Decks & the Lab
Test the deck before the table does.
The Lab opens on any decklist: a goldfish table that deals real hands off the real list, a Report Card that runs 10,000 of them and grades what comes back, a Mana Doctor that reads your colors turn by turn, and a Bench where one swap gets A/B tested before you commit to it.
No account needed to run the Lab on a tournament deck. The Bench runs there as a pure what-if, because the list is not yours to write to.
The Report Card loop
Your list, exactly as you built it
Shuffle, deal seven, mulligan down to the floor
The deck grades itself
A schematic of the loop. The hand count is the simulator's own default, and the right column names what a run reports back, in the Lab's own words.
The Lab
A microscope for a deck you have not sleeved yet
One list, opened up: a table you can actually play out, a graded run over it, a read on your mana, and a bench for testing a single change.
Smart Goldfish
The goldfish table, instrumented
Deal off the real list and the table reads the hand with you. The Mulligan Coach calls Keep, Risky, or Mulligan and shows the three factors behind the call. Cards your lands cannot cast yet glow amber and name the color you are short. London mulligans bottom for real, in the order you pick.
- Every hand carries a seed: replay this exact opening, or load one somebody sent you
- Rewind a draw and take the other line
- Or skip the shuffling and ask for a hand the Coach would keep
Mulligan Coach
A schematic of the call. The Coach reads a real hand, shows all three factors it weighed, and leaves the decision to you.
Report Card & Mana Doctor
The deck grades itself
Hit run and the Lab plays 10,000 opening hands, then writes the verdict as a sentence: how often you keep, how often you flood, how often you make every land drop through the early turns. The Mana Doctor takes the colors, reading your sources against your pips turn by turn and splitting untapped from tapped, so the tapland tax shows up instead of hiding inside a source count.
- One consistency read with its formula on screen: keepability, color reliability, and curve, each weighted and named
- When the read comes back streaky or fragile, it says which of the three is the weak one
Mana Doctor · one color
A schematic of the read. It measures whether you have drawn a source by a given turn, so it reports color access. Mana already on the battlefield is a separate question it does not model.
The Bench
Swap one card. Watch what moves.
Cut a land for a mana rock and the mulligan curve moves in front of you, before the deck ever hits a table. Both versions re-run the same 10,000 hands off one fixed seed, so the delta you read is the swap and not the shuffle. Apply it, or leave it as a what-if and walk away.
- Keepable hands, land drops, flood and screw, color access on curve, and whether the commander still casts on time
- Ask and it picks the cut for you, with a line saying which weakness that cut pays down
- Suggestions come from the diagnosed weakness, your own shelf, and the combos you are one card from
The Bench · one for one
out
a card you cut
in
a card you bench
Re-measured on both sides
A schematic of the swap. Both sides re-run the same hands off one fixed seed, so the delta belongs to the deck and not to the shuffle.
The workbench
Four ways to look at the same list
Text is the editable one, with quantity steppers, a remove, and a find-on-page filter. Stacks, grid, and compact are built for reading. Group by type, mana value, color, or rarity; sort inside each group by name, mana value, price, or quantity; narrow to what you do not own or what you have already wanted.
- Card picker: type for name matches, or press Enter to hand the query to the same engine the site's search runs on
- Printing chooser that floats your own copies to the top, with a Use cheapest jump and finish chips for foil and etched
- Commander slot that seats a pair, with Partner, Partner with, and Friends forever buildable and the union color identity applied to the whole deck
- Import by pasting a decklist, a Moxfield URL, or an Archidekt link, which brings its attribution along with it
Four views, one list
The real axes on the arrange toolbar. Text is the editable one; view, group, sort, and highlights all ride the URL, so an arrangement you like is a link.
Legality Report
Build the illegal thing
The write path makes no legality call at all. Put an off-color card in a Commander deck, run five copies in Modern, seat a partner pair the rules do not allow. The deck saves.
