The 3 Mistakes Team CFB Ice Made in Testing

Despite Eric Froehlich playing in his fifth Top 8 of the Pro Tour, Pro Tour Amonkhet didn’t go to well for CFB Ice, or its testing partners from team Face to Face.

Personally, I felt well prepared for Limited, and underprepared for Standard. You might be surprised to hear that, as I went 2-4 in Limited, normally my bread and butter, and 7-3 in Standard with a deck I had close to zero preparation with before the event.

I mentioned last week that I thought testing for the Pro Tour would be different from here on out with the early release of the format on Magic Online. The players of the world are getting better and better at solving new formats quickly, and Pro Tour Amonkhet showcased exactly the point I was trying to make.

Zombies (a fresh archetype highlighting a few cards from Amonkhet like Dread Wanderer, Lord of the Accursed, and Liliana’s Mastery) and Aetherworks Marvel (which featured basically no cards from Amonkhet) were the two stories of the tournament, and the two stock decks CFB Ice worked on the least while preparing for the event—despite myself, EFro, and Paulo playing Temur Marvelworks.

How did that happen? Well, I think we made three major errors as a team.

Error 1: Inconclusive Testing

All of the Magic Online data pointed to a format made up of four decks: Mardu Vehicles, Various Aetherworks Marvel decks, Zombies, and U/R Control.

Early in testing, Zombies was reported as being a fairly low power level deck that had some trouble against sweepers and wasn’t matching up very well against either Aetherworks Marvel or U/R Control decks. The deck was performing well against Mardu, so if we thought Mardu would be a portion of the metagame, maybe it would have been a reasonable choice if we didn’t find anything else. In a four-deck metagame though, we didn’t want a deck that was bad against half of the decks, had an even matchup against itself, and then was good against one other deck.

U/R Control was performing about as well as we expected it to, playing close games with Mardu, beating Zombies, and our first major error came in testing this deck we basically assumed would beat Aetherworks Marvel. When we tested the matchup, it wasn’t for nearly long enough, as the data from the Pro Tour actually had Marvel dominating U/R Control and even U/W Flash in the tournament.

Because of this, we didn’t test the matchup enough to discover that Aetherworks Marvel decks were actually good against control, or at least close. We also didn’t realize that we had been testing an inferior version of Marvel against a stock U/R Control deck, as the Marvel decks changed rapidly while the U/R Control lists didn’t change that much. It’s also possible that the Pro Tour statistics are an anomaly as there isn’t a huge sample size, but I think even though the sample was small we can still draw some good conclusions from the data.

At one point, we skipped right to testing an Elder Deep-Fiend version of Aetherworks Marvel, assuming it would help against control, allowing us to tap down the opponent’s lands so we could resolve a Marvel game 1, but I think we didn’t do a good enough job developing a more normal version beforehand and then testing it against control to see how the matchup felt. This is especially true for sideboarded games.

Next time, we should come to real conclusions about a matchup by playing it a lot instead of making assumptions with so little data to back it up.

Error 2: We Misjudged the Metagame

Earlier this year, at Pro Tour Kaladesh, our team wildly misjudged the metagame by assuming no one would play an all-in version of Aetherworks Marvel, the deck that turned out to be the most popular deck at the tournament. We thought the deck was just not consistent enough and others would come to that same conclusion. This time, we over-anticipated the amount of control decks we thought would be present in the field despite eventually not thinking the deck was good enough.

U/R Control was promising early in testing, and we actually really liked Kefnet the Mindful. Kefnet was a cheap win condition with a built-in card draw engine when the control deck manages to turn the corner, and a resilient threat that closes out games. Kefnet also has the ability to get underneath a Gideon, Ally of Zendikar, making it awkward for a Mardu Vehicles opponent to cast after Kefnet has hit the battlefield because it’s relatively easy to turn into a creature on the control player’s turn 4.

It got to the point in testing that we just removed all the Torrential Gearhulks from our decks because they lined up so poorly against Unlicensed Disintegration and added Kefnet and Pull from Tomorrow, which proved to be a powerful way to go over the top of midrange decks. We eventually moved off of this plan, because Torrential Gearhulk was still too good to leave out against non-Mardu opponents, and because having so few threats made it easier to get unintentional draws.

Most of the best cards from Amonkhet were also good control cards. Cards like Censor, Kefnet the Mindful, Pull from Tomorrow, and even Glorybringer could slide right into U/R Control.

