Announcement

Collapse
No announcement yet.

Data Mining All the Adversaries

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Data Mining All the Adversaries

    Hi everyone, mdlthree here again with overkill analysis.

    I did a similar analysis back in November with all the adversaries described up to that point. I was particularly interested in figuring out a monster HP formula, to which I was successful. With the release of the beastiary I now have a data set of 120 adversaries to analyze. Check out this quick tumblr picture of a correlation matrix between all the abilities plus some other derived stats. The most interesting parts I think are the red dots that mean negative correlation. HP and Def are negatively correlated, meaning the developers have a sense of balance that avoids really meaty enemies that dodge a lot.

    http://herdingdice.tumblr.com/post/1...adversaries-in

    Another interesting note that will be looked at in the future is the scatter plots of various abilities versus inferred level. You can see this in the last row of the matrix (Threat.Level). The most revealing part of this is the intersection of Def and Threat.Level. There was a fairly active thread this week about hitting and defense. One would expect that as adversary level increases that defense would likely rise as a result of more points in DEX or perhaps equipment. In fact the data shows a zero correlation between level and defense meaning that all 120 adversaries basically have a random defense. That random defense is balanced against overall HP however. For a natural progression of adversaries one might expect slight increases in HP and DEF and the same time to convey increased difficulty. PC's are definitely going to spend points on both hitting and defending, so why wouldn't the adversaries.

    Here is the list of abilities with decreasing correlation with level.
    • Will 0.75
    • Str 0.64
    • Int 0.58
    • Con 0.54
    • Fight 0.53
    • AR 0.51
    • Comm 0.42
    • Perc 0.35
    • Acc 0.26
    • Spd 0.15
    • Dex 0.10
    • Def 0.05

    A little weird that Will Power is first place and most consistently increases with level. Second place feel right with STR. CON is fourth and should be a consisent increase. HP is not on the list but is higher than WP at 0.78.

    More insight to come with this thread in the future.
    [URL="http://herdingdice.tumblr.com/"]Herding Dice[/URL] - A tumblr where I put things about Adventure Game Engine design.

  • #2
    Re: Data Mining All the Adversaries

    That makes sense that Def is low because DEX is low, too.

    Comment


    • #3
      Re: Data Mining All the Adversaries

      Originally posted by mdlthree View Post
      Hi everyone, mdlthree here again with overkill analysis.

      I did a similar analysis back in November with all the adversaries described up to that point. I was particularly interested in figuring out a monster HP formula, to which I was successful. With the release of the beastiary I now have a data set of 120 adversaries to analyze. Check out this quick tumblr picture of a correlation matrix between all the abilities plus some other derived stats. The most interesting parts I think are the red dots that mean negative correlation. HP and Def are negatively correlated, meaning the developers have a sense of balance that avoids really meaty enemies that dodge a lot.

      http://herdingdice.tumblr.com/post/1...adversaries-in

      Another interesting note that will be looked at in the future is the scatter plots of various abilities versus inferred level. You can see this in the last row of the matrix (Threat.Level). The most revealing part of this is the intersection of Def and Threat.Level. There was a fairly active thread this week about hitting and defense. One would expect that as adversary level increases that defense would likely rise as a result of more points in DEX or perhaps equipment. In fact the data shows a zero correlation between level and defense meaning that all 120 adversaries basically have a random defense. That random defense is balanced against overall HP however. For a natural progression of adversaries one might expect slight increases in HP and DEF and the same time to convey increased difficulty. PC's are definitely going to spend points on both hitting and defending, so why wouldn't the adversaries.

      Here is the list of abilities with decreasing correlation with level.
      • Will 0.75
      • Str 0.64
      • Int 0.58
      • Con 0.54
      • Fight 0.53
      • AR 0.51
      • Comm 0.42
      • Perc 0.35
      • Acc 0.26
      • Spd 0.15
      • Dex 0.10
      • Def 0.05

      A little weird that Will Power is first place and most consistently increases with level. Second place feel right with STR. CON is fourth and should be a consisent increase. HP is not on the list but is higher than WP at 0.78.

      More insight to come with this thread in the future.
      A while back I did a analysis of the adversaries and the defense (though no where near as indepth as yours). I too noticed that defense was almost an arbitrary number thrown in based on the thread level and rank of the monster. That is one of the things that makes creating your own adversaries so difficult. I haven't gotten the beastiary yet, but I would love a guide line for creating your own monsters.

      Comment


      • #4
        Re: Data Mining All the Adversaries

        This is fascinating. Thank you for taking the time to do this and to share it. Its raised some interesting questions and discussion.

        I just love our fans sometimes.

        Comment


        • #5
          Re: Data Mining All the Adversaries

          Originally posted by Jack View Post
          This is fascinating. Thank you for taking the time to do this and to share it. Its raised some interesting questions and discussion.
          I just love our fans sometimes.
          Thanks Jack! Provoking conversations is something exploratory data analysis is great at. If someone from GR wants a copy of the data set or wants to apply some different analysis to the data, I would be happy to collaborate.
          [URL="http://herdingdice.tumblr.com/"]Herding Dice[/URL] - A tumblr where I put things about Adventure Game Engine design.

          Comment

          Working...
          X