Deck analytics for Hearthstone
This site is still a work in progress; I'm still pulling all of the threads together. Consider the content on this page aspirational while I work through the kinks. Note that at this moment I am not planning on offering deckbuilding - the intent is to send you to hearthpwn or other socially-connected sites to actually build a deck.
- Cards are their basic dump of information, but we don't yet show card labels, nor do we show effective attributes alongside the card's.
- Hearthpwn URLs work to pull in decks, but I don't yet run the analysis backend on them.
- We don't yet have a complete index of the major deck sites' ratings and scoring.
Improve your decks... with science!
Small changes to deck composition can have dramatic effects on how your deck plays. The process of analyzing a deck is a complex and non-intuitive mathematical problem, and the rules of the game and the cards you insert change the probabilities dramatically.
Add to this that many cards actually break the rules of the game - exchanging one set of resources, like mana, for another - like cards from the deck. These cards change the speed of a deck, and alter the probability of seeing other cards.
Most people's grasp of basic statistics is bad enough; but the distributions that govern how decks of cards work - the probability of certain occurrences when extracting elements without replacement, and the hypergeometric distribution.
We perform complex distributional analysis of deck composition, using these and other statistical techniques to help you build a better deck.
The quality loop
We learn by doing; improving is a continuous feedback loop that involves building a deck, playing it, learning what worked and didn't, and applying that knowledge when you build the next one.
The problem, of course, is that - especially in the early stages - it can be very hard to know what a good deck looks like. Novice players have a habit of building a deck containing all of their favorite things, leading to a deck that contains a series of individually interesting choices but with no balance, no cohesion, no cooperation, and ultimately no wins.
We provide tools that help you learn faster, showing you how the cards in your deck interact. We analyze many attributes of the deck - the speed, the playability in both early and endgame, and help you make sure that, at every moment of the game, you're getting the cards you need when you need them.
The wisdom of crowds
While community reviews and reviewers at large can be a great source of information on what works, and what doesn't, it can take a long time - and a lot of luck - to get enough people to try your deck to get that feedback.
And, as we know, both quality and learning depend on feedback loops.
We've trained a machine learning model off of the reviews of hundreds of decks by hundreds of reviewers to develop a rating - a kind of "virtual wisdom of crowds" - in a score we call the deckacritic score.
A Virtual Value Town
While there are lots of players out there that stream, few of them are as focused or as good at teaching as TrumpSC - and he does a great job of documenting his thinking... and his decks.
We've trained a machine learning model off of TrumpSC's constructed and arena deck selections, to help you make better deck-building decisions.
If you find it useful, donate Trump some money for cookies.
(There are other virtual players, too, but he's my personal favorite.)