Hello! My name’s Henry Druschel, and I’ve been playing Magic for about 15 years. I play primarily online now, but I’m not very good, and I’m not here to offer strategic advice. Instead, I’m hopeful I can offer something that hasn’t really existed yet in the world of Magic coverage and analysis.

One of my other hobbies is watching baseball, and specifically writing detailed, quantitative analyses of it. Baseball lends itself really well to such a pursuit, as there are relatively few individuals responsible for the outcome of a given play, and almost every aspect of the game is tracked closely. In recent years, the understanding of what players are valuable and how they accumulate that value has skyrocketed. As someone who writes about baseball in that way, I’ve always wanted to bring a similar approach to Magic.

It’s a lot easier to think about than do, however. For one, synergy is basically nonexistent in baseball, while it defines Magic. A baseball player is basically as good on one team as he is on any other, but Myr Enforcer is a lot better in an Affinity deck than it is in a deck with no artifacts. That means identifying the best and worst cards is very, very difficult, since so much depends on context.

At a more basic level, though, Magic analysis is difficult because of the relative lack of data. The majority of deck lists from major events are not made publicly available, and when decks are made available, it’s usually 5 to 10 decks that performed very well, which is just not enough for real analysis. It’s possible to combine tournaments from week to week, but that would ignore changes in the metagame. In any case, detailed quantitative analysis of Magic results hasn’t really happened yet.

That was one of several reasons Modern Masters 2015 Weekend was exciting to me. Four massive GPs in the same format running simultaneously would offer an unprecedented amount of data, even if it was of the limited type usually made available. I thought this was the perfect opportunity to try to bring some quantitative techniques to Magic.

On the one hand, my hopes ended up a little too high. I had dreamed of accessing the Sealed pools and deck lists for 40+ players that went 9-0. Unfortunately, only coverage of Chiba had full pools rather than deck lists only, and I was unable to find a good chunk of the undefeated decks, including the 40th card of Klees Vlaanderen’s deck from Utrecht, all of Pascal Maynard’s deck from Vegas 1, and any of the decks from Vegas 2. That said, this still left 31 decks, which is far more than any other single GP weekend has offered in the past, and I would think enough for some real trends to be observed. Let’s find out! What traits do the 9-0 decks share, and what distinguishes them?

Color Distribution

I started simply, by looking at the colors each deck played and how much, and taking the total across all three GPs.

(Click image to enlarge.)

The thing that stands out most is the dominance of green. It was the most played among the 9-0 decks at each individual GP, with 65% (20 of 31) playing at least one green card in their decks. White was the only other color above 20%, with blue and black lagging particularly far behind.

Black made up only 2.2% of the 9-0 decks in Utrecht, with a Creakwood Liege and two Grim Afflictions the only black cards present, all splashed in primarily GW decks. Its total breakdown was salvaged by Chiba, where it was 21.7% of the undefeated decks, but it still seems clear that black and blue have the least to offer for Sealed.

Green’s dominance perhaps isn’t completely surprising, since it is the color of mana-fixing and ramp, and the chatter going into the GP was that 4- or 5-color strategies, often base green and splashing for every bomb or piece of removal opened, would be dominant. However, the results don’t seem to bear that assumption out:

4- or 5-color decks only made up a combined 22.6% of the undefeated decks (7 of 31). Indeed, the fewer colors played the better, as decks with playing two colors without any splashes were best represented, at 45.1% (15 of 31). Having watched coverage, this matches my intuition and what seemed to be the narrative of the day. The commentators mentioned several times that the slow, durdley decks were often getting run over by streamlined and consistent collections of synergistic cards.

My sense is that’s not the normal pattern for Sealed. The commonly heard advice is to not try to build an aggressive deck, or rely on synergies that are stretched too thin across your pool. Without data from previous Sealed formats, it’s impossible to do an objective comparison, and see if this is indeed a lower share of decks playing 3+ colors, but hopefully some work can be done in the future using this information.

Color Pairs and Archetypes

What about color pairs? Which combinations were most present in the 9-0 decks?

Note: this is an unduplicated count (e.g., a GWB deck counts for GW, WB, and BG), so the percentages sum to more than 100%.

Unsurprisingly, based on the above data, green combinations dominate the 9-0 decks, with three of the four most common pairings. GW is the most observed by a large margin, which again, fits the weekend’s coverage. In feature matches and conversations between the announcers, the token strategies seemed very, very powerful. BG is the second most common combination, and that often has token synergies as well, with cards like Dread Drone, Scavenger Drake, and Plagued Rusalka providing a nice complement to the GW strategy.

