,531 Chairperson of the Boards
    Yes, adjusting characters is certainly an acceptable tactic in making a change to their difficulty.
  • Making the opponent smarter is an example of bad difficulty. This is because there's nothing stopping the AI from playing flawlessly and you'd simply lose badly most of the time if they did that, which is why the AI does not play flawlessly.

    A long time ago I saw an article about this guy who made a pool game and the users complain about how the AI is too cheap because on the highest difficulty the AI calculates all the angles and knows exactly how to hit the ball so that it always pockets the next ball with optimal efficiency, which left the developer puzzled because on the highest difficulty, the AI simply pockets a ball with 100% accuracy and it always picks the closest one for the sake of simplicity (even though it can pocket any ball anywhere). The AI isn't cheating because it is simply good enough to pocket a ball 100% of the time from anywhere. Of course, the dev then thought about the users complaints and decided that having an AI that pockets the ball 100% of the time is probably not someone you want to play, so he went back and made the AI dumber.

    Programming a flawless AI is not hard. It's also not cheating in the same way that a AI pool player can pocket 100% of the time is not cheating. He is simply that good. You can simply have an AI that makes a move that results the maximum AP acquired based on the known board state every time (and it'd be pretty easy to maximize even against unknown tiles) and this literally takes no time to compute because all it has to do is simply make every move that can be made on the board (can't be more than 20 at most), while not allowing any new tiles to drop, and simply observe what happens to the existing board out of those moves and pick the best one. For some slightly more sophistication you can even calulate what player move can be made after all of your moves and take the move that has the greatest (your gain - player's gain). But is this what you want? Do you want the AI to always pick the best move that leaves you with nothing useful to match every single move? Again there is no cheating involved. Even a human player can calculate this if given enough time, and it's trivial for the AI to compute this, but it wouldn't be good gameplay.
  • Dormammu
    Dormammu Posts: 3,531 Chairperson of the Boards
    When it comes to match damage, I don't see why the AI can't be perfect. This game is won and lost with abilities and strategies that go way beyond missing a better match. But if it's a problem, the AI can be dumbed down. Add a coin flip to make it flawed - 50% of the time it makes the optimal match, 50% of the time it doesn't.

    Look, I'm not a programmer so the last thing I want to start doing is telling the developers exactly how to make it better. I don't know how. I just know the path they've chosen is not ideal.

    I'm not even saying the AI is the path to go. There are any number of other ways to add difficulty. What about adding timed matches, or adding more points awarded based on speed? How about two-on-three matches?
  • Phantron wrote:
    Making the opponent smarter is an example of bad difficulty. This is because there's nothing stopping the AI from playing flawlessly and you'd simply lose badly most of the time if they did that, which is why the AI does not play flawlessly.

    As long as the AI isn't cheating (by knowing the future gems that drop into the board), I don't see an issue. A flawless AI is as impossible as a flawless player, because neither one can know what the board will look like next turn. And in reality, coding a flawless AI is a lot harder than you might think.

    For example, a flawless AI would have to make the most optimal move on the board. Trying to describe what the most optimal move is would take many paragraphs of text, which would translate into pages and pages of code.
  • The problem with talking about scaling is everyone's situation is different. I am as effected as everyone else by community scaling, but I don't ever see enemy levels above 170 (on hard) over the course of an event. Which is very manageable and enjoyable to me. Maybe by the last day I will see a lvl 230 (goon) on the very last 'non-essential' node, but I just do them once and ignore them thereafter.

    So, that leads me to believe that it's the actions of the individual that drive up scaling to untenable heights. That along with their 'past sins' that can't be as easily expunged through damage/losing. That diminished effect is probably because there used to be crazy people that would play 12 hours a day and "tank" their PvE MMR for half it, to insure reasonable levels.

    But, I'm not looking over these people's shoulder to confirm/deny how they are playing an event. So, it's probably irresponsible for me to point at finger at them as their own worst enemy.

