Auro's Single Player Matchmaking is a failed experiment

Discussion in 'Game Design' started by vivafringe, Jan 6, 2017.

  1. vivafringe

    vivafringe Well-Known Member Staff Member

    When Auro's single-player matchmaking was first envisioned, I was psyched. It seemed to fix many traditional problems with videogames by providing a clear goal that scaled with the player's ability. Unfortunately, expectations have not met with reality. The dream is dead.

    In Greek mythology, Sisyphus was banished to hell and forced to repeatedly roll a boulder up a hill, only to have it roll back down again. Human beings hate seeing their efforts wasted, so it's a resonant tale. For a moment, put aside what you think of a grindy RPG where you try to get to level 100. Now, imagine a grindy RPG where you get to level 100, and the game immediately resets and puts you back at level 1. It's revolting.

    Single player matchmaking as implemented in Auro fails because the endgame is Sisyphean. You stabilise at a particular skill that bounces between two difficulties. The easier difficulty is just easy enough that you can beat it, while the harder difficulty is hard enough that you can't. So you slowly grind up your boulder to the higher difficulty, only to watch it roll back down again. It's excruciating.

    In this case, the player probably *can* grow his skill to the point where he can overcome the harder difficulty and climb out of his Sisyphean hellhole. But to do that, he has to spend at least half his time grinding the easier difficulty, to get a chance to practice the harder one. The player would almost certainly prefer to spend his time exclusively practising the harder difficulty, but the game does not let him, since it thinks it knows better.

    In practice, the single player matchmaking model has failed not just Auro, but a lot of Brain Good Games as well. Don't get me wrong - all of these games are still good, IMO, because they have other strong design characteristics. But this part, in particular, significantly hampers my enjoyment of all of them. I will play until I hit a difficulty I can't beat, then immediately quit. It's become such a common pattern that I'm beginning to quit before that happens!

    I'm unsure what the best style of single player matchmaking is, but I am fairly sure it can only move *forward*, not backwards. Personally, I would recommend copying Hearthstone's arena mode. Here's a possible implementation:

    - Players start at difficulty 1.
    - They push a button called "start campaign", where they play a series of games.
    - At 3 losses, they die, and must start over at the same difficulty.
    - If they get 9 (or 12?) wins before they get 3 losses, they advance to difficulty 2.
     
  2. BrickRoadDX

    BrickRoadDX Well-Known Member

    Interesting thoughts! I have been playing a bunch of Militia's "Timed" mode after getting to high ranks, but I do go back and forth between that and traditional ranked mode.

    I like the idea of a Hearthstone-ish "arena" run system. Seems fun! Food for thought for sure. I like the idea of eliminating games where you're not on the right difficulty.

    (Lol @ "The dream is dead" :p)
    "ded gaems"
     
  3. Waterd

    Waterd Well-Known Member

    I will say though you should make a point of why this is different than human ladders.For example in warcraft 3 ladders, I peaked at position 23 and then for the rest of the time bounced between 150 and 100 for most of my warcraft career, it wasn't a problem. I'm sure the same happens to most people in league of legends ladder, and it doesn't seem to be a problem.
     
  4. Toad_Racer

    Toad_Racer Well-Known Member

    I just stopped playing for Auro rank after a while. Rank only interest me as an indicator of how I've been playing recently, and not as a progress bar I want to fill.

    This doesn't really happen to me in Auro. I've been hanging out in the 21-25 range for months now and there's no grind-fall pattern. A rank 21 match is still difficult enough to lose, and a rank 25 match isn't just a brick wall. I'm basically happy at any point in that range.
     
    keithburgun likes this.
  5. BrickRoadDX

    BrickRoadDX Well-Known Member

    I think the point that something "feels" off is certainly important here. I wonder what systems can be put in place to further enforce that the goal isn't really to climb the ladder, and to provide some additional motivating desire to play when "bouncing around".
     
  6. Nomorebirds

    Nomorebirds Active Member

    I'm pretty sure that Starcraft 2 will not demote you down a rank once a season of ladder has started, at least that's how it used to be back when I played it. If you get to gold league, for example, you cannot go back down to silver, no matter how much you lose.

    But if you start doing poorly, you can always redo your placement matches.

    This is the only game that does this form of matchmaking that I can think of though, and Starcraft is pretty notorious at creating huge loss streaks, where players can lose like 90% of their games, but people don't want to rank down so badly that they will never re-take their placement matches even if they keep losing. I would imagine that this is the reason not many other games use this system.