The Legality Report is computed on read, not on write, and it is the only surface that judges. It names what is banned, what the format does not allow, what is over the copy limit, what falls outside the color identity, an ineligible commander, and an illegal pairing across all five Partner Mechanics, including the two the builder will not seat for you. Quantities roll up by Oracle Card before anything is checked, so four of one printing and four of another read as eight.
The report also says how much of the deck it checked. Commander and the six constructed formats get the full per-card check. Brawl and Oathbreaker get the size and singleton read, Limited gets the deck-size floor and nothing else, and the badge says so. Casual is skipped and the badge hides. A card whose format row is missing gets flagged unverified.
Combo Radar
The combos you hold, and the ones one card away
Your list is matched against Commander Spellbook's catalog by containment. The radar tells you which documented combos the deck already holds, and which sit exactly one card away, filtered to pieces that resolve to real cards and fit inside your color identity. It reads a combo list; it does not read card text and infer a loop, so it will only ever show you combos somebody has written down.
- Near misses dedupe to one combo per missing card, so a single missing staple cannot flood the tier
- A piece that has not resolved to a real card counts as missing, so a truncated list can never read as complete
Combo Radar · two tiers
A schematic of the read. Your list is matched against Commander Spellbook's catalog by containment; the radar reads a combo list, it does not read card text and infer a loop.
The read
What the deck is actually made of
The synergy lens drills the list four levels down, from a lens like what the deck is built around, to a facet, to a classifier tag, to the cards carrying it, with facets ranked by how many distinct cards land in each. Beside it, the consistency read and the ownership overlay answer the two questions that follow: how reliable is this, and what do I already own that would fix it.
- The collection overlay answers a mana diagnosis with cards off your own shelf, untapped lands first
- Share a deck by link and it stays out of search, with the archetype, the stats, and a Get these cards button on the shared page
- That button builds a list of exactly what the deck needs and you do not own, at Oracle Card grain
Synergy lens · four levels down
- lenswhat the deck is built around
- faceta related group of tags
- taga classifier the card carries
- cardsthe cards carrying it
A schematic of the drill. The five lenses are strategy, function, resource, role, and audience, and facets rank by how many distinct cards land in them.
Every deck you build is free, and so is every deck you share by link. Keeping a deck private is the paid side, and there is no self-serve purchase for it yet, so today only comped accounts have it. The pricing page carries that boundary and every other one.
Archetype classification
Named on the way in
Every list is scored against each strategy's recipe on three axes: how much of the deck carries the right classifier tags, how the curve fits, and which colors it runs. The label is explainable as a share of the deck, because the classifier is a set of declared recipes and there is no model to interrogate. When the top score is low, or two plans fit the list equally well, it says nothing at all.
- Runs at import, and again after every card you add or cut
- The badge carries its confidence, and it simply does not render when the classifier abstains
One decklist, read as concepts
- Removalcounted
- Rampcounted
- Card drawcounted
- Board wipescounted
A schematic of the read. The full classifier vocabulary is public; browse every concept it reads.
Tournament decks
The metagame, as data you can filter
Competitive decklists land from TopDeck.gg events daily, faceted by format, commander identity, and event window, and sortable by recency, top finish, or win rate. If you play cEDH, this is the deepest data on the site. If you play Standard, be warned: the events we ingest barely cover it.
- Open to everyone, no account, and the Lab opens on any of these lists as a what-if
- The cEDH leaderboard ranks on win rate behind a games-played gate, so a one-game sample cannot post a perfect record
- Copying a tournament list into a deck of your own is not built yet
cEDH metagame · top archetypes
- Simic Tempo3.7%
- Esper Tempo1.8%
- Witch-Maw Tempo1.5%
- Jeskai Tempo1.4%
Live from tournament results over the last 30 days. The classifier placed 18% of 5,362 decks, and shares divide by every deck in the window, so they sum to that coverage. As of 2026-08-27.
From owning to building
The bridge a spreadsheet never had
Because your collection, the market, and the metagame all key to the same cards, the jump from what you own to what you could build is one step.
Take a list apart
Bring a deck in from Moxfield, Archidekt, or a paste, and put it under the Lab. Or read what the format is actually doing first.