3/01/2010 Feature: "Pro Tour–San Diego Metagame Breakdown"

4 posts / 0 new
Last post
This thread is for discussion of this week's Feature Article, which goes live Monday morning on magicthegathering.com.
First of all, thank you so much for these articles.  I eagerly await their arrival all the time and I imagine there's a lot of effort put in to producing and resenting the data.  So, when I bring about my nitpicks they are only in the spirit of wanting to improve future articles and mostly to create discussion.

There are two majors things I want to point out here.  The first is that you overlooked another reason why Jund (and other Most Popular Decks of the past and future) struggle to keep their heads above 50%.  The reason is that everyone knows it's going to be the most popular deck, and most of the deck choices people bring will have it in mind to make sure they have a chance against it.  People will maindeck hate for the deck and definitely have a preplanned and tested sideboard strategy for it.

The other thing I want to bring about is how those White Weenie and Junk numbers are not surprising at all.  Math ahead!

Let's suppose that there was no sideboard at all.  You play the deck you play for all two or three games.  We will also assume that each deck has a pure probability p of beating each other deck (and a probability 1-p to lose to it).  Now, the ways that I can win a match are to go 2 straight, win-lose-win or lose-win-win.  The odds of getting the first are p^2, and the odds of getting either of the other two are (1-p)*p^2 (order doesn't matter here).  Summing all of these odds gives a total chance to win a match of 2p^2-2p^3.

Okay, let's take WW's game win percentage of p=0.5533 and plug it in to see what match win% we get: 0.5796 or 57.96% which is pretty close to the recorded 58.04%.  In fact, it's well within +- one match (0.39%).  Now let's see what the expected match win% is for Junk with p=0.5226: looks like we get 0.5339 or 53.39% which again is really close to the recorded 53.23% and again is within +- one match (0.81%).

Now, the assumption that all the games played are with the exact same deck is definitely not true, but by my account Isay that there is not enough data to support your claim that sideboard was the cause of the difference in match results between those two decks.

In fact, looking at all the decks with >100 matches, the ones that have wonky match win% compared to their game win% are the bottom four, most especially Bant and Vampires.  In fact, Bant has the incredibly bizarre result of having a losing game win% but a winning match win%!
Always a great resource, though I really liked the breakdown in sub-archetypes that you did for Austin.
There is some mistake in the presented data.  It doesn' make sense that Boss Naya has 2 matches against Blue-White Control but 7 games.

[math analysis snipped]

Summing all of these odds gives a total chance to win a match of 2p^2-2p^3.


Great insights.  But there is a typo in the formula above.  It should be 3p^2-2p^3.

Sign In to post comments