This is the iteration of the Data Generated Vintage Cube that was featured in the main event at CubeCon 2023. For updated iterations, see here
As a Magic The Gathering fan and a cube enthusiast, I understand the effort and time required to hand curate a paper cube. It's a labor of love to carefully select cards that work well together, fit the cube's theme, and balance the power level. For me however, as life's responsibilities started to pile up, I found it increasingly challenging to keep up with the task of evaluating the ever-growing number of new Magic The Gathering cards. I didn't want my cube to become outdated, but keeping up with new releases was becoming a daunting task.
That's when I decided to apply my data-driven approach to this problem. I wanted to create an algorithm that could evaluate and filter a vast pool of cards and provide me with a curated list that fits my cube's specific requirements. I was also excited about the prospect of using a method that would allow me to discover new cards and combinations that I hadn't considered before (As an example, I no longer follow spoillers and hadn't heard of the card Atraxa, Grand Unifier before being recommended to the cube)
Although I initially wanted to save time by automating the cube-building process; these days, I find myself spending just as much time, if not more, tweaking and refining the algorithm and data collection methodology. However I don't mind the additional time investment as it allows me to explore my passion for coding interesting solutions to a novel problem, and the excitement of discovering new card combinations keeps me motivated.
Overall, building the algorithm and cube has been an enjoyable and rewarding experience. It's not only allowed me to create a cube that stays current with new releases, but it's also given me an opportunity to apply my data-driven approach to a hobby that I'm passionate about.
The DataFundamentally, data is required in order to find common patterns among cards used , and in order to generate a relevant cube. There were many data driven approaches that could have been taken; For example, earlier I had considered analyzing cards via vector similarity. For this approach to generating a cube that is representative of other cubes of my desired target (in the powered vintage category) I decided to go with a simpler frequency analysis approach. This approach utilizes cube data from Cube Cobra (thank you fellow cubers) as the input source to the algorithm. Currently, to get this data I manually pull each cube's card information down as a csv and save to a directrory. In the future if Cube Cobra implements a /get cube endpoint, that could greatly speed up this process. Alternatively, I could write a scraper to manually pull down the data as well; but for now, manually pulling the data down will do fine, it's just a bit more labor intensive.
After deciding that I would use Cube Cobra data, I had to decide how I would select the cubes whose data I would utilize. This process has changed over the years. Initially I would run a Cube Cobra search with the folowing criteria.
Pre 5.9 update, this search returned quite relevant results and I would often take the first ~70 or so cubes (always excluding this cube) and use that as the input. There was some light discretionary filtering so as not to skew results too much. so for example, I excluded old-school or time-era specific cubes, cubes missing one or more colors , or cubes themed around one archetype or block/set. The beauty here is that as each quarter goes by, the results pulled back would change a little bit as new cubes were created or old cubes updated and in this way the input data to the algorithm was a good representation of how the cube community as a whole was building powered vintage cubes.
A note, on the search here. Post 5.9 Cube Cobra, the search functionality has been re-worked. To the best of my knowledge, we cannot currently submit searches that specify criteria such as the cube's category or number of cards. Fortunately we can still search cubes containing specific cards. Additionally, the methodology for how Cube Cobra ranks popular cubes seems to have been re-worked such that popularity is orderd by number of followers descending. I have been using this search link to get cubes, which I must additionally do discretionary filtering manually based on number of cards.
One of the issues of pulling data in this way is that there are some very "popular" cubes that have not been updated in years. Including these sorts of data points cause issues when one of the goals is to keep the cube updated as new sets are released. Additionally, there are some great cubes that are close to my cube's target size of 360 but not exactly that many cards. Including these cubes seems like a good opportunity. So I've now added two new additional criteria.
I still have challenges finding a good representative sample of cubes that I'm happy with. Interestingly taking the top 70 or more cubes, makes a really good overall representation of what the average powered vintage cube looks like, but simultaneously the frequency sampling method that I will describe in the next section can discount newer cards, which again finding these is one of the goals of building this algorithm in the first place. Through experimentation I've found that using a small sample of cubes that are widely followed and frequently updated provides me with recommendations on new exciting new cards. However, this is no panacea as using this small sample can also allow some quirky non-category representative cards to make the recommendation cut if the sample has many copies of the card. I'll detail how I currently overcome this in the algorithm description.
