How synergy
scoring works

You pasted a deck, and Deckaugur gave you a grade, a stack of suggestions, and maybe a few uncomfortable warnings. Fair questions follow: where do those numbers come from, and why should you trust them?

The short answer: every judgement traces back to the printed text of the cards in your deck. There is no language model guessing and no popularity statistic standing in for analysis. This article walks through the whole pipeline, from how a card is read to how the grade is assembled.

Every card answers two questions

When a card enters the pool, its oracle text is read and it is tagged along two dimensions:

• What does it enable? The effects it produces: creating tokens, making treasure, drawing cards, putting creatures onto the battlefield, filling the graveyard.
• What does it care about? The effects it rewards: a card that triggers whenever a creature enters cares about creatures entering; a card that gets cheaper as your graveyard fills cares about the graveyard.

Synergy is then a concrete, checkable claim rather than a vibe: card A enables something card B cares about. Two cards can share a keyword and have no relationship, and two cards with no words in common can be a working engine. The tags capture the function, not the vocabulary.

The tag dictionary covers the full Commander card pool with roughly ninety pattern rules plus hand-curated corrections, and it distinguishes cases that look identical at a glance. Cards that want a big graveyard are tagged differently from cards that play out of the graveyard, because a deck can be full of one and empty of the other.

How a suggestion earns its rank

The suggestions panel, WHISPERS FROM THE LIBRARY, scores every candidate card in your commander's colour identity against the deck as it stands. The core rule keeps the panel honest:

• A tag only contributes full points when your deck both enables and cares about the effect. A tokens payoff scores highly only if the deck actually produces tokens and rewards them.
• A card that merely produces a side effect your deck ignores gets a token score at most. This is why a landfall card does not get suggested into an elf deck just because both involve creatures hitting the table.
• Standout suggestions carry badges: a combo badge when the card closes a loop with cards already in the deck, and a mana-efficiency badge when it does its job cheaply.

The practical consequence: WHISPERS gets sharper as your deck gets more opinionated. With five cards it can only read broad strokes; with sixty it knows exactly which themes are live and stops offering anything generic.

Why a warning fires, and why sometimes it deliberately does not

SOMETHING TO RECONSIDER is the inverse check: it looks for cards whose conditions your deck never meets. A card that cares about instants and sorceries in a deck running six of them is dead weight most games, however strong it looks alone. Where the engine knows a better fit, the warning comes with a suggested replacement.

One refinement matters here. If two or more cards in your deck share the same unmet tag, no warning fires. Several cards caring about the same missing effect is not a mistake; it is a theme you are still building toward. The warning exists to catch the lone stray, not to fight your plan.

Clusters, combos, and engines

Beyond pairwise links, the engine looks for structure:

• Synergy Clusters groups cards that share two or more combo tags, scored by how many cards and tags participate. This is your deck's engine room made visible.
• Two-card loops are detected against a curated whitelist of combo-capable tags, and require the two cards to bridge different tags, which filters out false positives where two cards merely resemble each other.
• Engine detection flags hub cards, such as a mass-blink effect or a free sacrifice outlet, when they connect to two or more payoff cards. A classic example the engine finds on its own: Brago, King Eternal plus Village Bell-Ringer, a genuine loop built from a blink tag meeting an untap tag.

What the grade means

The deck grade summarises the density of working relationships across your ninety-nine: how many cards participate in the deck's themes, how many sit isolated, and how much of the deck pulls in the same direction. A high grade does not promise wins; Commander games are decided by pilots, politics, and variance. What it tells you is that the deck is coherent: whatever you draw is likely to relate to what you already have. The StarMap shows the same information spatially, and the loneliest node on that map is usually your next cut.

Why rule-based instead of AI

A deliberate choice. A rule-based engine is deterministic: the same deck always gets the same analysis, every suggestion can be traced to named tags on named cards, and when the engine is wrong the fix is a visible tag correction rather than a retrained model. The flag icon on every card in the app feeds exactly that loop: players report a wrong judgement, the tags get corrected, and the whole pool benefits. The trade-off is honesty about coverage: the tag dictionary is audited continuously, but not every archetype is documented equally well yet. When the engine misses something subtle, that is usually a missing tag, and it is fixable the same afternoon.

Further reading

• The Commander Deckbuilder Guide: how EDHREC, Moxfield and Archidekt work, and where Deckaugur fits among them.
• Audit your existing deck with paste-import: get a full synergy read on a deck you built years ago in about thirty seconds.
• Why your deck is missing removal: how interaction erodes out of a deck, and how much is enough.