Commander decks are usually something of a passion for anyone that plays the format. I've seen literal suitcases of decks—each a different commander and unique to its own theme—carted around events by eager players looking for a game. I've seen hundreds (perhaps thousands) of decks over the years as a Commander writer and content manager.

There is an infinite variety of decks, but there isn't always an infinite variety of cards.

That's the anecdote I've come to share from looking across all those decks. While deck diversity and commander choice seems to constantly increase, I've also argued too many cards show up too frequently regardless of the theme.

But is that true?

There are plenty of ways to look at Commander decks, but only one that functions as a recommendation and aggregation engine: EDHREC. Aggregating decks to find the most-played cards, identifying those most played together (synergy) and overall makes seeing popular cards—cards popular for good reasons—a snap.

What is Data?

Let's talk about what data is and how we can look at it. While there is a ton of nuance in statistical analysis, addressing basic assumptions and ensuring we're all on the same page is easy.

Any source of data, aggregated or not, isn't necessarily useful. What it is, what it tells us, and how reliable it is at both is always a consideration. Aggregating data—rolling it all together—typically makes it more reliable since outliers—suspect or highly unusual points of information—get washed away as the most frequent observations pile up. While sometimes those outliers are important (look at sales charts, for example), assuming your data sets follow a standard distribution and acting accordingly is safe. Since we're talking about tens of thousands of decks, each with one hundred cards in them, our sample size—how much data we have—is big enough to make these general assumptions.

Of course, that may not be enough. The only decks that make it into EDHREC come from two events occurring:

The biggest concern about EDHREC data holds true to any source of Commander decklists online: Not every deck is getting entered. Unlike competitive formats, where all of the most successful decks from a large sample of competitive play get entered into databases, there is no standard for Commander deck entry beyond an individual being passionate. Taken in whole—aggregate—it's likely trustworthy, but not every kind of Commander player enters their decks online (I don't!) and those that do reasonably don't enter every deck (Literal suitcases of decks, remember?).

If there is any skew, it's a skew towards players interested in making powerful Commander decks. That's the main benefit of pushing your list into EDHREC—getting recommendations of popularly played cards and their "synergy" with your commander—but whether that's enough to make the data unreliable is beyond reasonable analysis here.

What is Our Data?

I'll get to the takeaways and perspective I have on these facts at the end of the article, but here is some data provided by EDHREC, created directly by the owner last week. (Thanks again, Don. Seriously—this stuff is awesome!)

Let's start with the most-played cards by percentage of the decks with that card's color identity. The top 50 by percent played in that color identity.

CardPenetrationColor Identity
Sol Ring 0.768
Cyclonic Rift 0.502
Swords to Plowshares 0.465
Eternal Witness 0.449
Cultivate 0.375
Counterspell 0.37
Kodama's Reach 0.356
Demonic Tutor 0.35
Putrefy 0.341
Supreme Verdict 0.333
Izzet Signet 0.332
Sakura-Tribe Elder 0.324
Sun Titan 0.32
Rakdos Signet 0.319
Boros Signet 0.318
Orzhov Signet 0.317
Lightning Greaves 0.313
Phyrexian Arena 0.305
Mirari's Wake 0.301
Path to Exile 0.298
Dimir Signet 0.289
Swiftfoot Boots 0.287
Merciless Eviction 0.286
Utter End 0.284
Boros Charm 0.282
Acidic Slime 0.281
Azorius Signet 0.279
Enlightened Tutor 0.277
Beast Within 0.274
Simic Signet 0.269
Rhystic Study 0.267
Diabolic Tutor 0.267
Wrath of God 0.267
Brainstorm 0.266
Solemn Simulacrum 0.262
Anguished Unmaking 0.26
Krosan Grip 0.254
Aura Shards 0.252
Mortify 0.249
Golgari Signet 0.244
Mystical Tutor 0.23
Terminate 0.23
Blasphemous Act 0.229
Avenger of Zendikar 0.228
Nicol Bolas, Planeswalker 0.225
Birds of Paradise 0.222
Return to Dust 0.22
Gisela, Blade of Goldnight 0.219
Chromatic Lantern 0.214
Chaos Warp 0.213

That percentage played out of possible decks can be called penetration—how far something penetrates into all of something else. Penetration is basically frequency, but it's looking at the total possible appearances rather than just counting the number of times it did.

Speaking of that, here's the percentage that any given color combination is being played:

Color Identity% Decks

(If you want to play with it a bit nicer, I've made a Google Sheet available that's all of the above plus carries the penetration ranking all the way down through 300 cards. Enjoy!)

