Close

    • Dark Prophecies Balance 101



      Howdy Shadow Era peoples! Mojumbo here today to tell you all about the balancing of our new expansion set Dark Prophecies. In particular, I’m going to focus on how we go about making balance decisions based off tracking win rates, tracking card usage stats, and playing/watching games. Expect to learn a few interesting factoids from this article!

      Live Server Stats

      First off, to get an idea of how balanced a given metagame is we need to keep track of all of the heroes’ win rates. Of course, we track how often the heroes are being used as well so we can determine which heroes are very popular and which ones aren’t as popular. For example, Eladwen is currently the most popular hero as she represents 10% of games played. She currently holds a win rate of 51%.

      Now it’s great to know that the most popular hero is only winning 51% of the time but there is more useful data we can extract out of this data to make it even more meaningful. For instance, we can determine Eladwen’s win rate both going first and going second: Eladwen wins 52% of the time going first and 49% of the time going second. The fact Eladwen wins 3% more going first than going second does imply a minor advantage to going first (but 3% isn’t too bad still) but we can then design cards like Layarian Seductress that offer an advantage when you’re going second over going first.

      After we have determined the heroes’ overall win rates we then break it down hero by hero. For example, our live server stats suggest that Eladwen’s best matchup is against Elementalis and her worst matchup is against Amber. Using this data, we could determine that Eladwen needs tools in the near future to help out against Amber’s big weapons and likewise Elementalis may need tools to better fight off Eladwen’s mage prowess.

      Here are some interesting factoids about Dark Prophecies balance:
      • We strive to have every hero producing a 50% win rate +/- 5% (between 45-55%).
      • No hero is currently holding a win rate of 55% or higher.
      • First turn advantage generally amounts to a difference of 2-3% higher win rate going first.
      • In terms of games played, the hero pool is quite diverse in this metagame.
      • The stats don’t include games where people leave too early (where you gain no gold or xp).
      • The average time for a game across all heroes is between 11-14 minutes.

      Card Usage Stats

      Naturally, balance goes further than just looking at individual hero win rates. We also have to look at the popularity of individual cards. To do this we look at how often a specific card was used when it was legal for it to be in a deck. After that we determine how many copies of the card were being used. For example, we could determine that Ley Line Nexus is really popular as a 2-of in your deck whereas a card like Ironhide Karash is really popular as a 4-of in your deck.

      This data is important because it helps determine why some heroes are producing high win rates and also validates the feedback we receive from the community. If a card is rarely being used but is being considered overpowered by a few then it likely needs more active coverage. If a card is being used heavily and is being considered overpowered consistently, then it probably is.

      Here are some interesting factoids about Dark Prophecies card usage:
      Aldon the Brave is currently the most used card
      • Cloak of obscurity is current the least used card

      Balance Testing

      Stats can only take you so far. After we have run the numbers it’s time for us to get testing. Testing can both mean playing the game or watching replays of other players playing. The Player Focus Group (PFG) is invaluable when it comes to generating lots of highly skilled replays to watch so we can determine how powerful a card is (or isn’t) when it is used correctly. From there, we can make the necessary nerfs/buffs as required.

      It is only after we have live server stats, card usage stats, and have done the requisite balance testing that we can make informed decisions to change cards on the live server. We are happy with how Dark Prophecies has turned out from a balance perspective and hope you are too! There are lots of new powerful cards and combinations that will allow you to make new decks that weren’t quite viable during Call of the Crystals.

      Well that does it for my look at Dark Prophecies balance. Hope you enjoyed the article, good luck in your games, and make sure you have fun! This is Mojumbo signing out.


      This article was originally published in forum thread: Dark Prophecies Balance 101 started by Mojumbo
      View original post
      Comments 80 Comments
      1. huynguyenquang's Avatar
        huynguyenquang -
        First. Great to know this.
        But do you think 50% +/- 5% hero win rate is still big? We are talking about a spread of 10% between the best and worst hero.
      1. Preybird's Avatar
        Preybird -
        Good read!

        I must admit I'm pretty pleased with DP. When I can run Baduruu consistently in the 280 range I'm a happy boy.
      1. Index's Avatar
        Index -
        Interesting to read, and I remember you said won't reveal these status to us...
      1. meeklosh's Avatar
        meeklosh -
        Good article and good work. It's true, some of the Hero are finally more playable than in COTC. I hope, it'll stay that way in the future.
      1. Direwulven's Avatar
        Direwulven -
        A much needed insight!

        Are there any minimal threshold usage index for cards? For instance, you mentioned Cloak is least used but is there sufficient usage to allow for its (sad) existence to continue?
      1. Index's Avatar
        Index -
        Lol, I don't even know what that Cloak card is!