    The problem mainly seems to be that people see a number and they only want it to go up and never want it to go down. Secondary problem to this Auro thing, is that perhaps one of the rankings has too big of a jump in difficulty? I could definitely see it being a problem if like rank 10 was so much easier than rank 11 that you could have a 70-99% winrate in 10 but a 30% or lower winrate in 11. Creating this endless back and forth where you never get enough time to practice rank 11 before getting thrown back into a difficulty that has nothing to teach you. (But I want to add that this is certainly not the case for me, rank 12 seems close enough to rank 13, that I am always being challenged and always have more to learn.)
     
  7. richy

    richy Well-Known Member

    I agree this is just where the problem lies. Having it so the rank is a number that games award us. Like a touch on the forehead and blessing from God: "Well done, you are intermediate my child". If approval from something, even a videogame, is available in higher amounts, of course people want it. The best way is the simplest way, let players choose for themselves the difficulty level they play at.
     
  8. SwiftSpear

    SwiftSpear Active Member

    Nah, lots of things in life never end. It's less satisfying, but satisfying isn't particularly great for games. You need to need to have something the player is working towards, but it doesn't need to be tied up with a pretty bow. Getting better at the game is a totally valid goal for players to work towards.
     
  9. BrickRoadDX

    BrickRoadDX Well-Known Member

    @vivafringe certainly seems like you made a provocative post :)
     
    vivafringe likes this.
  10. vivafringe

    vivafringe Well-Known Member Staff Member

    When your ladder rank goes down, the rules of the game don't change. The opponents just get weaker. While the game still changes somewhat vs. weaker opponents, I'd argue it's less of a shift than a game like Auro, where monster compositions and score requirements can change pretty drastically.

    That being said, it can still be annoying psychologically in human ladders if it's emphasised enough. Starcraft 2 is actually a good example of people being really annoyed by dropping down in leagues. "Ladder anxiety" was coined around that time, of people getting their number up and not wanting to lose it.

    Fair enough. It's possible my experience is extreme, since I'm extremely vulnerable to extrinsic motivation. If there is not a clear goal I'm working towards, whether it's climbing to a specific leaderboard slot or finishing all my quests, I rapidly lose interest.

    In Auro I think I got to around level 17, where probably the jumps in difficulty are less smooth.
     
  11. Batlad

    Batlad Well-Known Member

    The ranking system in Auro doesn't actually try to estimate your skill, instead it just has some simple rules that determine whether you will rank up down or not. It could be possible to modify it so that it actually has an estimate or your rank and a confidence interval. It could then use a combination of the two to determine whether you should rank up/down or not. You could also bias the algorithm so that ranking down requires more confidence than ranking up.

    An alternative would just be to make ranking down optional, so that people will only rank down if they feel its in their best interests. The dialogue box for that choice could include advice.

    Edit: I should say I like this criticism and appreciate hearing other peoples thoughts on this subject.
     
  12. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    Nice clickbait headline! Got me to post while on vacation.

    The hearthstone thing you posted is like roughly the same as Auro, minor differences.

    Some points:
    - There are minor problems with the Auro system especially at the highest level
    - I do think it needs a "final level"; issue is just in determining where that should be but I don't want highest level players gambling. So I would be happy right now probably saying you've solved the game at rank 24 or something.
    - Every single player strategy game needs a system like Auro's.
    - For the vast majority of players, the system works great.
    - Even for the highest level players, it seems like you and Toad Racer disagree!

    I think options like:
    - Optional down-ranking
    - Allowing the player to choose their level
    - Just slightly tweaking the algorithm

    ... could all basically fix the problem.

    The idea that it's a "failed experiment" is crapola, because it's the first time anyone really tried to do this AND because all single player strategy games require it.

    With all that said, I appreciate the post.
     
  13. vivafringe

    vivafringe Well-Known Member Staff Member

    It's not at all clear to me that single player strategy games require matchmaking. The best systems I've experienced don't have even a hint of it. They include:
    • Imbroglio's Izu mode, where every 4 days players are given a new character and weapons to build their board with. Then they highscore chase. Since the highscore board rotates so frequently, you dodge the traditional issues with highscore play.
    • Super Hexagon, where you try to beat the game on all 6 difficulties. Simple but good.
    • FTL, where you try to beat the game with every ship (on normal, then on hard). Has a long shelf life, since new ships shake up the strategy space a lot.
    It sounds like your tweaks are pushing you more and more towards Super Hexagon's model of just having some difficulties, and letting players try to beat all of them. I agree that's an improvement for Auro. However, if that's the case, what is *left* of Auro's original system, as envisioned, that makes it so different from Super Hexagon's?
     
