I'm really feeling it!
Note: I’m not trying to cry havoc here. This is just a cool image.

Well, it’s happening. After conquering a slew of games both traditional and digital, the nascent forms of our future robot overlords are now, via Google’s DeepMind, pointing their brains at what will surely be their next conquest: fantasy card games Magic: The Gathering and Hearthstone. We have, however, been promised (by human underlings, so take this with a grain of salt) that there are currently no plans to build a program that can play these games - only one that can learn how the game works. But as I’m sure many Magic or Hearthstone players will agree, it’s not a robot player that we should be afraid of. If we’re going to be spending the rest of our lives in servitude to an all-knowing Magic or Hearthstone robot, it won’t be the one that plays the game - it’ll be the one that designs the decks.

I spoke to popular Magic: The Gathering deck-brewer and self-proclaimed wizard, Travis Woo, about the possible influences of artificial intelligence on the landscape of the game. Check out the video below for our conversation, or read on for my own thoughts.

For me, deck-building has always been the most enjoyable part of any trading card game. The process of discovery and experimentation is incredibly fun, and winning a match with a deck of my own financially-restrained creation fills me with a sense of fulfillment and satisfaction that can otherwise only be achieved with the help of a foot-long Sub. While my creations are never strong enough to propel me to the higher Hearthstone ranks or net me an undefeated night at a ‘Friday Night Magic’ event, expressing myself through an original deck is far more preferable than winning a bunch of matches with one of the flavor-of-the-month decks that everybody else is playing (not that there’s anything wrong with this approach, you bunch of no-good net-decking pieces of [etc, etc]). But in addition to the fleeting (and I mean fleeting) moments of fulfillment brought on by those rare victories is another, equally-important part, of what makes deck-building an engaging and worthwhile activity: the idea that maybe - just maybe - I’ll be the guy to craft that next big deck.


Let’s take a look at Magic (sorry Hearthstoners, but there’s no school like the old school), and one of its cards in particular: Descendants’ Path. It’s my favorite Magic card for a number of reasons: it’s green (green is the best color and don’t let anybody tell you otherwise), the art is bananas good, and it features a highly unique and (situationally) powerful effect such that it’s a fairly neat-o card to build a deck around. But according to every human method of card evaluation, Descendants’ Path is not a good card. It doesn’t actually do anything on the turn that it’s played, and the effect relies way too much on chance to be worthwhile. In certain competitive environments, investing the mana to play the card, only to have the effect turn up nothing on the following turn, can spell doom.


There are fourteen-thousand unique Magic cards. And human card evaluation has, historically, not always been on point (there was plenty of debate regarding the quality of the Avacyn Restored expansion’s fancy new ‘Miracle’ cards, none of which ceased until a deck based entirely around them won the most prestigious tournament in the professional scene). It would be fairly difficult to argue, definitively, that no existing possible combination of Magic cards involving Descendants’ Path could form part of a deck that could at least hold its own against the more competitively-tuned stacks of cardboard out there. Again, I’m not the most mathematically-gifted human being in the world, so I understand that a guy like me likely isn’t going to be the guy who makes that deck (if it even exists in the first place). But the knowledge that there might be a deck out there that nobody else has thought of is absolutely enough to sustain me. Call it naive hope, call it stupidity, call it whatever you want - knowing that I might be the guy to stumble across that one awesome Descendants’ Path deck is enough to sustain my efforts. I’m a monkey, and Magic is my typewriter - you can’t tell me that I’ll never write Macbeth. A computer, on the other hand, might do just that. And that’s not something I ever want to see happen.


Let’s say that DeepMind eventually crafts a program that understands how Magic or Hearthstone works - what exactly could that program be used to achieve? (We’re ruling out playing the game based on the assumption that DeepMind aren’t going to change their minds, though there is some talk on the feasible effectiveness of a Magic-playing computer in the video.) Well, assuming it has a complete understanding of all the cards, surely it would be able to discern the relationships between cards and thus evaluate the strength, or efficiency, of groups of them. It might be able to, for example, understand that Hearthstone’s Mechwarper works best with mech cards (as opposed to non-mech cards), or that Magic’s Norin the Wary works great with cards that do cool stuff whenever he enters the battlefield. What if the program was able to then work out the most efficient combinations from within particular subsets of cards, and then use these combinations to generate basic deck engines? If I let it do its thing with Descendants’ Path only to have it tell me that even the most efficient engine possible was still under-par, how would that make me feel about my own deck-building process? I’m sure I’d feel at least a little bit dejected.


Or what about a program that could simulate games? Using two user-determined decks, is it feasible that a program could run a set number of games, with the purpose of calculating each deck’s average winning percentage (assuming both players are playing perfectly)? Both Hearthstone and Magic pros put in hours upon hours of play-testing time - if a program could take over the human element of this process - thus speeding it up significantly - imagine the effect it would have on tournament-level play. Similarly, imagine the effect it would have on my fragile ego. I’d rather have my slow, janky creations beaten in real-time by all the net-deckers than be told that I don’t stand a chance against any of them. To put it another way, do you think all the folks who play the lottery would keep playing if they were told they were never going to win? Of course not - that kind of knowledge (however apt) would suck the fun out of the game entirely.


I’m no computer scientist, so maybe everything I’ve described in the above is impossible. But look: everybody who ever said that computers can’t compete with us flesh-bags has turned out to be wrong. The folks who said that IBM’s DeepBlue couldn’t beat Kasparov were proven wrong after the computer won the first game between the two, and later the second six-game match (heck, even the fact that it put up a fight was something that surprised many). Elon Musk said that it was originally predicted that a computer with the power to defeat a top Go player likely wouldn’t appear until around 2026 - meaning AlphaGo got there a decade earlier. In the view of this author, it’s feels far more appropriate to be optimistic about the power of future game-playing computer programs than it does to be pessimistic - because the pessimists have historically been in the wrong.

Deck-building in Magic and Hearthstone is a process of discovery that is, for some, just as enjoyable as playing the game itself. The idea that maybe I will be the person to craft that tournament-winning deck is something that sustains my interest in the game - sure, it might be a naive way to think (especially since I’m no genius), but it’s a valid goal nonetheless. While we don’t know for sure whether or not the DeepMind analysis will eventually result in a program that can effectively eliminate the capacity for this kind of naive hope, I pen this hyperbolic piece as a half warning cry, half call for cessation. DeepMind, there are other games to conquer. The inner child in me is not yet done with Magic or Hearthstone. And I’m sure yours isn’t either.


I’m Scott. Twitter.

Share This Story

Get our newsletter