As an aside, the algorithm creates a representative cube based on whatever set of input cubes you send to it; so a Legacy Cube, a Peasant Cube or whatever flavor of cube you enjoy most are all fair game.
The AlgorithmBit of a hack job that...
The source code for the algorithm is not publically available; If you are interested in the code you can reach out to me.
I use the following color categories in the algorithm description and analysis that follows
Color Category | Symbol |
---|---|
White | ![]() |
Blue | ![]() |
Black | ![]() |
Red | ![]() |
Green | ![]() |
Multi-Colored | ![]() |
Colorless | ![]() |
Land | ![]() |
The algorithm does the following to generate a cube.
Load all the data gathered in the previous section in a DataFrame object.
For each color category in the table above do the following sub setps.
Here is a partial example output for blue at this stage
Concatenate all cube color frames into a single object.
**I don't have time to write more details on this step, so feel free to ask me about this at cubecon.
Cube BlacklistWhen building a cube, you have the option to provide a blacklist to the algorithm. If you choose to use this feature, the cards included in the blacklist will be excluded from the cube. This can be particularly useful when building format-specific cubes that have a ban list, such as Legacy. Wizards of the Coast provides format-specific ban lists that you can refer to when building your cube.
In Vintage, there is also a ban list that includes cards that reference Ante, Chaos Orb, Falling Star, and Shahrazad. However, I am currently not using a blacklist for these cards. Additionally, these cards are not frequently played (even in vintage cubes) and are not ever recommended for inclusion by the algorithm due to low play rates. So in absence of a banlist, the results generated in this cube are a faithful representation of card usage across the cube community.
AnalysisWhat follows are analysis and visualizations of the results of using an algorithm to generate a cube. Data in this analysis was current as of 2023-09-12
Inclusion Rate DistributionAnalyzing the distribution of inclusion rates can provide valuable insights into card popularity and help identify both popular and niche cards. By examining the relationship between inclusion rates and card counts, an interesting distribution is shown where inclusion rates are contant accross the majority of the distribution but on both tails (i.e. high and low inclusion rates) there is a drop off. Generally speaking though, cards that have high inclusion rates can be considered core to Vintage cubes.
However, an interesting observation in this distribution is the presence of a drop in cards included in the cube. Based on the graph, this phenomenon indicates that the core of the cube consists of cards that are included at rates above 50%, suggesting their widespread acceptance and popularity within the community. On the other hand, cards with inclusion rates below this threshold may represent newer staples that haven't yet gained widespread recognition or cards that provoke disagreement and varying opinions among players.
I'd say that generally by studying the distribution of inclusion rates relative to card counts, cube designers can make informed decisions about card selection and adjust the composition of their cube. They can ensure that the core includes highly regarded and frequently used cards while also considering the inclusion of emerging staples that may have the potential to reshape the cube's metagame. Understanding the dynamics of inclusion rates helps me strike a balance between established favorites and fresh additions.
ELO DistributionELO scores are not normally distributed. This makes presenting visual analysis on these less interesting (normalizing these scores still results in heavily skewed visual representations) and as such, when we discuss ELO scores and their analysis I have filtered out cards which are two standard deviations above the mean ELO score, this causes the remaining cards to be fairly close to a normal distribution. Further, the fact that there are a few cards which dominate other card's ELO is in line with experience that vintage cube is often centered around opening and drafting really powerful cards. Here are the cards that are two deviations above the mean ELO score.
