The ‘lucky’ Greg

1234568

Comments

  • joerginger
    joerginger Posts: 198 Tile Toppler
    What you just described is happening all the time. Sometimes you're forced to stare at the screen as the AI keeps matching. Which is even more annoying when it's doing that with an empty hand because it's already played all its cards after the first cascade.
    This is not a game. This is torture.
  • FindingHeart8
    FindingHeart8 Posts: 2,731 Chairperson of the Boards
    well...that escalated quickly :o 
  • hawkyh1
    hawkyh1 Posts: 780 Critical Contributor
    the thing that bothers me in the battles is the reckless
    nature of the rng. the best way I can explain it is let's
    say both the player and the ai have a 50% average.
    if the range was from 25-75%(average 50%) then a
    bad dice roll would mean 75%/25% or at worse case
    a 3-1 disadvantage for that one turn. now if the game
    was to be coded to range 1-100%(average 50%) then
    a bad dice roll results in 100-1 disadvantage for that
    turn. this is a problem especially during streaks of
    bad luck of 2 or more turns. it's no good saying that
    my luck will change to get a 100-1 advantage in my
    favour(it evens out the 100-1 against me) on turn 5
    when I'm dead by turn 4?
    rng should never on it's own govern the possible
    outcome of a battle. it should be a combination of
    deck/cards, rng and how well each player played
    that determines the possible outcome.
    this is why I always feel an unfairness in the gem
    matching despite others presenting the case that
    technically there should be no advantage. averaging
    50% in both cases, there's a big difference between
    rolling a 1d100 and a 25d3.

    HH
  • hawkyh1
    hawkyh1 Posts: 780 Critical Contributor
    octal9 said:
    I've posted several screens of myself gaining 3, 4, even 5 extra swaps in a row (yes, without GAH/Aggravated Assault shenanigans) on the TP9 discord. It's not unheard of for players.

    The AI weights own mana gain + player mana denial over gaining extra swaps. A swap that generates two matches like this
    r   g
g<->r
r   g
    is weighted as a better swap than one like this
    r   g
r   g
g<->b
r   g
r   g
    when the mana differential in the first is greater than the one in the second.

    I thought a match 5 was a rare occurrence and that's
    why you are given an extra turn. being able to chain
    5 match 5's before passing up your turn represents
    too large of a spread in mana? more powerful cards
    usually cost more mana(versus similar rarity cards).
    this just encourages it to be a game of luck where
    any player(including ai) can just win by having a
    very lucky turn?

    HH
  • FindingHeart8
    FindingHeart8 Posts: 2,731 Chairperson of the Boards
    octal9 said:
    I've posted several screens of myself gaining 3, 4, even 5 extra swaps in a row (yes, without GAH/Aggravated Assault shenanigans) on the TP9 discord. It's not unheard of for players.

    The AI weights own mana gain + player mana denial over gaining extra swaps. A swap that generates two matches like this
    r   g
g<->r
r   g
    is weighted as a better swap than one like this
    r   g
r   g
g<->b
r   g
r   g
    when the mana differential in the first is greater than the one in the second.
    I'd imagine that it wasn't an impossible feat for players, more a bout of stupendous luck.

    I've never surpassed the luck of 4 swaps without the help of gem-conversions or other tricks.  To see the AI pull 5 natural swaps on a fairly evenly distributed gem board was a "Woah" moment for me.  Not saying anything conclusive from the experience, merely taking note.
  • Bil
    Bil Posts: 831 Critical Contributor
    edited March 2018
       I'm overall skeptical about the omniscient IA theory but i must confess the rate of AI cascades since the fix is quite confusing. 
       I'm not sure the upgrade of the IA was just about better swaps among the ones on screen. Could he also be able to deal with probabilities, thus maximizing the chances of cascades?
  • FindingHeart8
    FindingHeart8 Posts: 2,731 Chairperson of the Boards
    Bil said:
       I'm overall skeptical about the omniscient IA theory but i must confess the rate of AI cascades since the fix is quite confusing. 
       I'm not sure the upgrade of the IA was just about better swaps among the ones on screen. Could he also be able to deal with probabilities, thus maximizing the chances of cascades?
    I would suspect that is more likely than omniscient AI.  Presuming the gem drop is entirely random, the AI might be able to calculate the probability of (if multiple gem swaps are available) which swap would yield a higher probability of a cascade.  This could be its general programming; where it may be be superseded from conditional board states that other players have observed (an if/condition1 exists, make swapA, if condition 2 exists, make swapB, if neither condition exists follow general programming).
  • FindingHeart8
    FindingHeart8 Posts: 2,731 Chairperson of the Boards
    edited March 2018
    octal9 said:
    FindingHeart8 said:

    the AI might be able to calculate the probability
    it might be able to but it doesn't

