How Deckcipher's Commander tools work

This page explains where the card data comes from, how each tool reaches its answer, which parts are official Commander rules and which parts are Deckcipher's own read, and what these tools cannot tell you. It is written so you can check our work and argue with it.

Last updated August 9, 2026. Bracket rules version 2026-02, the February 2026 Commander update, last reviewed August 6, 2026.

The short version

• Card dataScryfall
• Bracket rulesVersion 2026-02
• Game Changers tracked53 cards
• Named combo packages5 packages
• Where it runsYour browser
• Model-generated resultsNone

Where the card data comes from

Every card name, type line, mana value, color identity, rules text, printing, and price comes from Scryfall. Deckcipher keeps no card database of its own, so a card that was printed yesterday is available as soon as Scryfall has it.

Card lookups pass through a Deckcipher cache in front of Scryfall. That cache exists for two reasons: it makes a repeated audit of the same deck fast, and it keeps Deckcipher inside Scryfall's request guidelines instead of firing one request per card per visitor. Cards are cached for seven days, and prices are re-checked every twenty four hours. Your browser also keeps a short-lived copy of the last few decks you resolved, so moving between tools on the same list does not re-download a hundred cards.

If a card name does not resolve, Deckcipher does not quietly drop it. It counts as an unresolved line, it is listed back to you, and it lowers the confidence of the result that depends on it.

What is an official rule and what is a Deckcipher read

This is the distinction that matters most, and the tools carry it through to the result pages. Some statements come from the published Commander bracket framework. Others are Deckcipher's interpretation. They are never presented as the same thing.

Deckcipher is not the official source of the bracket framework. Every bracket result prints its rules version so you can tell which revision produced it.

How the bracket estimate is built

The Commander Bracket Calculator runs these steps in order. Nothing here is random, so the same list always produces the same bracket.

1. Parse the list. One card per line. Commander and Deck headers, set codes, collector numbers, and common CSV exports are understood. Sideboard, maybeboard, and Considering sections are excluded from the deck.
2. Count Game Changers. This is the official part. Zero Game Changers starts the deck at Bracket 2. One to three moves it to Bracket 3. Four or more moves it to Bracket 4, because four is already past the Bracket 3 limit.
3. Look for a known two-card win package. If the list holds one, the floor is raised to Bracket 4 and the result names the exact two cards that did it. A deck that can win on the spot from two cards plays at Bracket 4 pressure whatever else it is doing.
4. Apply any pressure you confirmed. Mass land denial, repeated extra turns, wins that regularly threaten before turn 6, or a compact combo plan each raise the floor to Bracket 4. These come from the optional refine questions, and the result says so.
5. Check for a cEDH shell, then stop at Bracket 4. When the card mix looks like cEDH, Deckcipher says what it saw and holds at Bracket 4. It never promotes a deck to Bracket 5 on its own.
6. Consider Bracket 1 only if you asked for it. If you mark the deck theme-first and no pressure signal was found, the estimate can drop to Bracket 1. A detected combo package blocks that drop, because a theme-first label cannot undo a two-card win.
7. Report a confidence level. High means you stated an intent, so nothing is being guessed. Medium means there is card evidence one way or the other. Low means no Game Changers and no pressure signals were found, which is the weakest call the tool makes.

How combo detection works

Deckcipher keeps a short, deliberately conservative list of named two-card win packages. Matching is by card name, so it needs no rules text parsing and cannot be fooled by a reprint or an alternate printing. The current list includes Thassa's Oracle with Demonic Consultation, Thassa's Oracle with Tainted Pact, and Heliod, Sun-Crowned with Walking Ballista, among others.

When a package matches, the result is stated as a fact about your list and the two cards are named. When the tool only finds loose combo pieces without a confirmed package, it says that too, as a warning rather than a ruling.

One detector is shared by the Deck Audit, the Bracket Calculator, and the Bracket 1 to 4 Checkers. That is on purpose. These tools used to disagree with each other about the same deck, and a shared detector is what stops that from happening again. The Combo Finder searches a much broader documented combo database and is the right tool when you want every line in a list rather than the few packages that move a bracket.

The list is short on purpose. A long list would flag decks that cannot actually assemble the pieces, and a false Bracket 4 label costs a player a game they were allowed to play.

How each tool reaches its answer

What these tools deliberately do not do

• They do not rank decks as good or bad. A Bracket 2 deck is not a worse deck than a Bracket 4 deck. It is a deck for a different table.
• They do not tell you which cards to cut. The Deck Audit surfaces cards worth reviewing and leaves the decision to you.
• They do not replace a Rule 0 conversation. The result is meant to make that conversation faster, not to end it.
• They do not put a deck in Bracket 5 from cards alone. Only you can say a deck is built for cEDH.
• They do not use a language model to produce results. Every output is deterministic, so the same list gives the same answer every time.
• They do not model a real four player game. No tool here accounts for interaction, threat assessment, or politics.

Known limits

These are the places where a Deckcipher answer is weakest. They are listed here rather than buried, because knowing where a tool is soft is what makes the rest of it usable.

• Bracket boundaries are fuzzy by design. A deck that reads Bracket 3 with low confidence may well be a Bracket 2 deck to the people who actually play against it.
• Unresolved card names weaken everything downstream. If ten lines fail to resolve, the bracket, the price, and the ratios are all computed from ninety cards. Deckcipher says so, and lowers its confidence, but the result is still thinner.
• The audit model targets Brackets 1 through 4. A genuine cEDH list needs a different evaluation model, which is why detected cEDH shells stop at Bracket 4 and ask you to confirm.
• The two-card package list is short. It will miss real combos. That is a deliberate trade: a missed combo is a conversation you still have, while a false positive is a bracket label you did not earn.
• Archetype detection can be wrong. When it is not confident, it stays quiet rather than guessing, but a misread archetype shifts the land and ramp benchmarks with it.
• Prices are a snapshot. They reflect the printings your list resolved to at the time of the lookup, not what you paid or what your store charges.
• No published accuracy study yet. Deckcipher has not published a validation run against a labeled reference set of decks. Until it does, treat the bracket as an informed estimate rather than a measured one.

Where your decklist goes

The analysis itself runs in your browser. To do it, card names from your list are sent to the card data service so the cards can be looked up, and when you paste a public deck URL that deck is fetched from the site hosting it. Beyond those lookups, your decklist is not stored on a server.

Decks you keep are saved in your browser's local storage on that one device, which is why the current deck card says "This device only". If you create an account and choose to save a deck, that deck and its card list are stored with the account, and you can export or delete everything from My Decks. See the privacy policy and your data for the full detail.

How updates work

The encoded copy of the bracket rules carries a version number, an effective date, and a last reviewed date. All three travel with the result, so a bracket you screenshot today can be traced to the revision that produced it.

The current revision is 2026-02, the February 2026 Commander update, effective February 1, 2026 and last reviewed August 6, 2026. When Wizards of the Coast or the Commander Format Panel publishes a change, the encoded copy is updated and the version is bumped, so a stale answer is visible rather than silent.

Sources

• Commander brackets, Commander Format Panel
• Official Commander format page, Wizards of the Coast
• Scryfall API, the source of all card data
• The current Game Changers list as Deckcipher encodes it

Related

Start with the Commander Deck Audit for a full read of a list, the Commander Bracket Calculator for a bracket estimate, or the full tool directory to find the one that answers your specific question. The site FAQ covers the tools themselves rather than the method behind them.