Card Name | Inclusion Rate | ELO | Type | Color Category |
---|---|---|---|---|
Black Lotus | 98.98% | 16651.9 | Artifact | ![]() |
Ancestral Recall | 95.92% | 6828.1 | Instant | ![]() |
Sol Ring | 96.94% | 6553.9 | Artifact | ![]() |
Mox Sapphire | 97.96% | 6020.6 | Artifact | ![]() |
Time Walk | 92.86% | 4925.5 | Sorcery | ![]() |
Mox Ruby | 97.96% | 3046 | Artifact | ![]() |
Mox Pearl | 97.96% | 2942.4 | Artifact | ![]() |
Mox Jet | 96.94% | 2928.4 | Artifact | ![]() |
Mox Emerald | 97.96% | 2757.3 | Artifact | ![]() |
Mana Crypt | 90.82% | 2353.7 | Artifact | ![]() |
Ragavan, Nimble Pilferer | 87.76% | 1742.6 | Creature | ![]() |
Mana Vault | 88.78% | 1720.5 | Artifact | ![]() |
The One Ring | 31.63% | 1707.7 | Artifact | ![]() |
Mox Diamond | 87.76% | 1705.4 | Artifact | ![]() |
Strip Mine | 86.73% | 1704.4 | Land | ![]() |
Mana Drain | 87.76% | 1704.4 | Instant | ![]() |
Oko, Thief of Crowns | 82.65% | 1690.6 | Planeswalker | ![]() |
Demonic Tutor | 94.90% | 1688.4 | Sorcery | ![]() |
Minsc & Boo, Timeless Heroes | 61.22% | 1686.3 | Planeswalker | ![]() |
Library of Alexandria | 83.67% | 1677.6 | Land | ![]() |
Sheoldred, the Apocalypse | 62.24% | 1659.8 | Creature | ![]() |
Mind Twist | 80.61% | 1650.9 | Sorcery | ![]() |
Once outliers are excluded from the analysis, visualizing ELO scores can provide insights into the relative power level of the remaining cards. Here we explore whether certain types or colors of cards tend to have higher ELO scores. First, the general distribution of ELO scores among cards in the cube is shown below.
The above graph still shows that there is a slight positive skew to card ELO, meaning that the cube tends to be characterized by individually powerful cards. I would be interested in replicating this analysis on cubes for other formats such as Pauper, Modern and Legacy and overlaying the density estimation to compare the relative ELO distributions per format. For now, we can still see that the average ELO is (subjectively) high, clocking in at 1378.
Next lets looks at ELO ranges based on card color categories.
The box plots are ordered by median ELO rating ascending. Here we can see that land cards have the highest median ELO and that some categories such as colorless cards have average ELO ratings, but several outlier cards being represented as diamonds on the plot.
Next we can see the elo distributions broken out by card type. Interestingly Enchantments have high median ELO scores but a lot of mediocrity in terms of the range of ELO scores for the card type. In case you are curious the single diamond outlier for creatures is Orcish Bowmasters, which was just on the cutoff of being two standard deviations of the full data set's mean ELO score and included on the outliers list.
Something I've found interesting two do is to look at the relationship between inclusion rate and ELO score. In particular I am curious if there are staple cards in cube that perform poorly on aggreagate. For this analysis I took a staple card to be any card that has a high inclusion rate and poor performance to be a low ELO score relative to that inclusion rate. Below is a table with cards that have the highest differential in inclusion rate and ELO ranking.
Card Name | Inclusion Rate | ELO | Type | Color Category |
---|---|---|---|---|
Temple Garden | 92.86% | 1280.7 | Land | ![]() |
Overgrown Tomb | 92.86% | 1318.3 | Land | ![]() |
Breeding Pool | 92.86% | 1319.2 | Land | ![]() |
Reclamation Sage | 80.61% | 1246.7 | Creature | ![]() |
Thalia, Guardian of Thraben | 94.90% | 1339.5 | Creature | ![]() |
Sacred Foundry | 92.86% | 1331.4 | Land | ![]() |
Knight of Autumn | 67.35% | 1169.9 | Creature | ![]() |
Eternal Witness | 86.73% | 1300 | Creature | ![]() |
Faithless Looting | 95.92% | 1359.6 | Sorcery | ![]() |
Savannah | 93.88% | 1349.2 | Land | ![]() |
Interestingly a wide range of non-blue/Abzan lands are on thislist... I guess they call it Junk for a reason ¯_(ツ)_/¯. Other than that, depending on your preferences none of these cards are exciting, all are pretty strong staples in vintage cube that I would imagine are not being rotated out anytime soon. Note that with the exception of Knight of Autumn, all of these cards have solid ELO scores relative to most cards you will find in an average cube.
I also did the same analysis in reverse, finding cards that have low inclusion rates by cube designers but high ELO ratings relative to their inclusion rates.
