Commander Brackets Explained

The labels we use are the official ones: 1 Exhibition, 2 Core, 3 Upgraded, 4 Optimized, 5 cEDH. What separates them is not how fast a deck kills. It is which categories of card are on the table: Game Changers, mass land denial, extra turns, and two-card combos.

Brackets 1 and 2 allow none of those. Bracket 3 opens the door to three Game Changers and a single extra-turn effect. Bracket 4 removes the restrictions entirely — anything Commander-legal is fair game. Bracket 5 shares that same card pool; the difference between 4 and 5 is not permission but density and intent, which is why our test for 5 is a conjunction of several measurable facts rather than a single trigger.

Our deck analyzer starts every list at Bracket 2 and evaluates these in order:

1. One or more Game Changers → Bracket 3. Brackets 1–2 allow zero, so the first one is a hard line.
2. One or more cards whose oracle text contains extra turn → at least Bracket 3.
3. More than three Game Changers → Bracket 4. Three is the Bracket 3 ceiling; the fourth breaks it.
4. Any mass land denial → at least Bracket 4.
5. Any two-card combo in the deck → at least Bracket 4.
6. Three or more extra-turn cards → at least Bracket 4, on the reasoning that a chainable Time Warp package is a different card than a single copy.
7. Bracket 5 requires all of: four or more Game Changers, and three or more fast-mana cards, and either two or more two-card combos or five or more tutors.

Each rule that fires writes a line into the reason list printed under your bracket on the power level page, naming the cards responsible. If none fire you get the explicit "no Game Changers, mass land denial, extra turns, or two-card combos detected" line rather than a silent 2.

Worked example. A Meren list running Demonic Tutor and Survival of the Fittest has two Game Changers. Rule 1 fires and sets Bracket 3; rule 3 does not, because two is not more than three. Now add Mikaeus, the Unhallowed and Triskelion. Commander Spellbook returns that pair as a two-card combo, rule 5 fires, and the deck lands at Bracket 4 with two reasons listed. Note what did the work: the Game Changers only got you to 3. Cutting Triskelion alone — one card out of a hundred — puts the deck back at Bracket 3.

The thing that surprises people most: fast mana never moves the bracket by itself. Sol Ring in an otherwise clean deck is still Bracket 2. Fast mana is counted (seventeen cards, including Mana Vault, Grim Monolith, Ancient Tomb, Lotus Petal, Dark Ritual and Jeska's Will) but it only matters inside the Bracket 5 conjunction and in the Speed subscore, where each copy adds 0.9.

We do not maintain our own list. The set is fetched from Scryfall's is:gamechanger tagging of the official Wizards list, cached for 24 hours. The only reason the embedded snapshot in our source (53 cards, dated 2026-07-17) ever gets used is if Scryfall is unreachable or returns nothing. When Wizards revises the list, brackets across the site move without a code change.

Names are normalized before matching: lowercased, and split on the // separator so only the front face is compared. A modal double-faced or split printing still matches its entry.

One overlap is worth internalizing, because it catches people out. Several Game Changers are also fast mana: Ancient Tomb, Chrome Mox, Grim Monolith, Lion's Eye Diamond, Mana Vault, Mox Diamond, and Jeska's Will. Sleeving a single Ancient Tomb into a colorless-ish deck puts you at Bracket 3 immediately and contributes to the cEDH fast-mana threshold. That is one land.

Our fast-mana list deliberately omits Mana Crypt and Jeweled Lotus. Both are Commander-banned, and counting cards you cannot legally play would inflate every historical decklist we import.

Land denial is detected from oracle text rather than a curated list, because the effect has too many wordings to enumerate. Twelve patterns cover the families: destroy-all (Armageddon), return-all, untap denial via lands don't untap and can't untap more than [^.]*(land|permanent) — which is how Winter Orb and Static Orb read, and yes, an orb that freezes every permanent is denying lands too — plus the Pox family of forced sacrifice.

That last family is where the interesting decision lives. The pattern is a subject of (each|target|that|all) (player|opponent) followed by sacrifices?( \w+){0,4} lands?. The ( \w+){0,4} is doing all the work: it spans at most four word tokens, and a comma is not a word character.

So it matches Smallpox — "…then sacrifices a land" — and it stops dead on "sacrifices an artifact, creature, or land." That second clause is Braids, Cabal Minion, and excluding her is deliberate, not an oversight. Braids makes each opponent give up a permanent every upkeep and lands are on the menu, but the opponent picks. Someone holding a spare Treasure or a chump blocker never loses a land to her. Armageddon takes the choice away. A forced land sacrifice is a resource lock; a menu is a tax you pay in whatever currency you happen to have most of.

