Oh man, reader. After taking the initial plunge, reviewing data, histograms, values, ratios, and more, I’ve gotten much more excited about the prospect of continuing this research project. I’ll do a quick good news bad news, and readjust our goals after explaining a lot of the data gathered for this step.
Bad news: sample size is a serious, serious problem, and this is more complicated than previously thought.
Good news: There are ways to mitigate sample size error (and in doing so, we can also solve some of the further complications that have arisen).
So, before I get into the nitty gritty, let’s talk about some data:
- There are only 5 non-tribal, no modifier minions in the entire game. These 5 are Wisp, Magma Rager, Chillwind Yeti, Boulderfist Ogre, and War Golem.
- Including tribals, there are only 14 no modifier minions in the entire game. (No modifier means no card text.)
With that in mind, in order to have a large enough sample size, I’ve introduced minions with two extra variables: Tribals (or Creature Types), and Taunt. With these two categories, the number of minions we can draw data from rises to 25. This is still a small number, but it’s better than the 14 we would normally be stuck with. Also, there are some precautions I can take to make sure they don’t cross-contaminate the data. Overall, introducing these extra variables is fine, because they are the two easiest to break down (both logically and relative to the data). But before I can explain why that is, I have to explain something entirely different, something I didn’t expect to have to do for another step, and that’s the value of a card. Read More…
I’d like to start by explaining exactly what I’m planning to do with this series.
This is a series of articles (and accompanying videos, given time) focusing on the concept of determining the “mana value” of certain stats and effects on Hearthstone cards, using the mana value as a metric.
When I say metric, I mean using mana in the same way you would measure the height of something in meters; the meter is an established standard, and by measuring something in terms of that standard you can have a uniform method of measuring anything.
This is definitely not about measuring the meta value or potential value of a card; both of those determined mostly by the standards of popular play, which is difficult to be objective about. The questions that kind of study would answer are things like the worth of running an extra BGH to counter handlock, or a Kezan Mystic to counter Echo Mage. Interesting things, yes, but not what I’m after; this series is about trying to be as objective as possible, entirely through statistics or data comparison. Through these things we can arrive at an approximate value of the stats of minions, or certain effects on cards; with the knowledge gained from this article, one would be able to measure whether changing Argent Commander from a 4/3 to a 4/2 was a statistically significant change to the card’s value (and therefore too hard or too light), or be able to create an entirely new card with an appropriately balanced effect for its mana value.
This is a study that has a rather hefty amount of variables, and there are certainly going to be outliers that test my ability to objectively measure things. I wholly expect to be stumped when attempting to break down
- why Magma Rager may have “good stats”, number wise, but is objectively one of the worst cards in the game
- why Injured Blademaster is nearly a Priest must-have, but no one else runs him
- how to judge random effects in any consistent manner
- and the value of card draw.