What Does Our Data Say?

Let's talk about that second chart first: For just becoming widely available, four-color decks have a surprising amount of play. The Atraxa, Voice of Praetor's combination—not necessarily Atraxa decks, though—is as popular as Jund decks. That's huge, and shows how quickly the new combinations and their accessibility through preconstructed decks have been picked up.

Mono-black and five-color decks being popular struck me as odd, but there are several reasons it makes sense. Mono-black can tutor, ramp and combo successfully while having plenty of removal and ways to recover quickly by abusing the graveyard. Without more information that's all I can speculate, but assuming players are looking to build powerful, resilient decks fits with the data feeding.

Five-color's popularity is likely driven by both the popularity of Slivers and the desire for players to have it all. You can just play the best cards—or at least all your favorites—if your deck just plays all five colors. It's strange as I've rarely seen five-color decks in the wild during my travels, but plenty of mono-black decks. This is a good way to check my assumptions from anecdotes: Being surprised means I need to constantly reevaluate my other thoughts from playing Commander.

Overall, the spread of color identity by deck seems great. There's no runaway color or choice dominating, and even the newest color combinations have great representation. Deck diversity looks healthy from these counts.

Moving back to the first chart, there's some fascinating facts here. Let's start with the biggest one: Sol Ring is played in 75% of all Commander decks.

It makes sense that Sol Ring is so heavily played. It's in 100% of preconstructed decks, and is typically one of the cards suggested as the defining staple of the format. It's cheap to acquire, obviously powerful and virtually unplayable outside of Commander.

The data here can't tell us if Sol Ring is actually effective in those decks, but looking at analogous situations from other Constructed formats would be something like this: If a card appeared in 24 out of 32 possible Top 8 slots in every tournament, would the card get banned?

Data isn't always appropriate to argue a ban—Swamp appears in 93% of all Commander decks running black—since context matters, but it does reinforce the question I ask myself when I build: Does this deck need Sol Ring? I've started to warm up to the idea that Sol Ring itself isn't a game-breaking card, but perhaps it's simply overplayed given the perception it's "needed" in every deck.

Looking at a different card, Cyclonic Rift is the most-played card across all decks with blue, at 50%. It's also the colored card with the highest penetration for its possible decks. While the data alone can't tell us this, we know context around these cards: Cyclonic Rift is different from Sol Ring. At many points in a game, Sol Ring is a marginal—or even dead—draw. Cyclonic Rift isn't. If you're ahead or behind, Sol Ring likely isn't the difference maker for getting back into the game or putting it away for good. Cyclonic Rift is often backbreaking to opponents at parity, and is one of the few ways to come back from a losing position.

When the data is telling us a card is in half of all possible blue decks but it's about a card that isn't as innocuous as Sol Ring, I think it's a stronger discussion point for banning. Having every deck start with commander and Sol Ring, then Cyclonic Rift if you're in blue, seems wrong. There are just too many possible cards for any one to be in so many decks—at least that's what I believe.

The second most-played card within its color identity surprised me: Swords to Plowshares. It shouldn't be surprising to anyone that's listened to enough of the Command Zone podcast as Swords to Plowshares is one of the go-to spot removal cards every time the subject of removal comes up. It's played to almost the same penetration in white decks as Cyclonic Rift in blue decks, but I don't feel the same about Swords as I do Rift.

Why? Path to Exile's penetration is "only" 30%—still high—but is effectively the same card in most cases.

The difference between a card like Cyclonic Rift and cards like Sol Ring and Swords to Plowshares is that what Ring and Swords do is easily replaceable or effectively marginal against alternatives. Cyclonic Rift is an instant, asymmetrical board wipe for all non-land permanents, while Swords to Plowshares can only answer an individual creature and doesn't help against a variety of other threats.

Discussing the penetration for so many cards is beyond what this article should get into, but having cards like Cultivate, Sakura-Tribe Elder and Kodama's Reach near the top matches up to anecdotes about ramp being so big, while Demonic Tutor, Supreme Verdict, Counterspell Phyrexian Arena lend to the idea that control decks—players looking to answer and take over games—drive some of the data for decks.

Ultimately, data is just information. What we read from it and what we should do based on it is a bigger conversation. It's why political parties and business plans can have such variety in actions, and why we're so passionate about discussing those differences. Understanding the data is step one, but I leave a lot here for your own conclusions.

I look forward to your comments this time.

Adam Styborski