What’s most interesting to me, however, is the relatively high occurrence of the more synergy-based aggressive color combinations, specifically WU (Affinity) and RW (double strike). Those archetypes seem much better suited to Draft than Sealed, but their associated color combinations still managed to put up a decent showing among the 9-0 decks.

There are two factors that make me hesitant to conclude they’re actually decent Sealed color combinations, however. One—by no means is every RW deck the “double-strike archetype,” but that can’t really be defined, and so while color combinations are a flawed measure, they’re all that’s available. There were 6 Viashino Slaughtermasters in these decks and 11 Faerie Mechanists, so one should be cautious when using color to draw conclusions of what archetypes performed well.

Secondly, with linear archetypes like these, it’s possible there’s a big jump in win percentage once a certain threshold of cards is crossed. It’s conceivable that a UW deck with 20 good affinity cards might win 40% of the time, but a UW deck with 24 wins 65% of the time, since those 4 additional solid cards for the archetype push it to the critical mass required for consistency and explosiveness. (Compare to a hypothetical “normal” non-synergistic deck that wins 50% of the time with 20 good cards and 55% of the time with 24.)

If that’s the case, there might be a relatively high number of UW decks among the pool of undefeated players, but that wouldn’t mean it was actually a good archetype, or one that you should be looking to go into before opening your packs for Sealed. Without decks for players with worse records, or full pools rather than starting 40s only, it’s impossible to say. Those are two reasons, though, I wouldn’t conclude from this analysis alone that synergistic, aggressive decks are actually not bad in this format.

Converted Mana Cost and Curve

A deck’s curve is another way to draw conclusions about speed. What does that data show us? I looked at the percentage of nonland cards in each deck for three categories of converted mana cost: 0-3, 4-5, and 6+. About 69% of nonland cards had a CMC between 0 and 3, or about 15-17 cards in the average deck, compared to 23% from 4 to 5 (5-6 cards) and 8% 6 or higher (1-2 cards). That feels lower than I would expect for a Sealed format, but again, without past data to provide context, it’s pretty much impossible to say for sure.

I also made three categories based on the mix of CMCs in each deck. I called them low-, middle-, and high-curve. You can think of them as aggro/midrange/control, though each individual deck might not fit that label perfectly. Low-curve decks were those where more than 75% of their nonland cards had a CMC between 0 and 3, middle-curve decks were those with more than 60% and less than 75%, and high-curve decks were those with less than 60%.

Of the 9-0 decks, 35.5% were low-curve, and only 22.6% were high-curve. These categories are totally arbitrary, and again, without context, it’s impossible to say how these compare to a “normal” Sealed format. That said, this feels like a high proportion of low-curve decks. A deck with 75% of its cards at 3 CMC or below is definitely somewhat extreme. That translates to between 5 and 6 cards at 4 CMC or above, which is not a lot, particularly for Sealed. This would seem to suggest that some pools have aggressive strategies that can in fact perform very well.

Individual Cards

Finally, I want to look at are the individual cards in the 9-0 decks. What cards seemed to perform particularly well? Without deck lists that didn’t perform as well, this analysis is limited, since there’s no way to be sure that these cards are showing up more frequently in good decks than bad decks, as opposed to showing up a lot in every deck, but it’s a start. First, the most-played commons, uncommons, and rares (no mythics, since there were at most 3 copies of any mythic):

This is absurd. 5 green cards, 3 white cards, a GW card, a GR card, and 5 colorless cards (one of which is basically white, Blinding Souleater). The colorless cards are not surprising—I would think most Cathodions, Mortarpods, and Golems that get opened are played—but the total absence of blue, black, or (mono-)red cards is crazy. Kozilek’s Predator and Overwhelming Stampede run away with their rarity categories, and it seems that the dominance of GW was not overstated.

To allow a comparison between categories, I calculated the expected number of a card of each rarity opened in 31 Sealed pools, and divided the number of each card played by the expected number opened to get a standardized figure. This is not a perfect metric, so rather than examining the top 10 or 15 and stressing over slight differences, I think it’s more interesting to look at the general trends.

Overwhelming Stampede is the top card, again by a large margin, followed by Elesh Norn (the only mythic with 3 copies). The first red, black, or blue card was Savage Twister, at #8. The first mono-black card was Bitterblossom, tied with the first mono-red card Comet Storm, both at #12. The first mono-blue card was Cryptic Command, in a cluster at #43. The earliest non-mythic red, black, and blue cards were Lightning Bolt (#22), Profane Command (#27), and Faerie Mechanist (#61). In case you haven’t got the message yet: play green or white, preferably both, and don’t play blue.

