Who Beat Whom in Modern at MCII?

The coverage of this year’s second Mythic Championship—MCII for short, though not to be confused with 1,102—featured an unprecedented amount of data analysis. This article has a little more.

The Metagame, the Mulligan, and the Information Policy

In my article last week, I looked at the latest Modern trends in the run-up to the big show. One notable insight was that Humans had become more and more popular over the course of the last few GPs. In another shocking development, Izzet Phoenix may have passed its peak. The deck once reached an unheard-of metagame share of 13.4%, but recently didn’t make it past 9.1% at either GP Calgary or GP São Paulo. The latter event also marked the first time Esper Control showed up as a relevant metagame factor.

All of these trends continued in London. The MCII field exhibited much less diversity than what you typically find at a Grand Prix, or rather, it was much more focused. Consequently all the major decks claimed larger shares. But even when we account for that, Humans still took a surprisingly large piece of the cake.

An image of the most played archetypes at MCII.

The same is true for Esper. Esper Control had been the 16th most popular deck choice with players in São Paulo, but it was the 12th pick among Mythic Championship competitors. In particular, Esper completely eclipsed Jeskai.

Izzet Phoenix, in contrast, played less of a role in London than many had expected. Calls for the banning of Faithless Looting may have been premature. Sure, Surgical Extraction weighed in with the highest number of copies across all cards players registered, but it barely cracked the Top 50 of main-deck cards. I don’t think 199 main-deck copies of Surgical Extraction spread over 515 MC decks constitute proof of a problem.

The most popular deck at MCII instead turned out to be Tron. The new mulligan rule must have had something to do with it, although just in so far as tipping some scales. Tron had already been the third most represented deck at GP Los Angeles, Tampa Bay, Calgary, and São Paulo, as well as number four in Bilbao. And that’s despite GPs being notoriously overrun by Burn. Evidence also suggests that Tron wasn’t even such a great choice for the tournament. The archetype’s win rate remained far below expectations.

Ahead of the event, there was speculation that the London Mulligan would inordinately benefit strategies in need of specific cards at the expense of strategies that need a certain number of cards. The event mostly allayed those fears, as the mulligan seemed to have very limited impact on deck selection all around. It clearly didn’t hurt the Humans deck, which requires a critical mass of, well, Humans, or White-Blue Control, which is rather heavy on resource requirements too. Wobbly combo decks such as the one going for Goryo’s Vengeance into Narset, Enlightened Master into Brilliant Ultimatum were nowhere to be found.

In the end, it might have been more relevant that players got to see their opponents’ deck lists ahead of matches. The benefit to Meddling Mage may be minimal, because Mythic Championship competitors should be able to identify decks quickly and know what to name. But it’s a huge deal for control players to be aware during mulligans whether creature removal will be at a premium or mostly useless in the game to come.

It would have been interesting to see one rules change in effect without the other, because one possibly counteracted the other. But I suspect the new mulligan to be a net positive and I love knowing what deck everyone at the Mythic Championship played, so I appreciate the decision to move ahead and implement both.

Conversion and Win-Rates

Information on what decks advanced to the second day doesn’t allow us to draw big conclusions. The Day 2 metagame breakdown says so itself. With three of eight rounds played in Limited, a strong Draft record can completely obscure the strength of an archetype’s performance in Modern. This even applies to the Top 8.

Much more telling are the actual win-rates exclusively measured across the Modern rounds. Here we find Ad Nauseam adherents on top followed by Hardened Scales supporters. If you wonder why none of them made it to the playoffs: Well, the Scales crowd won 67 of their 132 Draft matches and the Ad Nauseam folks only won 18 of 42.

Humans and Jund also won more than the average share of matches in Modern, although the latter did so at a largely irrelevant sample size. In total, there were five Jund players in the running, compared to 53 humans on Humans. That’s part of the reason why the Top 8 ended up with three Humans decks.

Most archetypes’ win rates landed somewhere in the 46–54% grey area. Among the reasonably popular decks, Grixis Shadow and Amulet Titan narrowly missed at 45% and 44.8%, while Bogles avoided “average” by a wide berth and clocked in at 38.5%.

Including playoffs and mirrors, Tron players completed 530 matches, by far the most of any one archetype in the tournament. Despite the massive sample, its low win-rate of 47.9% wasn’t statistically significant by common standards. A deck with a fifty-fifty chance of winning each match has an 18% chance to go 254-276 or worse, and 254-276 happens to be how Tron went.

