About the MTG Arena tools

Legality

A pasted deck is checked against Standard, Pioneer, Historic, Timeless, Brawl and Standard Brawl in one go: deck size, copy limits (with cards like "any number" exceptions), sideboard size, banned and restricted cards, commander rules, and whether every card exists on Arena. A companion's deckbuilding condition is checked too, and the deck list names the cards that break it (Zirda's "has an activated ability" test reads the rules text, so it can be fooled). Legality comes from Scryfall's per-format flags in the card data of 18 Oct 2026. Alchemy is not supported, and a card the index cannot find is reported with "did you mean" suggestions instead of being guessed.

Curve and mana

The curve counts nonland cards by mana value (the front face for double-faced and adventure cards). Colour checks compare the coloured sources in the deck with Frank Karsten's targets: the number of sources needed to cast a spell on curve about 90% of the time. For a 60-card deck:

Coloured sources needed, by coloured pips (rows) and mana value (columns)
Pips	MV 1	MV 2	MV 3	MV 4	MV 5	MV 6+
1	14	13	12	10	9	9
2	—	21	18	16	15	14
3	—	—	23	21	19	18
4	—	—	—	24	23	21

Dual lands count for each colour they make. A fetch land counts for what it can actually find in this deck: "an Island or Swamp card" finds a Blood Crypt as well as a Swamp, and finds nothing at all in a deck without Islands or Swamps. Modal double-faced cards with a land back count as colour sources, and as half a land when the power tier judges the land count. Lands that always enter tapped are listed separately because they cost tempo.

Card roles

Roles (removal, counterspell, sweeper, card draw, ramp, tutor, threat, protection, lifegain) come from patterns in each card's rules text, with reminder text removed. Every tag names the rule that matched, so you can see why a card was tagged and disagree with it. The patterns are deliberately simple; an unusual card can be missed or mis-tagged. Treasure counts as ramp only when you can rely on it: the spell makes it, the permanent makes it as it enters, or an ability you activate makes it. Treasure from combat damage, attacks or saga chapters is a bonus, not ramp.

Cards a casual table tends to dislike (extra turns, mass land destruction, lock pieces, mass discard and the Commander "Game Changers") are marked feel-bad in the deck list; milder ones such as counterspells, tutors and hand attack are listed as scored down. The list comes from grindstone's casual-etiquette rule file and only matters for the Friendly casual upgrade version.

Power tier

Each part of the deck gets a score from 0 to 100 with the reasons behind it, and the weighted total sets the tier. The deck's plan (aggro, midrange, control or ramp) is inferred from its curve, creature count, interaction and ramp, and the targets follow that plan. A ramp plan needs something to ramp into: at least 8 ramp cards and an average mana value of about 2.8 or more, or 6 or more cards at 5+ mana.

Constructed (60-card): what is scored
Part	Weight	What it looks at
Mana base	20%	Coloured sources against the Karsten targets, land count against the archetype, and lands that enter tapped.
Speed / curve	20%	Average mana value, the share of cheap spells and the number of expensive ones, compared with what the archetype wants.
Interaction	15%	How many removal spells, counterspells and sweepers the deck runs, and how many of them are cheap.
Card quality	25%	With meta lists: how many of your nonland cards the meta decks also play, and how widely. Without: an estimate from popularity, rarity and age (low confidence).
Consistency	20%	Keepable openers, land drops and castable colours from the hand simulation, plus how many cards are 4-ofs.

Tiers
Tier	Label	From score
1	Jank / Precon	0
2	Casual	30
3	Focused	50
4	Optimized	70
5	Competitive	85

Targets by plan (60 cards)
Plan	Avg MV	Lands	Interaction
Aggro	2.1	21	6
Midrange	2.8	24	9
Control	3.0	26	13
Ramp	3.4	25	6

Brawl and Standard Brawl use a separate rubric: mana base (15%), commander synergy (20%), ramp (15%), card draw (15%), interaction (15%) and card quality (20%). Commander synergy checks how much of the deck supports the themes and creature types named in the commander's text.

Confidence. Without meta lists, card quality can only be estimated from how popular a card is on Scryfall (its EDHREC rank), its rarity and its age. That estimate is marked low confidence and the analyzer says so in a banner. Add a few current lists on the Meta page and the tier becomes much more meaningful.