    FindingHeart8 said:

    This could be its general programming 
    It could be but it's not

    I've posted previously what the AI takes into consideration when making a gem swap. I'll do so again. These are in no particular order after the first:

    <b>Whether the swap is a match (number of gems doesn't matter, 3=4=5 to the AI at this point)</b><br>
Whether an activated gem will be hit.<br>Whether there's an extra swap available.<br>Whether the match will fill a card in the AI's hand.<br>The color of the match (i.e. bonuses for AI/player). a calculation is made here to quantify swaps by mana differential: mana gained for AI + mana denied to player.<br>Whether a support will be hit.

    If it's not on this list, it's not in the AI's calculation. Period.

    edited for the following disclaimer: all claims are accurate at the time of their posting. They may not hold true in the future.
    ah, good to know.  I must've missed that explanation.  Thanks for sharing. :)


  • Deamon1
    Deamon1 Posts: 17 Just Dropped In

    I was playing HoR last node my Ob against Bolas. I had a shadows of the past out. I watched round after round how Greg would match greens and white over ANY gem that was trapped with SOP. It wasn’t until there was no other option that Greg matched the traped gems. How and why would this happen if he didn’t know what was on the board?  
  • rafalele
    rafalele Posts: 876 Critical Contributor
    edited March 2018
    Brigby said:
    GrizzoMtGPQ said:
    Maybe the AI could just be this hard for Platinum players? That would be cool. Let the mortals play against the easy AI.
    The AI is primarily being toned down for content aimed towards newer players. Events will still have strong AI, especially the further into the event players go, however not to the extent that Ixalan had.

    So as Brigby said here there are different levels of IA. Not everything is in our minds.
    https://forums.d3go.com/discussion/comment/782628/#Comment_782628

    This explains to me why TG is easier than events.

  • bk1234
    bk1234 Posts: 2,924 Chairperson of the Boards
    edited March 2018
    Deamon1 said:

    I was playing HoR last node my Ob against Bolas. I had a shadows of the past out. I watched round after round how Greg would match greens and white over ANY gem that was trapped with SOP. It wasn’t until there was no other option that Greg matched the traped gems. How and why would this happen if he didn’t know what was on the board?  
    As far as I've ever seen AI has gone to great lengths to avoid my colors when I play anything with trap gems (just as I do when AI plays trap gems). This has been this way since the beginning. 

    Haven't seen this in awhile because I haven't played anything with trap gems -- but I can tinker in TG this week. 

    It used to be even worse, AI would even target my colors when I had supports out.  Koth would hit a blue match every single time if it meant taking out my support. 
  • ArielSira
    ArielSira Posts: 520 Critical Contributor
    Deamon1 said:

    I was playing HoR last node my Ob against Bolas. I had a shadows of the past out. I watched round after round how Greg would match greens and white over ANY gem that was trapped with SOP. It wasn’t until there was no other option that Greg matched the traped gems. How and why would this happen if he didn’t know what was on the board?  
    Shadows says "trap 4 black" so Greg tries to avoid black gems, just like a person would. Or am I missing something else here?
  • DBJones
    DBJones Posts: 803 Critical Contributor
    Greg used to even purposely target gems you trapped. Hopefully the new change will put the AI somewhere between these two extremes, otherwise trapped gems will be nearly worthless.
  • Mojo
    Mojo Posts: 44 Just Dropped In
    I have witnessed the canscdes that come after a 5 match has been ignored.  I have witnessed new Greg knowing which gems are trapped making trapped cards kinda useless. I have also noticed how much harder my activated gems are to match now compared to before while Gregs are easy pickings.  I like the challenge, however  they should disclose these changes.
  • ZW2007-
    ZW2007- Posts: 812 Critical Contributor
    octal9 said:
    The AI knows which gems are trapped.  ;)

    this is is the closest it comes to cheating
    Thing is, I do too. Takes a bit of trial and error to find them but you can find all the enemy trapped gems. I did enjoy when Greg used to prioritize my trapped gems like he was depriving me of hitting my own activate gems.
  • Gormhaus
    Gormhaus Posts: 190 Tile Toppler
    This is what is getting my goat. Round 1 me 6 mana. Greg 53 mana 9 loyalty. I get presented with one move and his move is a monster cascade.
  • Sarahschmara
    Sarahschmara Posts: 554 Critical Contributor
    Gormhaus said:
    This is what is getting my goat. Round 1 me 6 mana. Greg 53 mana 9 loyalty. I get presented with one move and his move is a monster cascade.
    Why does RNJeesus hate you?!?

Categories