    To me, an ideal situation would be one where those that grind down every node/refresh to 1 point are punished by scaling. Those that leave every battle fully healed are punished by level scaling. Those that wait until the last hour to join are punished by level scaling. But those 'punishments' are not remembered the next time they join an event.
  • Any AI that is sufficiently awesome is the same as cheating, to the player. The 100% accurate AI in pool is not cheating. It is just that good, but good luck convincing the guys playing against the said AI that this isn't cheating.

    And if you have to throttle on how awesome the AI can be, then I feel MPQ already throttles its AI pretty well. The unknown tiles in MPQ are actually very easy to model. Basically all you do is tell the AI that matching on the 'X' on something that looks like 'RRX' is good (because R can drop and cascade again, and the worst that can happen is someone flips that around into a match 3 red which is something you can live with), and 'RRXRR' is very bad (because if R drops above or below the X your opponent flips that into a match 5, and chances are twice as good the R will drop above or below the X as opposed to the right row). If you put all that together what's going to happen is you'll have an AI that seems to get every lucky cascade (because it can compute the best move that leads to them) and you're usually stuck with matching 3 junk tiles. That alone should be enough for the AI to win comfortably even with its current skill usage, but it's not like it's that hard to teach AI concepts like 'Call the Storm = good'.
  • Infrared
    Infrared Posts: 240 Tile Toppler
    I would much rather have a scaling system where the AI progressively plays better the more times you beat it. Than one where enemy levels are simply boosted to far above the level cap, all while the AI consistently makes the same gameplay mistakes. Just as exploiting Spider-man stunlock becomes boring after a while, so too does exploiting AI weaknesses take away some sense of accomplishment from winning a tough match.
  • Phantron wrote:
    Any AI that is sufficiently awesome is the same as cheating, to the player. The 100% accurate AI in pool is not cheating. It is just that good, but good luck convincing the guys playing against the said AI that this isn't cheating.

    And if you have to throttle on how awesome the AI can be, then I feel MPQ already throttles its AI pretty well. The unknown tiles in MPQ are actually very easy to model. Basically all you do is tell the AI that matching on the 'X' on something that looks like 'RRX' is good (because R can drop and cascade again, and the worst that can happen is someone flips that around into a match 3 red which is something you can live with), and 'RRXRR' is very bad (because if R drops above or below the X your opponent flips that into a match 5, and chances are twice as good the R will drop above or below the X as opposed to the right row). If you put all that together what's going to happen is you'll have an AI that seems to get every lucky cascade (because it can compute the best move that leads to them) and you're usually stuck with matching 3 junk tiles. That alone should be enough for the AI to win comfortably even with its current skill usage, but it's not like it's that hard to teach AI concepts like 'Call the Storm = good'.

    If the AI made matches "perfectly," the player still: Attacks first, can choose what heroes to target and are targeted, use boosts, strategize with abilities... every advantage is still on the player's side.

    And since this "improvement in AI" would be applied to everybody, what it means is that... the people who play better will do better. Oh noes!
  • over_clocked
    over_clocked Posts: 3,961
    Feeding enemies AP from goons seems an example of bad difficulty to me too.
    I sure as heck don't want to play against a perfect AI, but having it a tad smarter would go a long way. At least make it not destroy its own special tiles (barring Redwing or moves leading to cascades). Playing against level 300 opponents (especially 1*s) feels like tiptoeing around a stupid angry troll, NOT being sufficiently (intelligently) challenged, because you can only allow them to make so many moves before they kill you with simple match damage or cheapass abilities that require no spark of thought, like Headbutt or Prehistoric Bite or Pheromone Rage.
  • Getting AP fed sometimes make sense if it's an interesting ability like Illusions. Unfortunately usually it's like lazy Thor getting fed green or Spiderman getting fed blue and there's absolutely nothing interesting about that.

    I think goons should not generate AP unless all the guys who can move the board are unavailable (stunned, dead, or not present in the first place).

    For the AI destroying their own tile bit, it looks pretty dumb but if you think about it, if you got an imminent match for your valuable special tile, and it's a useful color, you might as well take it because the human player sure will take it next turn if the AI does not. I guess the AI can attempt to try move that match away, but most of the time it's simply not possible.