        Edit: checked, I've got a foil piece of that! lol!
      1. Kross's Avatar
        Kross -
        Does this mean that we can expect a minor buff to cloak of obscurity?
      1. Flycheung's Avatar
        Flycheung -
        Apparently it's either Campfire Stories was not taken into count due it's CotC card (then why aldon was included) or I did not aware a new deck archetype involved lotsa campfire stories.
      1. Index's Avatar
        Index -
        Quote Originally Posted by Flycheung View Post
        Apparently it's either Campfire Stories was not taken into count due it's CotC card (then why aldon was included) or I did not aware a new deck archetype involved lotsa campfire stories.
        Campfire Stories is included in starter deck. A lot of starters are running it.
      1. Arikiro's Avatar
        Arikiro -
        Good article! now we know a bit more about what you do to balance the game, but what happen with heroes like Nishaven, Logan, ... maybe one day will be competitive?
      1. Direwulven's Avatar
        Direwulven -
        Quote Originally Posted by Flycheung View Post
        Apparently it's either Campfire Stories was not taken into count due it's CotC card (then why aldon was included) or I did not aware a new deck archetype involved lotsa campfire stories.
        That's because players like you don't need to slug it out in the 80-200 ratings zone and miss out all those eye-openers such as Campfire Stories.
      1. Rayzie's Avatar
        Rayzie -
        So you said that no hero has a win rate of >55%, but do any heroes (not naming names) have a <45% win rate?
      1. Flycheung's Avatar
        Flycheung -
        Quote Originally Posted by Index View Post
        Campfire Stories is included in starter deck. A lot of starters are running it.
        Quote Originally Posted by Direwulven View Post
        That's because players like you don't need to slug it out in the 80-200 ratings zone and miss out all those eye-openers such as Campfire Stories.
        I stand corrected.
      1. pyrogene's Avatar
        pyrogene -
        Hmm... so no differentiation when looking at stats for different ranking sub-groups? I do believe there would be a significant difference in these stats if you look at say, 300+ only games. And it might be more meaningful to look at game balance for more "competitive" players rather than the entire player group.
      1. Nataku's Avatar
        Nataku -
        Quote Originally Posted by Index View Post
        Interesting to read, and I remember you said won't reveal these status to us...
        kyle had already said most of the stats shown there

        but they aint going to show the full stats for each hero
      1. Index's Avatar
        Index -
        Quote Originally Posted by Nataku View Post
        kyle had already said most of the stats shown there

        but they aint going to show the full stats for each hero
        I see.
      1. Raphael Majere's Avatar
        Raphael Majere -
        Quote Originally Posted by pyrogene View Post
        And it might be more meaningful to look at game balance for more "competitive" players rather than the entire player group.
        I would say game balance for the 'largest' group. Suppose ratings from 0 to 400 are divided into groups of 80 pts. That would make 5 groups. If it was up to me, I would look at game balance for the largest group, and that could be from 160 to 240 or 240 to 320.

        Or even 200 to 280.

        This may or may not change over time, depending on how often the players from 250 onwards play.

        If the 250+ players play very often (getting slower these days, honestly), the lower rated players get exposed to 'good plays and good card usage' and hence certain cards/deck archetypes can grow in popularity and that would affect how often a certain card gets purchased and used in-game.
      1. pyrogene's Avatar
        pyrogene -
        Quote Originally Posted by Raphael Majere View Post
        I would say game balance for the 'largest' group. Suppose ratings from 0 to 400 are divided into groups of 80 pts. That would make 5 groups. If it was up to me, I would look at game balance for the largest group, and that could be from 160 to 240 or 240 to 320.
        I don't think that's right. For the sake of argument, let's say there is a card that is obviously overpowered in competitive play but for some reason, underplayed at lower ratings (maybe it's expensive, on a hard to play hero etc.). Evidently, that's not ideal for game balance even if stat-wise there is no anomaly for the entire player base. Game balance should always be assuming a certain level of competence (and rating as a proxy for that).

        Another way stats overall can be biased. Let's say there is a Hero A that is considered underpowered and difficult to play. As such, he is very rarely seen at 300+ rating. However, because he is considered harder to play, typically more experienced and better skilled players will use him. It therefore results in a case that with 2 players of identical skill rating, the one playing hero A will have a lower rating. However, his skill level is still higher than those in that lower rating (who use the more conventional heroes). Therefore, because he is a better player, he still can maintain a 50%+ win rate with that hero.

        Now overall, the stats shows that hero A has a 50%+ win rate. However, individually, hero A is still weak. Filtering by ranking would isolate for player skill to look at actual game balance.
      1. Alejandro's Avatar
        Alejandro -
        I think that balance if not divide in groups can not be correct. The designer Team must see which decks are playing in differents ratings if there arent for instance ant deck nish in 300+ Rating But there are 1000000 in 150-220 Rating its bad information says that nish is balance because there are 49% players With nish. When be able to that some people start With that deck But its imposible up to hight rank and leave this hero.

        Its my opinion sorry my english is not good But i have estadistic university studios.
      1. Drakkon's Avatar
        Drakkon -
        Quote Originally Posted by Rayzie View Post
        So you said that no hero has a win rate of >55%, but do any heroes (not naming names) have a <45% win rate?
        This was my question
      Untitled Document