Braids still lands the deck at Bracket 3 — she is a Game Changer. Different rule, different reason, and the reason list says which. The mirror-image case is Wave of Vitriol, which says "sacrifices all" and therefore removes the choice; a separate pattern with the literal word all catches it. A third pattern handles the Balance and Cataclysm shape, where each player keeps a few lands and sacrifices the rest.

Honest limitation: this is text matching, so it answers "does the card say the thing," not "is this an MLD deck." A deck that grinds lands away through a repeatable land-only edict never trips the pattern until something says "each player sacrifices a land." And an Armageddon you have built to break scores identically to an Armageddon someone jammed because it was in the box.

Combos are not guessed from card text. Your exact decklist is checked against Commander Spellbook, and we keep only results of two cards or fewer. A three-piece line you can assemble with a tutor chain does not move your bracket; a two-card line does. That is the official rule and we implement it literally — which also means our combo detection inherits Spellbook's coverage. A genuinely novel two-card loop that nobody has submitted is invisible to us.

Tutor counting is narrower than people expect, and it matters because tutors are one leg of the Bracket 5 test. We match search your library for a card and searches that name a card type. Demonic Tutor and Vampiric Tutor count. Goblin Matron, which searches for "a Goblin card," does not — a tribal tutor finds one of thirty interchangeable creatures, not the exact answer a cEDH list needs. Land searches are classified as ramp instead, so Rampant Growth never inflates your tutor count.

Bracket 5 is a conjunction, not a threshold, and it is intentionally hard to trip. A deck with eight Game Changers, nine tutors, and two fast-mana cards is Bracket 4, not 5 — it fails the fast-mana leg by one card. We would rather under-call cEDH than hand a Bracket 5 label to a strong casual deck, because 5 is a claim about which table you belong at, not a score.

Because brackets gate cards rather than outcomes, speed is almost invisible to the system.

Take a Krenko goblins deck: 34 lands, Sol Ring, a low curve, Goblin Matron and Goblin Recruiter to find threats, no Game Changers, no land destruction, no two-card combo in the database. Bracket 2. It can kill a table on turn six.

Now take a five-color enchantment pile whose only notable cards are Rhystic Study and Smothering Tithe. Two Game Changers, Bracket 3. It wins on turn twelve if nobody interferes.

The higher-bracket deck loses to the lower-bracket deck most games, and both readings are correct. The bracket describes what is in the deck; the 0–10 power score describes how fast it is. Goblins posts a high Speed subscore off its low curve and Sol Ring while sitting at Bracket 2. Read both, or browse public decks to calibrate against builds you recognize.

The same logic explains the case people push back on hardest: one Armageddon in an otherwise precon-level upgrade reads Bracket 4 on its own. We do not soften that, and socially it is the right answer — the card that can end a game on turn five is precisely the card everyone deserves to hear about before shuffling.

The bracket badge that updates while you edit in the deck builder is a lighter scan, not the full engine, and it will not always agree with the bracket saved on your deck page.

Known divergences, all deliberate: the live scan starts at Bracket 1 rather than 2, so an empty or signal-free deck reads Exhibition. It uses the embedded Game Changers snapshot instead of the live Scryfall sync. It treats two Game Changers as Bracket 4, where the full engine allows three at Bracket 3. It cannot call Commander Spellbook on every keystroke, so it sees no combos at all. And its fast-mana set is a different, shorter list where any single hit is a Bracket 4 signal. Net effect: the builder over-reads Game Changer and fast-mana decks and under-reads combo decks. The saved analysis on your deck page is the authority — recompute it when you are done editing.

The number is the least useful thing you can say. Brackets are a permissions list, so the honest version of a rule zero conversation is a list of permissions you are exercising:

• Name your Game Changers out loud. "I have Rhystic Study and Smothering Tithe" tells the table more in six words than "it's a 3."
• Disclose two-card combos and roughly how quickly you assemble them. "There's a two-card line but no tutors for it" is a real and useful distinction.
• Disclose mass land denial every single time, even if you rarely draw it. This is the one nobody forgives you for omitting.
• State your realistic kill turn. Brackets do not measure it and it is what everyone actually wants to know.
• When you are genuinely between two brackets, quote the higher one. Overselling costs you a slightly slower game; underselling costs you the table's trust for the rest of the night.

One thing we cannot compute for you: Bracket 1. Our deck analyzer starts every list at 2 and only moves up, so it never outputs Exhibition. There is no card check that proves a deck is not trying to win — that is a promise you make about how you will pilot it, not a property of the 99. If you built an Exhibition deck, you will have to say so yourself.

To see the rules applied to a real list, run one through the power level analyzer, which prints every triggered rule and the cards behind it. The weighting behind the 0–10 score is documented on the methodology page.