Another mistake was believing that this deck, or similar control decks, would represent a bigger share of the metagame than we anticipated. Ivan Floch and Paulo Vitor Damo Da Rosa are both players with strong biases toward control decks, and after a few days even they weren’t willing to sleeve up U/R Control when we didn’t have too many good options. This alone should have given us an indication that control would be underrepresented at the Pro Tour. In the end, a deck we thought might make up between 10-15% of the field ended up only taking 4.2% of the pie. This was a mistake that steered people in wildly different directions. Some members played Mardu because it had a solid control matchup, other chose to play a B/G Cryptolith Rite deck that had a solid matchup against both control and Mardu while being quite poor against Aetherworks Marvel and Zombies. Misevaluating how good we thought this deck could be early led us to make poor decisions about our own deck choices and others’ deck choices.

In the future we need to do a better job of understanding how the Pro Tour will break down, and I think Magic Online 5-0 deck lists laid it out pretty well for us prior to this event, we just need to trust the information from Magic Online more.

Error 3: We Tried Too Hard to Be Different

In testing, we spent many days trying to develop new decks instead of just working hard to tune existing decks. While we did have a group working solely on Mardu, Marvel didn’t get much attention until the last week of testing. People on the team were hesitant to play with Aetherworks Marvel and embrace the variance a card like that brings to the table.

We worked hard on control decks, a Temur Emerge deck, and the B/G Cryptolith Rite deck that some members of our testing team played in the event. While it’s true that if you play a new archetype people are more prone to make a mistake against you, I still think you want to just play the best deck.

During the last week of testing, I actually spent a lot of time playing a Marvel deck that was similar to the one I played at the Pro Tour, but I played it against brews instead of against the decks we expected to face. In other words, I was playing the enemy instead of testing the deck for use, and it never developed against decks we anticipated playing at the Pro Tour. I did convey that I thought Marvel might be the best deck for the tournament, but I don’t think I voiced a strong enough opinion to sway people from B/G Cryptolith Rite, which soaked up a lot more of our time than it needed to.

In the end, I played no Marvel mirrors, very little Marvel against Mardu, and no Marvel against control or Zombies in testing.

The night before the Pro Tour, EFro and I were really the only ones considering the deck, but without much of an idea of how to build it, and it felt hopeless to build it myself. Paulo sent me a message on Facebook finally showing interest in playing Marvel, so we set to work putting together a list with the very little knowledge we had of the deck against any of the stock decks.

Shuhei Nakamura had a version of the deck built from Magic Online and we just went off of that and made some very minor adjustments.

This is what we landed on:

Temur Marvel

When it came time for the tournament, we all joked about how none of us had any idea of how to sideboard with the deck, which was completely true because none of us has really tried to figure it out in testing against the other stock decks. I had never felt less prepared with my deck, but I was confident we made the right call when it came to which archetype to play, and as it turns out I think we were right.  Despite the deck not winning the Pro Tour, I think it was the best deck in the tournament, but likely not the best version of the deck.

In the future, we need to do a better job of developing every stock deck, so that if it comes down to it, we have a good version of every deck we could consider playing, and we need to spend less time trying to “break it.”

Where Will Standard Go from Here?

The Top 8 at the Pro Tour showed us that currently Standard is a two-deck metagame. Zombies and Aetherworks Marvel were the stories of the weekend. Mardu Vehicles had a horrible weekend, and may get pushed out of the format if nothing changes.

Aetherworks Marvel is having a similar effect to the one Saheeli Rai and Felidar Guardian had on the format in that midrange strategies are just obsolete by comparison. Aetherworks Marvel is like the midrange version of Smuggler’s Copter—every midrange deck is just better off having Aetherworks Marvel with an energy package and Ulamog, the Ceaseless Hunger to spin into because it goes over the top of everything else.

I think Zombies will be a poor choice moving forward, at least this weekend. While it has some resiliency, it relies on flooding the board against Marvel to make casting even an early Ulamog, the Ceaseless Hunger ineffectual. I would expect the numbers of Chandra, Flamecaller and Radiant Flames or Sweltering Suns to go up, both in Marvel decks and other decks like Mardu, making a plan like this far less feasible.

There’s some talk of Aetherworks Marvel needing to be banned, but I don’t think so, at least not yet. I think it’s possible to beat both Zombies and Aetherworks Marvel with a single deck, and we need to wait a while to see what happens before we rush to this conclusion. Aetherworks Marvel is certainly breaking a lot of the rules in Standard right now, so I’m sure they’ll be keeping an eye on it.

I have a choice this weekend of staying home and playing the MOCS or going to GP Montreal, and I haven’t decided which of the two I’ll do yet, all I know is that whichever event I choose, I’ll be activating Aetherworks Marvel, and I suggest you do too.

Scroll to Top