  14. BrickRoadDX

    BrickRoadDX Well-Known Member

    @vivafringe I think it goes back to engineering experiences. Many people do not know what is best for them/what will provide the most value over time. While this is a contentious claim for difficulty, it's not nearly as contentious for things like card design, systems design, etc, and I think they're actually very closely related in terms of creating an overall experience for a player.
     
  15. vivafringe

    vivafringe Well-Known Member Staff Member

    Yeah, I think it's pretty risky to completely control difficulty for the player. There's too much of a knowledge gap between the player and the game designer. The player is going to know if he is drunk, if his sister is playing, how fast he is learning, whether he is starting again on a new account, etc., and the game designer just has to make a guess about all this stuff or hardcode in stuff that takes it into account somehow.

    Ultimately I agree to some extent that players shouldn't be making lots of game design decisions. Civilization's map-tweaks are a good example of just way too much pregame customization. Ultimately "pick a difficulty" seems fine to me, though. I mean, the extreme of your position would be that players shouldn't be allowed to pick which game they want to play, haha.

    Notably, the tweaks Keith is suggesting mostly seem to be relinquishing control back to the player.
     
    Jon Perry likes this.
  16. Hopenager

    Hopenager Active Member

    Maybe this is something that would be better discussed in it's own thread, but in my opinion the foundational idea behind the whole single-player ladder thing, i.e. that there should always be a 50% winrate, is flawed.

    Of course, if you treat games as if the player only gets feedback from the victory/loss output at the end of the game, then the 50% winrate is obviously ideal. But that way of thinking about feedback is flawed. The player is learning about the effectiveness of their strategies while they play the game, not after it is done. Even in Auro, if I make an obviously stupid move, it isn't as if I don't realize that the move I made was stupid until the game tells me "you lose," I can tell which moves were good or bad immediately, based on how they affect my health and score.

    Also, the player doesn't interpret the win/loss output as being completely binary. Losing in Auro after having only scored 10 points is obviously worse than losing after scoring 50, and in general losing very early in a game is worse than losing near the end of a game. If you don't think of the feedback the player gets from winning/losing as binary, the idea of striving for a 50% winrate falls apart. In general I think it's better to have a very low winrate, and allow the player to judge how effective their strategy was based not just on if they achieve the goal, but how close they come to achieving it.
     
    vivafringe likes this.
  17. RyanRothweiler

    RyanRothweiler Active Member

    You only know what moves are bad because you know they lead you closer to a lose state. No choice in a system is inherently good or bad. The definition of the win / lose states is precisely how you know what moves are good or bad.
    Yes exactly, because the lose state has been define as "When you run out of health".

    The player does only get feedback from the definition of victory / loss states, but not only from/during the loss screen at the end of the game.
     
  18. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    I agree the current version is not perfect, although I think it is much better than Viva makes it sound. Actually, I called it clickbait before but actually the content of the post runs with the sort of false clickbait headline.

    Is there any problem with a Version 3.0 of Auro (which I'd someday like to make) that has the following tweaks to the system:

    1. Increase number of ranks (let's say, 1.5x to 2.0x as many ranks, covering the same skill area)
    2. Designating one rank near the top as the "final rank", beyond which you cannot level up
    3. Allowing players to play clearly demarcated "Custom Games" at any skill level, including ranks above the final rank.
     
  19. keithburgun

    keithburgun Administrator, Lead Designer Staff Member

    I think there are approximately 5 threads on this. :)
     
  20. richy

    richy Well-Known Member

    So much is said about the win or loss state being the most important piece of feedback, and yet in Auro at the end of a match, equal prominence is given to the ranking up or down. More prominence really, because it's a higher-quality graphic as well as coming after the win/loss notification, like "you won the match, and so, what that means is, your rank went up".

    Saying the built-in Elo system makes matchmaking transparent or whatever, and you don't have to worry about your rank, let the game sort all that out, doesn't ring true. The presentation is effectively telling us to see our ranking up/down as the most important consequence of each match.

    Not that I agree with the default mode being dynamically-adjusted difficulty, but if it has to be, what would be wrong with keeping all the Elo workings confined to the stats screen rather than being so prominent? Just let it do what it's there for and dynamically adjust the difficulty. As it stands, it seems to be shoving it in our faces all the time really just to show off that it has such a clever system, rather than it really being necessary.
     
    alastair, Redless, Hopenager and 2 others like this.

Share This Page