Sealed Fate

https://twitter.com/thepchapin/status/476402547045834752

Yesterday, Hall of Famer and recent Pro Tour Champion Patrick Chapin posted the above to Twitter. It spurred some conversation from Hall of Fame hopeful Paul Rietzl, CFB Pantheon member Matt Sperling and up-and-comer Kai Budde along with various other members of the #mtg community about the merits of sealed deck vs constructed. We all know that opening a sealed deck is less skill-intensive than building a coherent constructed deck, right? So the question came up: do Pros (capital P) perform better in the lottery that is sealed deck or are they better at the allegedly more demanding constructed formats?

Paul Rietzl posits that he has a larger edge in sealed vs constructed because there is so much information readily available on constructed but sealed is significantly less explored – especially within the public strategy-site space. There’s a wealth of resources to learn a constructed format, and far less to learn a sealed format. So a pro should be able to exploit this either via a super team who can pound countless sealed games and share the findings or simply via being better at building sealed decks (format agnostic). Patrick posits that he has no edge in sealed because he hates sealed and the variance in it is stinky-poo-poo (my words, though I suspect Patrick is on board with the sentiment).

So which is it? Does the variance in opening bombs outweigh the skill edge a pro has or is that skill and preparation more exploitable because information is less available to the public?

First, I had to define a Pro. For our purposes here I took a simple cut-off of 150 career Pro Points. This is an easy demarcation since it is what makes someone eligible for the Hall of Fame ballot. I’m adding the stipulation that I’ve further reduced the field to only people with 150 career Pro Points who were qualified for PT Journey into Nyx. This is because I had that file handy and didn’t want to really dig through a few million records to find any others that this excludes. This left us with 59 people, which should give us a pretty good idea. Finally, I only went back to the 1997-1998 season, records from the first couple of seasons are a little murky.

Finally, I needed a way to control for quality of opponent. Not all events are created the same. Constructed is played in virtually every type of event, on both day 1 and day 2 and in the top 8 whereas Sealed deck is only ever played on day 1 of a Grand Prix. The average day 1 field at a Grand Prix is weaker than the average day 2 field is (duh) and weaker than the average Pro Tour field (also duh). So I’m only looking at the first 9 rounds of Grand Prix (there’s no easy way to break off day 1 vs day 2 in the data I have, so I’m using 9 rounds as a baseline. I recognize this won’t be exact but we’re just having fun here so it should suffice). Overall, this exercise is examining roughly 19,000 matches. So at the very least I can say we have a reasonable data set from which to draw conclusions.

Here are some numbers:

Season	Constructed	Sealed
1997-1998	69.1%	58.3%
1998-1999	68.8%	59.4%
1999-2000	62.2%	58.3%
2000-2001	58.6%	60.2%
2001-2002	63.5%	64.7%
2002-2003	67.4%	64.5%
2003-2004	62.3%	61.7%
2005	64.0%	61.4%
2006	63.3%	65.3%
2007	64.6%	59.3%
2008	64.0%	63.1%
2009	64.7%	63.4%
2010	61.7%	64.5%
2011	64.7%	62.9%
2012	65.4%	65.9%
2012-2013	62.2%	63.2%
2013-2014	62.4%	62.0%
Grand Total	63.6%	63.0%

 

So there you have it. 0.6% difference, in favor of pro’s performance in constructed. Case closed. Except is it? A lot of those numbers reflect an era of Magic when information wasn’t as freely available. From 1997 – 2009 constructed was consistently favored, sometimes by wide margins (look at 2007!) with little fluctuation (9 of 12 seasons were in favor of constructed). Then things got less consistent.

The first Star City Games Open Series was in January of 2010. The first GP with video coverage[i] was GP Oakland in February of 2010. Streaming coverage, and expanded coverage in general of a wider variety of events, led to a proliferation of information available to everyone and anyone building a deck. And while, yes, sealed GPs got coverage too, it’s not as easy to digest a sealed decklist as it is a constructed one. There aren’t nearly as many articles dedicated to building a sealed deck as there are about what the 74^th and 75^th cards should be in BG Devotion and why. Between the expanded coverage and the continued growth of Magic Online, constructed got easier. You don’t need to be a master deckbuilder to succeed. You just need to take 75 from any number of places on the internet and learn your deck along with everyone else’s. And you can! All of the information is out there. Since the 2010 season, the numbers have changed:

Season	Constructed	Sealed
2010	61.7%	64.5%
2011	64.7%	62.9%
2012	65.4%	65.9%
2012-2013	62.2%	63.2%
2013-2014	62.4%	62.0%
Grand Total	62.9%	63.4%

Since the coverage explosion, Pros have been performing better at sealed than they have in constructed. Prior to the 2010 season there was a 1.7% gap in favor of constructed and since then it’s 0.5% in favor of sealed for an overall change of 2.2%. That’s a pretty big difference. Across an average pro’s season that’s about and extra 2 wins, which can be the difference between top 16 ($5,000 and 15 pro points) and finishing outside of the money at a Pro Tour.

In this light, I have to think my namesake is onto something. It’s very possible that limited GPs are a higher EV play for pros than constructed ones are, assuming they’ve done their homework of course. Giving away their secrets for free to the public, however, may not be advisable:

 

This entire exercise got me curious though, which formats have historically been more generous to the Pros out there? I used the same guidelines as above for record selection and came up with this:

Format	Total
Extended	65.2%
Standard	63.7%
Block Constructed	63.3%
Sealed	63.0%
Legacy	61.8%
Modern	61.4%
Vintage	60.0%
Grand Total	63.3%

This is since 1997. Notice that there are no draft formats in here since we’re restricting this to only GP day 1’s to control for strength of opportunity. If I apply the same ‘post-2010’ filter we get this:

Format	Total
Extended	64.9%
Standard	63.8%
Sealed	63.4%
Modern	61.4%
Legacy	60.9%
Block Constructed	57.6%
Grand Total	63.2%

And this, at least for Standard, seems to move away from the previous findings since Pros do better at Standard than sealed (by a small but relevant margin). Modern, Block and Legacy all greatly support the hypothesis while Extended only had 5 GPs which were all in the 2010 and 2011 seasons so I suspect there’s not much to talk about there on either side of the argument.

Overall these are all interesting and fun numbers but I can’t say that anything has been proven. I can understand where both Patrick and Paul are coming from. Maybe they’re both right.

Player	Sealed v Constructed Delta
Patrick Chapin	-2.0%
Paul Rietzl	+1.3%

 

 

[i] That I can find, I could be wrong