MPQ AI - A Worthy Opponent

13»

Comments

  • scottee
    scottee Posts: 1,610 Chairperson of the Boards
    ShionSinX wrote:
    The staff lied before tho.

    Were there posts saying one thing and then later saying something else? Or was this proven some other way? I don't remember that particular incident.
  • Xenoberyll
    Xenoberyll Posts: 647 Critical Contributor
    simonsez wrote:
    GuntherBlobel wrote:
    It must be disappointing for them to release interesting new characters just to hear "Vision sucks! Quicksilver sucks! Starlord sucks! GSBW sucks!" when it's really the AI that sucks.
    Don't blame the AI for the suckage of Vision and QS

    The issue with GSBW/MMags like characters would be much lighter if they'd just place their tiles next to each other in a straight line. So they'd gain something guaranteed.
  • nwman
    nwman Posts: 331 Mover and Shaker
    Someone suggested this already but I'll say it again.

    The ai doesn't need to be perfect. Players, especially tired ones will miss match 5s from time to time.

    So have the ai on a randomized schedule of using smart ai or dumb ai.

    This makes it harder to predict and more like playing a person. If a match 4 or 5 isn't guaranteed to be matched it changes a lot. B
  • XandorXerxes
    XandorXerxes Posts: 340 Mover and Shaker
    scottee wrote:
    I like how in a thread that has a developer definitively prove that the AI doesn't cheat...we have people who are sure the AI still cheats.

    The source code they referenced isn't even close to the full code and isn't relevant to the thing that we consider "cheating" - the cascades.

    According to the code and the interpretation they post, when only Venom is left he should never match red over black. Yet, I've seen it done. However, that's anecdotal at best so let's look at some math. Note these percentiles won't be exact - I'm not factoring in the lower percent chance of the AI dropping in team-ups, instead I'm treating everything as having a 15% chance to drop. Therefore these odds should be a bit on the high side.

    The odds for a 5-in-a-row dropping after a match 4 according to the article is 0.050625%. That means every 10,000 match 4s, the AI should drop in a match 5 about five times. Those odds are nowhere close to what we see now. If we assume 20 match 4s a battle (both sides) that should happen once every 500 battles.

    How about a more common scenario - 4 tiles in a row are dropped in. The math on that is 0.3375% each time a match 4 is made, or about 3 times every 1,000 match 4s. Using the same 20 match 4 per battle, that's about 50 battles.

    Lets assume someone takes 75 battles in a PVP (I imagine this number is a bit high except for the highest-scorers). They should see a perfect chain 4-drop once or twice in any given PVP and a full 5 drop in only once every 6 or 7 PVPs they play. That's not just for the AI - that's at all.

    I'm fairly confident even without gathering the data that it happens a lot more often than that. If the code is truly designed to be random, then their random number generator isn't sufficiently random.

    Disclaimer: this does not account for the "exact tile" a player needs to drop into a 4 or 5 match landing in the right spot, only for 4 or 5 tiles in a row of the same color dropping in.
  • xellessanova
    xellessanova Posts: 183 Tile Toppler
    I was reading this thread and the comment thread on the blog post and thinking, "Man, why are people so conspiracy theorist all the time here."

    And then I realized that this is an audience that is specifically self-selected for strong pattern-matching abilities and Pavlovian-conditioned to take note of large color matches.

    It's basically the equivalent of giving a couple hundred thousand people who love to consume toast a part-time job testing toasters for best overall toasting, and then telling them that no, the toasters aren't supposed to burn designs in the toast, it just looks that way because you're saving every toast with something like the face of Jesus/Queen Elizabeth/Homer Simpson on it and eating all the ones that don't.
  • scottee
    scottee Posts: 1,610 Chairperson of the Boards
    XandorXerxes wrote:

    I'm fairly confident even without gathering the data that it happens a lot more often than that.

    I think you need to reconsider your testing criteria.

    And if you want worthless anecdotes, I just ran up to 1100 in Rag pvp in one sitting and didn't have a single match-4 drop after a previous match-4. I actually haven't seen it in a while. But my anecdotes are worthless without lots of people meticulously recording results.
  • Xenoberyll
    Xenoberyll Posts: 647 Critical Contributor
    scottee wrote:
    XandorXerxes wrote:

    I'm fairly confident even without gathering the data that it happens a lot more often than that.

    I think you need to reconsider your testing criteria.

    And if you want worthless anecdotes, I just ran up to 1100 in Rag pvp in one sitting and didn't have a single match-4 drop after a previous match-4. I actually haven't seen it in a while. But my anecdotes are worthless without lots of people meticulously recording results.

    How are the odds for 3, 4 or 5 match 5s falling from the top? I've seen that happen multiple times. I wish i was this 'lucky' playing the lottery icon_e_wink.gif
  • Unknown
    Unknown
    I really enjoyed this blog post; I've always wanted more insight into the inner-workings of this game. Force Justin to write more!

    The current dumb-AI serves its purpose quite well. Most people just want quick, easy games during their daily grind.

    Still, I'd love to see the smart-AI get some use in the future. I can imagine playing against a seemingly standard Thanos enemy -- then suddenly noticing all the smart **** he was doing. That would definitely bring a smile to my face. It would feel like a real boss fight.

    Maybe add a single-node "Grandmaster Challenge" event featuring the smart-AI, then slap it next to Deadpool's Daily Quest and rotate it once a week?
  • Arctic_One
    Arctic_One Posts: 133 Tile Toppler
    I read it on the internet, it must be true. lol
  • Linkster79
    Linkster79 Posts: 1,037 Chairperson of the Boards
    I wouldn't mind seeing a hard mode Prologue that is only accessible after you have attained say 2 or 3 rewards on every node of the original Prologue. Would give us something else to do. Of course hard mode would have to offer far better rewards.
  • atomzed
    atomzed Posts: 1,753 Chairperson of the Boards
    Ummm, if a clear post by the D3 with the source code cannot convince ppl that the AI is not cheating, then there's nothing that anyone can do to convince this group of people.
  • TLCstormz
    TLCstormz Posts: 1,668
    GuntherBlobel wrote:
    But whatever, I totally understand why a smarter AI will never happen. See TLCStormz's reaction above.

    I have no idea where you are coming from, but yes.....see my reaction to the REALITY of the game for a non veteran / normal person.
  • TLCstormz
    TLCstormz Posts: 1,668
    DukeUpdate wrote:
    One thought on the enhanced AI possibilities is that the AI could be improved when certain characters were on the team.

    For example, if Beast.....Doc Ock.....

    It could add a new dimension to the game by improving some underpowered characters.....

    Now THAT sounds amazing!!!!! :+D

Categories