The last thing I examined was the curve of the decks each card was played in. Savage Twister was one of the top performing cards, for example, but it seems like a card that wouldn’t fit into every deck, or at least needs to be built around to perform to its full potential. What cards are the most generically powerful, and what cards are the most skewed toward low- or high-curve decks? I’m again using percentage of cards with CMCs from 0 to 3 to determine curve, but splitting it into two categories, below 67.5% or above 67.5% (or about 15 cheap cards, close to the average for these 31 decks). I’m limiting this to only the 62 cards that appeared 5 or more times, and split the data into three charts.

Hopefully that’s legible, but what’s being charted is the split between low-curve and high-curve decks. As you move from the top to the bottom of each individual chart, and from left to right between charts, you move from cards seen more frequently in high-curve decks to cards seen most frequently in low-curve decks.

It’s nice when data fits intuition, and indeed, Savage Twister has the fourth-highest skew towards high-curve decks, behind a ramp card, a 7-drop, and a 4-drop that’s sometimes an 8-drop. The cards at the extremes of this metric are the ones that might look great in some decks, but shouldn’t be jammed into every deck of those colors. Not every red deck wants Gorehorn Minotaurs, and not every green deck wants Pelakka Wurm.

Cards toward the middle are the ones that might pull you in one direction, but are by no means bad in decks on the other end. Bouncelands, for example, were somewhat skewed toward high-curve decks, likely because having a bounceland or two makes it more likely a deck will have a higher curve, not because they can’t be played in low-curve decks. Burst Lightning is a similar card that’s slightly skewed toward low-curve decks but will never be cut from a higher-curve deck either.

Finally, the cards in the middle are those that were played equally in low- and high-curve decks. Some have traits that make them useful to either side of the spectrum, but for different reasons: Nest Invader, for example, is a 2-drop for aggressive purposes, but also adds a mana for ramp purposes. Other spells are just very flexible, and should be played in almost all decks of those colors. Arrest and Narcolepsy are powerful, all-purpose removal spells, Cathodion is a colorless 3-mana 3/3 with slight upside, and Overwhelming Stampede is bonkers.

Conclusions

That concludes this deep dive into the data from Day 1 of Modern Masters Weekend! As was hopefully clear, there’s a lot that can be gleaned even with the limited information available. That said, without years of results to compare this data to, it’s hard to draw firm conclusions. This format feels different than Khans Sealed, without a doubt, but is it more or less different than Fate Reforged Sealed? It’s impossible to definitively say.

Going forward, I plan to do analyses like this whenever adequate data is available, and hopefully contribute to a foundation for other interested individuals to build off of and supplement. A big step in the right direction would be more complete and comprehensive data from paper tournaments. Obviously, inputting deck lists or entire Sealed pools by hand is miserable, and I’m incredibly grateful toward the coverage workers who do so! That said, all this information is already stored somewhere, on paper registration sheets, and I have to imagine it might be possible in the near future to digitize the information and make it public at the end of the event. I can dream, at the very least.

I hope this was enjoyable! If you have any questions, feel free to get in touch with me on Twitter at @henrydruschel. While I Tweet almost exclusively about baseball, I’d be glad to talk about Magic!

Methodological Notes

I included only 9-0 deck lists.

• Color ratios were calculating by counting each mana symbol in the costs of cards in the deck.
  • Hybrid counted as half, so Dimir Guildmage contributed 1 to blue’s count and 1 to black’s count, Spectral Procession contributed 1.5 to white’s count, etc.
    • I made an exception if a deck was completely incapable of casting a hybrid card using one of the colors. This came up twice: Wataru Okada played Creakwood Liege in a purely WG deck, and Jasper Boelens played Selesnya Guildmage in a purely RW deck. In those cases, I counted the cards as mono-black and mono-white, respectively. Decks with any mana of the other color were not adjusted. Takehiro Fujimoto, for example, played Selesnya Guildmage in a primarily BG deck. He played no Plains, but did play a Selesnya Sanctuary, so the Guildmage was classified as a WG card.
  • Phyrexian mana was not counted, and cards castable with colorless and Phyrexian mana only were considered colorless. They also had their converted mana costs adjusted. e.g., Dismember is classified as a colorless card with a CMC of 1.
• For a deck to count as an example of a color pair, it had to have at least 15% in both colors. This translates to about 4 or 5 colored mana symbols, in an effort to exclude splashes.
• Alternately, there was no minimum when calculating the number of decks playing 5 colors, 4 colors, etc. A WR deck splashing one card from blue, black, and green would be considered 5-color, as would a deck playing 5 cards of each color.
• Affinity cards had their converted mana costs halved for the curve calculations.