25 comments

[ 4.2 ms ] story [ 29.7 ms ] thread
This has been done a lot in the card game Dominion (http://stackoverflow.com/questions/10867765/genetic-algorith...). Dominion has orders of magnitudes less cards than Magic the Gathering.

I've always wondered how a machine would solve for these complex games... How do you build AI to play a combination and reaction reliant game?

Brainstorm is no surprise. That card is the reason so many people enjoy the format still.
Also, this type of ranking can be somewhat misleading. Ad Nauseum doesn't appear anywhere on the list, however all of the cards that support it do (Infernal Tutor, Lion's Eye Diamond, Lotus Petal).

This is because any deck playing the card typically only plays one copy, because it is bad to play in multiples.

That's because the cards that support it are the actually powerful cards! Without them, you can't pull it off. Without Ad Nauseum, there is doubtless a slightly less powerful alternative to be found.

For example, brain geyser was an easy finisher in a deck built around generating stupid amounts of mana, and it also had secondary utility. But there are many cards that can do the same job almost as well.

You'll notice the trend if you peruse lists of restricted cards. It's always the catalyst that is restricted.

Yeah, there have been countless combo decks built around pretty much the exact same set of support cards. Individually the support cards usually don't appear to be a key part of any given deck, but to a large degree that's because no one would ever call a deck a "brainstorm deck" even if brainstorm was the central card the deck was built around. They'd name it after the other things that go well with brainstorm, because that's the part that's actually different.
The point is that this ranking factors in the number of copies of a card played, but that use as a metric can be misleading.

Would Vendilion Clique rank higher on the list if it weren't legendary? Probably. It's not that it's a less powerful card than others higher up the list, it's that you get diminishing returns for playing more than one.

The card text on Ad Nauseum prohibits playing more than a few copies in order to make it powerful. If the card worked a different way, such as if you paid 1 life for 1 card, you would certainly be playing four of them. Would it rank higher than the catalysts? No. But that's not the point.

And following suit, Jace the Mind Sculptor.
Wasteland is an interesting result. I've watched it come and go as the metagame changes, and I also can't help but wonder is it the most powerful land? Or truly the second most powerful card.
I thought the article was mostly about "most used" instead of "most powerful", but I admit not having taken the time to understand it well enough.

If it's just about "most used", then it's probably because Wastelands fits almost everywhere and it's good disruption.

It's interesting that the fetch lands are rated higher than the dual lands. With Deathrite / Brainstorm / Tarmogoyf etc are they actually better?
I think a lot of it has to do with the fact that they can retrieve dual lands. (disclaimer : I'm not an expert)

You could, for example, run a two-color deck with 4 dual lands and 12 fetchlands. You retrieve the dual lands first (if you don't care about Wastelands) and then you get basic lands.

Does that mean that fetch lands are 3 times as popular ? Yeah, kinda, but if you had a choice between using only dual lands or only fetch lands, maybe the decision would favor dual lands in many contexts.

Besides, fetch lands allow you to dry your deck. After you use a fetchland you are less likely to draw a land card.
I believe they are more popular since it offers more flexible.

You can get dual lands, effect lands like wasteland, and regular lands instead of being stuck with any given one of the listed.

You also get the benefit of lowering the odds of drawing a land since you've cleared another one out of your deck.

Finally you get a chance to shuffle (which might matter in some decks).

Fetchlands are so good because Brainstorm is so good. If you Brainstorm without a fetchland, you put your two worst cards on top of your library, meaning you're doomed to see those cards the next two times you draw. With a fetchland, you're no longer "Brainstorm-locked."

Deathrite and Tarmogoyfs are considerations too, but they're not played nearly to the extent that Brainstorm is.

Deck-thinning is another oft-cited use, but in practice, the extra land out of your deck is unlikely to give you a statistical edge unless you're shuffling and drawing a lot that turn. It's far better to save your fetches to shuffle away Brainstormed cards.

Basically every multicolored legacy deck runs 5-8 fetchlands, some even more. Almost no one runs that many dual lands though; even the greediest decks rarely have 4 of any single land, but every B/U deck is going to have 4 Polluted Delta to get their basics. Additionally, Delta can be played as a 9th fetch in a B/G or U/W deck too, because it still gets islands and duals, but you wont see an Underground Sea in that list.
I guess given that they're looking at deck lists things that were insanely broken aren't going to be represented. Skullclamp is possibly the best card to ever exist and nowhere to be found. Jace at 19, I guess maybe it's not played across lots of formats? Though, given just how truly broken it is I'd expect pretty much any blue deck in any format to have 3 or 4 of them. Very cool stats though, I loved looking through it and going "oh man yea, that was an AWESOME card" :)
This is specifically looking at the legacy format, in which Skullclamp is banned, thats the same reason that Jace is where he is. Many slow blue decks in Legacy want to play Jace, but that is not an overwhelmingly huge set of decks.
Back when I played (15 yrs back?) Black Lotus was the consensus "top" card... I don't even see it on that list, so I'm guessing the game has changed a lot...
There are now many formats where only certain cards are legal. This post is about Legacy, where Black Lotus is banned due to its insane power level. Lotus is really only legal in Vintage, and even then a deck can only have 1
I read a comment from Richard Garfield in the days of "Revised edition" and he said that one of the flaws of Black Lotus was that there was no deck that could not be made better with the addition of Black Lotus.

It's an auto-include. A no-brainer. Everybody wants one. Every deck benefits from having it, and that's why it defeats the spirit of the game and should be left out of future editions.

So, you could say that it's because it's so powerful that it's an auto-include, but then that leads to a discussion about cards that are powerful on their own and what cards are enablers.

That's very true, especially considering the members of the Power Nine. None of these cards win you the game by themselves by dealing damage or milling, but they are just huge enablers for almost any strategy.
As someone who for a time bought MTG cards, in the secondary market, there is real money to be made in apps with analyzing financial data.

The secondary market is like a stock market only without regulations, has high fluctuations and swings. Oh and the God of world, Hasbro can reprint most cards in new sets except for a few on the reserve list at wil. Overnight a card can rise or fall. There are many people trying to predict the next swing and trying to get in on the ground floor.

If you're good at analyzing data, can grasp the game quickly and understand why certain cards are more valuable than others, then there is a great opportunity to "sell shovels to gold diggers".

The main market is http://magic.tcgplayer.com/ for the US. http://www.starcitygames.com/ is the one everyone watches. If they begin to buy up cards and force up prices or slash prices, the market notices. If you have a spare weekend, take a look. Might make a good side project.