Hands and mulligans

Sample hands and the simulation shuffle with a seeded generator, so the same seed always deals the same cards and a shared link replays the same hand. The simulation plays out thousands of openers with the London mulligan (down to five cards, bottoming lands when flooded, else the most expensive spell). A seven is "keepable" when it has two to five lands and a spell of mana value 3 or less that those lands can cast. The optional best-of-one smoothing is an approximation of Arena's hand smoother: of two shuffled openers it keeps the one whose land count is closer to what the deck's land ratio predicts. It also reports land drops on curve, the chance to cast each card on curve, and any "card or role by turn N" checks you add, next to the exact hypergeometric odds for comparison.

Wildcards and upgrades

Wildcard cost counts every copy in the main deck, sideboard, commander and companion (at most four per card), each at its cheapest Arena rarity; basic lands are free. Whetstone cannot see your collection, so you can enter the wildcards you have to see what is still missing.

When the card data has paper prices (Scryfall's, at each card's cheapest printing), the wildcard section also shows what the same cards cost on paper. Arena has no cash prices, so this is only a guide to card value.

The upgrade planner builds one of grindstone's three versions of your deck. Same deck, best cards keeps every card's job, card type, colours and tribe, at the same or one less mana. Most competitive goes for strength and also upgrades the lands. Friendly casual swaps out feel-bad cards and never asks for a rarer wildcard than the card it replaces (unless you prefer mythics and rares). Cards that have to go (banned, not on Arena, breaking the companion) are replaced in every version.

How much of the deck may change follows the version (about 25% for the same deck, no limit for the most competitive, about 15% for friendly casual) unless you pick about 7%, 15% or 30%, or no limit, yourself. You can also lock cards you want to keep and set a wildcard budget that is never exceeded. "Prefer mythics and rares" leans toward premium cards when there is a choice and is remembered in this browser.

The planner avoids downgrades. In a low-curve deck (average mana value 2.4 or less) a swap never costs more mana, and an instant is only replaced by an instant or a card with flash. Without meta lists the new card has to beat the old one clearly on the popularity, rarity and age estimate. Those swaps are not ticked for you, and each reason says what the estimate preferred. "More widely played" appears only when the new card really is. Every applied swap can be undone.

Meta lists

Whetstone does not collect decklists from sites whose terms forbid it. Curated lists are read by hand from decklists Wizards of the Coast publishes on magic.gg, copied as card names and counts only, and added to the repository with the link and date of the article they came from. They are refreshed about once a month and checked against the card data every day, so a ban or a rotation shows up as a warning. You can also keep your own lists in your browser and import or export them as grindstone snapshot JSON. The comparison reports the closest list by shared nonland cards, the core cards your deck is missing, and your unusual choices.

Limited ratings

Limited pages are built from 17Lands' public game and draft datasets (licensed CC BY 4.0), downloaded and summarised once a day when a file changes. For each card: games-in-hand, opening-hand, drawn and never-seen win rates, improvement when drawn, average pick when last seen and when taken, and colour-pair win rates. 17Lands users are stronger than the average Arena player, so win rates run high; the differences between cards are what matters. Rows with fewer than 500 games in hand are marked as low sample, and while you sort by a win rate they are listed after all the others, so a lucky handful of games never tops the table.

Sources and privacy

Card data, paper prices and images: Scryfall, from its daily bulk data. The trimmed card index is an internal part of the tools, not a dataset. Card images load straight from Scryfall's image server (cards.scryfall.io), shown whole with the artist credit.
Limited data: 17Lands public datasets, CC BY 4.0, credited at the top of every Limited page.
Curated meta lists: decklists published by Wizards of the Coast on magic.gg, linked from each list on the Meta page.
Analysis engine: grindstone-mcp, MIT licence.
Your browser keeps a copy of the card data in IndexedDB (database whetstone-mtga), replaced when the data changes, so the tools don't download it on every visit. Saved meta lists and tool settings (the meta comparison choice, the wildcards you own and whether you prefer mythics and rares) stay in local storage under whetstone:mtga:. Clearing this site's data removes both.
The analyzer only contacts api.scryfall.com when you tick "Ask Scryfall about names this site doesn't know", and then only with the names it could not match. See the site's privacy page.

Fan Content Policy

Whetstone is unofficial Fan Content permitted under the Fan Content Policy. Not approved/endorsed by Wizards. Portions of the materials used are property of Wizards of the Coast. ©Wizards of the Coast LLC. Magic: The Gathering and MTG Arena are trademarks of Wizards of the Coast LLC. Whetstone is free and has no ads.

Card counts on this build: 16,284 Arena cards, 5,177 legal in Standard.