Not statistically significant either, but uncharacteristically strong was Burn’s performance. Burn had posted a negative record at six of the last eight Modern GPs. Sometimes it even set a record for worst record. In London, on the other hand, it went 57-51. What I wrote about Burn remains true nonetheless: evidently more people play the deck at GPs than should play the deck at GPs. The reason, however, may be that it really isn’t a good pick for less experienced players.

Clean Cuts and Murky Matchups

So Tron beat everything else in sheer numbers. Ad Nauseam and Scales outdid everything else when it came to winning. But who actually beat whom in heads-up combat?

I ran the numbers. What I found is that, compared to a Grand Prix, ten rounds of Constructed play and a smaller field create sample sizes a bit on the low side. Not only that. A few matchups that looked very lopsided at the GP level received more nuance at the mythic level. The lack of clean-cut affairs is of course valuable insight in itself. When players pilot their decks expertly, almost no pairing seems unwinnable. Modern truly remains a wellspring of viable archetypes.

The following combines TitanShift and Titan Breach as well as The Rock and Jund, doesn’t differentiate between the various shades of Eldrazi or Shadow, and spreads a big tent over all control decks based in white and blue. The resulting “W/U/(x)” then shoots into third place as far as metagame share is concerned. Here is how the players of each of the major archetypes performed against the players of the others:

A grid of matchups versus matchups.

One matchup had a very clear favorite, and that was Ad Nauseam in the battle against Tron—the best performing deck overall versus the most played deck overall. If the matchup were even, we’d see a result this extreme or more extreme with a probability of only 1.95%. As such, we can be virtually certain that Ad Nauseam is at a sizable advantage here. Ad Nauseam had also beaten Tron 9-2 the first time it showed up in sufficient numbers at Grand Prix São Paulo, another statistically significant result.

Another deck with a statistically significant winning streak versus Tron at the Mythic Championship was Infect. One player doesn’t do much until they play Karn Liberated on turn 3, while the other will every so often simply kill on turn 3. Doesn’t look like a fair fight to me.

That midrange decks based in black and green did just as well against Eldrazi is a little more surprising—not as in, “I can’t believe it,” but as in, “I didn’t know that.” Previous results from the GP level had pegged The Rock and Jund as no more than slight favorites.

And that’s it. In theory we should stop here, because no other matchup reached statistical significance. A 5-1 or even 8-2 record isn’t significant in itself, although we can look at previously gathered data to try and corroborate some findings. Specifically the results from São Paulo and Bilbao are of note here, because for these two GPs we have complete information.

Humans versus Burn makes for the perfect example. Humans did have a significantly positive experience with Burn at both GPs, with scores of 26-11 and 33-9. The 6-2 performance at MCII matched the corresponding win percentages perfectly—75% compared to 70.3% and 78.6%. It is very likely that Humans would have achieved a significant victory over Burn in London too, if only the deck had faced Burn another five times.

Something similar is going on with Izzet Phoenix versus Hardened Scales. Izzet had won 70.5% of matches against Scales in Bilbao and later 66.7% of matches against Scales in São Paulo, both at statistically significant sample sizes of 78 and 42 matches. Then, in London, Izzet Phoenix’s win rate in the matchup came in at 59.1%. There may be a trend here: either the Scales players have adapted or, more likely, the frequent inclusion of Surgical Extraction over Gut Shot on the Izzet side changed the math. Regardless of whether the improvement constitutes a trend or just fluctuation, the matchup probably remains bad for Scales.

Ad Nauseam versus Hardened Scales is another opportunity for past results to corroborate MCII findings. Ad Nauseam’s 4-0 record versus Scales in London wasn’t significant, but its earlier 13-3 run against Scales in Bilbao had been highly significant (p=0.011).

Finally, White-Blue/Esper/Jeskai Control versus Whir came closest to the goal post in London without scoring a goal. Going 8-2 is impressive, but science maintains it’s not significant. Going 10-3, on the other hand, is significant, and that’s exactly how White-Blue Control had performed against Whir at GP São Paulo.

Final Words

The Mythic Championship results didn’t provide a lot of useful information on their own, but they added some valuable data to our aggregate. The good doctor Frank once tried to explain to me how to use “a Beta distribution as a Bayesian prior”—and failed. (See what I did there? The failure is all his.) Though I suspect I actually did something vaguely related above. Possibly wrong.

Tell me in the comments and join me next week when it’s time for Legacy!

Scroll to Top