Card Name | Inclusion Rate | ELO | Type | Color Category |
---|---|---|---|---|
The One Ring | 31.63% | 1707.7 | Artifact | ![]() |
White Plume Adventurer | 36.73% | 1647.7 | Creature | ![]() |
Orcish Bowmasters | 38.78% | 1650.2 | Creature | ![]() |
Lutri, the Spellchaser | 26.53% | 1550.9 | Creature | ![]() |
Undermountain Adventurer | 26.53% | 1512.5 | Creature | ![]() |
This is interesting; Lutri, the Spellchaser is the only card that has been in print for more than 15 months. All the other cards are relatively new to Magic. Though after more than a year the low play rate of initiative cards is interesting and I wonder if this is because cube designers find the swingy gameplay of mechanics like Initiative to be unfun - that would be saying a lot for a format which is defined by opening powerful and busted cards in your packs. FWIW, I am not personally a fan of initiative.
Finally, I am curious about the overall relationship between inclusion rate and ELO. For that we have a scatter plot with regression line to show this relationship.
A correlation coefficient of 0.23 between card ELO and inclusion rate suggests a weak positive relationship between a card's power level as measured by ELO and its likelihood of being included in other cubes. I find this intriguing because it indicates that while card strength does play a role in card selection, it is far from the only factor we cube designers consider. The relatively low correlation suggests that other elements, such as thematic coherence, card interactions, or the desire to include a diverse array of strategies, might also significantly influence card inclusion. It's plausible that many cube designers of vintage cube are looking beyond simple power level to create a more nuanced and engaging gameplay experience. This could involve including cards that might be less powerful in isolation but contribute to richer, more complex deck-building puzzles or offer more interactive gameplay. The data hints that the art of cube design is a multifaceted endeavor, incorporating a variety of considerations that go beyond merely optimizing for the most powerful cards.
However, due to the construction methodology of this cube, we don't get to sample other ways that cube's are constructed. I would be interested in the future to quantify things like synergy or tackle this problem from a different angle (e.g. performing cluster analysis).
Card type distributionHere is a wheel showing the percent distribution
These numbers seem fairly standard as we would expect from a cube that is generated off of the aggregate data of other cube designs.
The ArchetypesWhat follows below is an almost verbatim copy of @LSV's ultimate guide to vintage cube. The cube supports all the following draft archetypes and LSV's explanations are great to read. If you prefer to read the original articles on ChannelFireball, here are the links:
Note that these articles are now quite old, and vintage cube has gotten faster over the years. With this in mind many of the individual cards mentioned might have different power levels than when the aritcles were written. That said, all of these archetypes still exist in vintage cube and are quite fun to play!
Now let's get started on archetypes. There’s a lot of nuance here, especially with knowing how to blend Reanimator and Sneak Attack, or Big Mana Blue and Storm, or pivot from one of these to another, but knowing the base decks is a huge upgrade to your Cube skills. These are specific to Vintage Cube, but some, like Mono Green or White Weenie, are pretty much the same in Legacy or Modern Cube as well.
First, let’s start in the most logical place – blue decks. Blue is the best color in Cube, hands down, and the average Cube draft can support five blue drafters without much trouble. Since Cube has so many playables, you can be cut on blue in both directions and still have a lot of powerful blue cards. Plus, blue gets some of the busted cards like Ancestral Recall, Time Walk and Mana Drain, so there’s a good reason that it’s the base for multiple different archetypes.
This is a bunch of different decks rolled into one, and is the deck most people think of when they imagine Vintage Cube. This is the deck you end up in when you just take good blue cards, and as such once you learn the core, you can customize it with various win conditions. This is a hard one to learn because of how many ways it branches (drafting Mono Green is much simpler, for example), but once you know how to draft a good blue deck you can always have a good fallback plan.
Game PlanDisrupt the opponent with counterspells, and develop your mana base/find your win conditions using card draw spells.
Key CardsThe core of “the blue deck” are the cards I mentioned in part one, the cheap interactive and flexible cards, plus card draw like Fact or Fiction, Compulsive Research and Thirst for Knowledge. If you start with these, filling in a win condition isn’t too hard.
This is the baseline which I start every Cube draft with, at least until I pick up a more specific engine card that pushes me into a theme. This strategy is compatible with all the combo decks and only doesn’t mesh well with aggro (Mono White and Mono Red) and Mono Green Ramp. If you grow confident in identifying and taking these baseline blue cards, you have a good base to start with that lets you branch into all the decks I’ll be talking about next.
Now that we have Generic Blue Cards on our mind, let’s get more specific.
This deck relies heavily on artifacts and is the best deck when you have picked up good acceleration. Bonus points if you spike a Sol Ring or Mana Drain, as this uses those cards better than any other deck.
