The grades and win rates come from real game results: 323,114 games of The Hobbit Premier Draft on MTG Arena (11 Aug to 22 Sep 2026), published by 17Lands. Reviewers' opinions sit alongside them, quoted word for word with a link. Some things are our judgment: the coach's pick % in the practice draft, the deck write-ups and the beginner tips.
17Lands and Scryfall
Game results and picks from 17Lands
Files
game_data_public.HOB.PremierDraft.csv.gz (one row per game) draft_data_public.HOB.PremierDraft.csv.gz (one row per pick)
Period
11 Aug to 22 Sep 2026, Premier Draft on MTG Arena
Games
323,114, of which 17Lands users won 56.2%
Draft picks
2.4 million
Thanks to for publishing these files for anyone
to use (17Lands public datasets). We use only these public files.
Card text and images from Scryfall
Cards
193 cards (188 without basic lands), with names, , mana costs and rarity
Served from our own server so they load fast. Following Scryfall's image guidelines, we never crop them and keep the artist and copyright line.
Which cards are in the packs
Pack contents (Play Booster):MTG Scribe's Hobbit Play Booster fact sheet (based on Wizards' fact sheet) and WPN product page. Each pack has 14 cards: 7 commons, 3 uncommons, 1 rare/mythic (mythic about 1 in 8), 1 wildcard slot (usually a common), 1 foil of any rarity, 1 land slot (common dual land or basic, 50/50). The practice draft builds packs this way. The wildcard and foil odds are approximate.
Where the grades come from
Win rate when drawn (): of all games where the card was in the deck and was in the opening hand or drawn, how many were won. The average is 56.2%. That is more than half because 17Lands users are better than the average Arena player. It is the number draft players quote most.
(improvement when drawn): win rate when drawn minus win rate when the card stayed in the library, within the same decks. This shows how much the card itself changes the game.
(average taken at): the average pick number at which Arena drafters took the card, so a card with a high win rate and a late ATA is underrated.
Per-deck win rates: the same metric, counted only in games of one two-color deck (needs 800+ games), which picks out cards that are average overall but win more in one deck.
Cross-check:Limited Grades, an independent site, also grades from 17Lands data. Its top cards match ours: Fíli the Pathfinder A+, then An Unexpected Party, Kíli the Resourceful, Gollum, Riddle Master, Smaug the Magnificent, Thorin, Mountain-king, Beorn the Fierce and Bard's Company.
How a grade is computed
Shrink small samples toward the set average. A card's is blended with 400 games at the set average (56.16% over
323,114 games): adjusted = (wins in hand + 400 × 56.16%) ÷ (games in hand + 400) A card seen in 400 games lands halfway between its own rate and the average. One seen
in tens of thousands barely moves.
Compare it with the rest of the set:z = (adjusted − 55.70%) ÷ 3.67 The two numbers are the and the (population) of the adjusted rate over the set's
188 nonbasic cards.
Not used: the Limited Resources set review (LSV & Marshall) is audio only, and the video reviews we looked at had no transcript we could download.
Reviews written before release are guesses. Where a guess and the data strongly disagree, the card page shows a warning note. Crude Bent Blade is the clearest example. Draftsim rated it 2/10 before release, and it turned out to be one of the best commons in the set.
What we wrote or tuned
These lean on the data and the reviewers above, but they are our opinions. The app says so
where they appear and marks a single figure or verdict with .
The coach's pick % in the practice draft: how it weighs a card's strength
against your colors, curve and creature count. We set the weights by hand. "Same as the coach"
counts how many of your picks match it.
Bots: the same model with randomness. Casual bots also overvalue rares.
Likely decks and signals: how the practice draft reads which colors are open.
Pick timing: on Lookup and All cards, whether a card is worth an early pick
or can come late. How early drafters take it and how well it wins both come from 17Lands.
The cutoff and the advice are ours.
Deck score and the recommended deck: the score out of 10 and the build we
suggest, from 17Lands win rates.
Deck write-ups and "how to play it" tips: written from the reviewers above
and the per-deck win rates.
Learn: each card's role in a deck, why it's good there, what replaces it
and the drills.
Crash course: the whole page, apart from the grades and win rates on its
cards.
Rules refresher: the advice on how to draft and on playing a beginner table.
The rest is measured or quoted. Grades, win rates, games and the average pick () come from 17Lands. Card text and images come from Scryfall, and reviewer quotes are
word for word. The removal tag comes from searching card text, checked by hand.
Caveats
The data is from Arena: best-of-one games, good players, 8-player drafts. Your table is beginners playing in paper. We think simple, efficient cards (, good creatures) matter more there, and cards that only work in a deck built around them matter less.
Win rate when drawn favors cheap cards and card draw a little, since they're in hand more often in games that go well.
The format changed over time (people learned that Goblin Army is the best deck). This is an average over the whole period.
Cardsplain 