Improving Deck Color Classification

Ever looked up a deck on the site and noticed unexpected splash colors?

The classification logic has served us well for years by optimizing for simplicity and straightforward rules, but the way it handled hybrid cards, activated abilities and a handful of edge cases was worth revisiting. We recently reworked how deck colors are determined. We’ll explain what changed and why.

Why it matters

Deck color labels aren’t just cosmetic: they decide how your draft is categorized in the aggregate data, feeding color-based win rate breakdowns and archetype filters across the site. A mislabeled deck is a small annoyance on its own, but over thousands of drafts it quietly distorts the numbers.

What was happening

Our legacy logic was relatively straightforward: If there was at least one mana source of a given color, we count up the cards with that color identity. 4 or more cards was a main color, 3 or less was a splash. However, this logic led to unintuitive results more often then we’d like.

Previously: 

Now: 

Decklist ↗

Take this ECL deck for example, playing only Islands/Mountains, featuring Deceit and Tam, Mindful First-Year
This was previously labeled as URbg. Hybrids were counted as both colors, and a single treasure-making source in Flamekin Gildweaver was enough to consider them as splashes.

Under our new logic, this would be classified as a UR deck, which would match how most players intuitively classify it. Let’s break down some of the changes in more detail.

#1. Hybrids will pick one color

Previously: 

Now: 

Decklist ↗

This UG deck plays seven different hybrid cards spanning every color. Great Forest Druid is a five-color mana source (more on this later), which was previously treated as a license to treat everything as a splash. So both sides of each hybrid card produced splashes and even a main color with 4 “red” cards. However, by making hybrids pick a main color, this deck conforms with player expectations as a Green-Blue deck.

New Rule:
Hybrid cards now pick the color with more sources.

Previously: 

Now: 

Decklist ↗

A subtler case in this GW deck: Taster of Wares is a genuine splash. The hybrid is handled by the previous fix, so red is no longer a splash enabled by the any-color mana producer. The hybrid commits to white given the mana base. Finally, note the two copies of Trystan with a black activated ability backed by a Swamp. Together that used to contribute to classifying black as a main olor, but now reads as a 3-card black splash.

But what if you were playing a card like Trystan without expecting to use its ability?

#2. Abilities need a land

Previously: 

Now: 

Decklist ↗

This UR deck is playing Oko which has a activated ability, but not a single Forest. While the deck can technically activate it using Treasures or Foraging Wickermaw, we decided it was better to recognize commitment to splashing the card by playing a land of the required color instead of relying on conditional sources.

New Rule: Ability splashes now need a dedicated land.

Multi-color abilities work the same way: every off-color still needs its own land.

Previously: 

Now: 

Decklist ↗

This UBG deck plays Everything Pizza but has no Plains or Mountain. Previously, a single five-color source like Omni-Cheese Pizza was enough to enable the splashes. Now, without a dedicated land for either color, both drop.

---

Additionally, TMT featured a lot of off-color abilities and we were seeing decks labeled with more main colors than their mana base actually committed to.

Previously: 

Now: 

Decklist ↗

This UG deck plays four copies of Everything Pizza, which has a WUBRG activated ability, with Frog Butler providing the off-color sources. Previously this would be treated as having 4 cards with WUBRG casting cost. Now, ability colors can splash but aren’t considered for main color, so we consider as splashes.

New Rule: Abilities no longer count for main colors.

If you made it this far, there’s one topic left: non-land sources.

Should one Manalith or Treasure maker be enough to splash every color?

#3. Sources required per pip

Unchanged: 

Decklist ↗

This WR deck splashes with no Island: the off-color pip is covered by Treasure sources. Previously a single non-land source would have been enough, even for multi-pip cards. This deck has multiple, so it keeps the splash.

New Rule: Two non-land sources now cover one pip.

This rule was aimed mostly at cube, where it’s more common to play cards with no intention of casting them: Reanimate and Flash decks often feature these.

Previously: 

Now: 

Decklist ↗

This BR cube deck plays Woodfall Primus, Torsten and Sin: heavy off-color creatures the deck plans to cheat into play, not hard-cast.
Take Woodfall Primus’: Multiversal Passage is the deck’s only green-producing land, covering one pip. The remaining two would need four non-land green sources and the deck has just Fable of the Mirror-Breaker. Same story for white and blue, so under the new rule all three splashes are dropped.

How it all fits together

%%{init: {'flowchart': {'nodeSpacing': 25, 'rankSpacing': 35}, 'themeVariables': {'fontSize': '13px'}}}%%
flowchart TD
    classDef main fill:#4a7fb5,stroke:#2c5282,color:#fff
    classDef splash fill:#3a9e8f,stroke:#276e62,color:#fff
    classDef ignored stroke-dasharray:4 2,color:#888

    A([For each color: W U B R G]) --> E{"Requires this<br/>color to cast? #1"}

    E -- No --> G{"Used by an<br/>ability?"}
    G -- No --> NONE1[Ignored]:::ignored
    G -- Yes --> H{"Has a land for<br/>this color? #2"}
    H -- No --> NONE2[Ignored]:::ignored
    H -- Yes --> SPLASH[Splash]:::splash

    E -- Yes --> F{"Has enough sources<br/>for this color? #3"}
    F -- No --> NONE3[Ignored]:::ignored
    F -- Yes --> C{4 or more<br/>cards?}
    C -- No --> SPLASH
    C -- Yes --> MAIN[Main color]:::main

Beyond the three big rules, a handful of smaller cases also got cleaned up: lands with off-color abilities like Hanweir Battlements and Gavony Township now register as ability splashes, split cards like Life // Death where one half is uncastable get correctly classified, and multicolor cards like Atraxa are skipped entirely when uncastable instead of adding partial splashes.

---

I hope you enjoyed this look under the hood at how we approached the problem. A couple of these cases involve some ambiguity, so we optimized for simplicity over precision: Simple is better than complex.

We expect these rules to hold up well across future formats, but will continue adjusting if the situation calls for it.

In the coming weeks we’ll re-aggregate the recent formats most affected by these changes, like ECL and TMT.

Found a deck that still looks off? Let us know in our Discord.