Game PlanUse artifact ramp to turbo out a finisher and go over the top of whatever the opponent is doing. Example Draft.
Key CardsThis deck is generally blue/red, as Inferno Titan, Wildfire, Wheel of Fortune, Fire // Ice and Dack Fayden are exactly what you want to pair with your good blue cards. This deck lives or dies on its artifact ramp, and needs 5+ ways to generate a lot of mana.
Ways to Get Into the DeckThe Sneak Attack deck is my pick for the most reliable broken deck in the Cube. It’s a two-card combo and Sneak is fast, hard to interact with and combines with a ton of other cards, so the second part of the combo isn’t even very specific. I basically won’t pass Sneak Attack at this point (and I’m sure by writing this I’m making it less likely I see it in the future).
Game PlanUse Sneak Attack or Through the Breach (or Channel) to turbo out Eldrazi or other giant monsters.
Key Cards*if you’re a Channel deck, move up colorless creatures appropriately
This deck is usually Grixis, mainly blue-red with a light black splash. It doesn’t mind having some acceleration, but it’s not nearly as critical as in the big mana deck. Part of the beauty of this deck is that the whole Sneak package is very small – the addition of Sneak Attack, Vampiric Tutor, and two Eldrazi can make any blue deck into one with a busted endgame. It’s the perfect finisher for decks with good card draw and disruption.
Under no circumstances should you play Show and Tell (and especially not Eureka). There are zero recorded instances of a player casting Show and Tell and winning, so don’t get tricked. Casting Show and Tell and having to wait a turn while also letting the opponent put their best permanent into play is just a recipe for disaster. It’s a fine sideboard card against low-curve aggro, but for the love of Jace, don’t take this card early or play it in your main deck.
Ways to Get Into the DeckThis deck uses the combo of Kiki-Jiki or Splinter Twin plus one of Exarch, Pestermite, Zealous Conscripts or Restoration Angel (note that Twin plus Angel is the only combo that doesn’t work). It’s Izzet with a white splash for Restoration Angel or Recruiter of the Guard and usually plays like a control deck with a combo finish (though some versions can be more aggressive, with creatures like Brazen Borrower and Vendilion Clique as pressure).
Game PlanCombo the opponent out, with a backup plan of getting value from Kiki / Twin / Resto on creatures with good ETB effects. Example Draft.
Key CardsLike the Sneak deck, this deck can be a finishing package for any blue control deck, and as such is a desirable place to be. It even dovetails nicely with Sneak, as Sneak Attack makes the combo very cheap to assemble. I’m willing to take Kiki very early as a result, which unsurprisingly is one of the ways to end up in this deck.
Ways to Get into the DeckAhh, Storm. This is my favorite deck, though I’m in no way saying it’s the best deck (as I’ve stated, I think that honor belongs to a good Sneak Attack deck). I love drafting Storm because of how fun it is to solve the puzzles it presents and because sometimes you end up winning with off-the-wall combinations of cards.
The upsides of Storm (in terms of winning the game) are that it can be extremely fast, hard to interact with and doesn’t care at all about what the opponent is doing. The downsides are that it’s hard to play, requires a lot of very specific cards and can easily not get there even when you do have a good storm deck. Now that I’ve sold you on drafting the best archetype, let’s dig into it.
Game PlanGenerate mana with Mana Flare effects or Ritual effects, cast draw sevens and win via Tendrils of Agony or Brain Freeze (hopefully). These are the main three engines, and you want as many of these as possible in your storm decks. Example Draft.
Storm is a bit trickier than most of these decks, so I’m going to approach it a little differently. It’s less about key cards than key combinations, so let’s talk those first:
Brain Freeze is the best win condition, followed by Tendrils of Agony, though you can string together all sorts of crazy ways to win. The trick with Storm isn’t winning the game, it’s generating mana and cards – once you have infinite mana and all your cards, winning the game isn’t that hard to accomplish. I don’t prioritize win conditions much and am perfectly content in slotting in any of the combos like Sneak Attack, Kiki-Jiki or just Eldrazi plus Mind’s Desire to finish the game. I’ve even hard-cast Emrakul off Turnabout and Time Spiral, though my favorite has been Upheaval plus Emrakul the Promised End in the same turn (I forced them to discard all their lands).
How You Get into the DeckI love Storm, and love how complicated it is. This is not an entry-level deck, so I’d highly recommend the other decks I’ve talked about before Storm if you are dipping your toes into the format.
This deck is heavy black, and in the case of the best versions, even mono black (though blue-black and black-red are are more common than mono black). It’s capable of some of the fastest draws in the format and does a lot of sweet things that can overlap with the Sneak Attack deck.
Game PlanPut a giant monster into your graveyard, then reanimate it. Example Draft.
Key CardsThe core of the deck is black, while blue and red offer various discard possibilities to supplement that. Red also offers Sneak Attack/Through the Breach, which are a natural fit into a deck with tutors and big monsters. This deck doesn’t need artifact ramp as much as other decks, so move that down if you are going this direction.
Note on Survival of the Fittest: I haven’t found traditional Recurring Nightmare/Survival versions of the deck to be great. Besides Survival, there are very few green cards that you actually want in the deck, so I usually stick to blue or red.
Ways to Get Into the Deck:Besides broken cards, Entomb, Griselbrand and tutors, this deck doesn’t have a lot of cards I like taking very early. I won’t usually start with a pick two Reanimate and I’m not taking Putrid Imp or Woodfall Primus in my first few picks. However, if you see a few reanimator cards in your first few picks, take generic good cards and note if they wheel. If they do, this deck is open and can be quite good when it is.
This deck is better in Legacy Cube than in Vintage Cube because it doesn’t gain much from the transition, but it’s still a great deck when it’s open. This is also one of the easier decks to get proficient with because, well, it’s mono green. Take the green cards and you’ll do a decent job (as long as you avoid traps like Thrun, the Last Troll).
Game PlanUse mana ramp creatures and artifacts to accelerate out a fast finisher or disruption. Example draft.
Key Cards:Opposition is a special case. It’s mostly a subset of mono green (though there is a blue/white Opposition tempo deck
out there), and it’s one of the best cards for mono green. If you get an early Opposition, the main changes you want to make to your pick order are that Tropical Island and mana dorks that tap for blue become critical (alongside fetches that get Trop) and blue cards like Remand, Mana Leak, Fact or Fiction and Brazen Borrower become viable picks. You also want to lean towards a more creature-heavy deck and Deep Forest Hermit becomes awesome instead of just good.
One of the risks of this deck is that it’s pretty focused and doesn’t overlap with many other decks. As a result, I don’t get into it all that often, though it’s very good when it is open.
Fair warning – I love this deck but it’s rarely the right deck to draft. It’s a lot of fun but it’s a gamble that you’ll see the right cards, and even then it doesn’t always come together. It does do a lot of broken things when you have a good version though, so have no fear that the power isn’t there.
Game PlanGo super hard on artifact ramp and use sweepers like Balance, Wildfire or Upheaval to wipe the opponent’s board. It also has Tinker for early giant artifact monsters.
Key CardsThis deck tends to be near mono blue but often splashes one or two cards of various colors. You want room for good colorless lands and because of all your random signets, it’s easy to play all these sweet cards. I’ve played mono blue with Demonic Tutor, Balance and Sylvan Library before.
Ways to Get Into the DeckI really enjoy playing this deck since it’s a nice change of pace.
Mono White or “white weenie”, is a deck full of underpowered cards that actually manages to win a ton. When it’s open, which it usually is, this deck can be very good.
Game PlanPlay cheap white creatures and back them up with mana denial and equipment. Example Draft.
Key Cards:This deck is pretty straightforward and one of its advantages is that most of its cards are interchangeable. As long as you have some mana disruption, what you pressure the opponent with doesn’t matter quite as much.
I’ll admit I don’t draft this deck a ton but I do have a lot of respect for it. As long as you lean into one-drops and away from Brimaz.dec, you can crush all the idiots like me by casting a single Thalia.
Ways to Get Into the DeckI used to force this deck in Cube, for reasons unclear to me. Maybe I just like blowing up peoples’ lands? I even drafted it at the World Championship in 2012, the only time (that I know of) that Cube was played on the professional stage. This deck is also quite good and also tends to be underdrafted.
Game PlanAttack the opponent with one drops, disrupt their mana and finish them with burn. Example Draft.
Key CardsThe deck is pretty single-minded. Attack them early and blow up a land or an artifact and you usually have enough time to burn them out. It’s very important to note that this deck gets much worse if it’s two colors, so I only splash if I have three or four fixers and only splash one to three cards.
Ways to Get Into the Deck