The question is how the card performs in the deck playing it. A high win rate on its own doesn’t tell the whole story.
Compare with the deck.
Δ = card GIH win rate − deck baseline win rate.
If a card’s GIH win rate is 60% and the deck baseline is 56%, that is +4 percentage points. It belongs in Top cards, provided it meets the selected sample minimum.
GIH means games in hand: games where the card was in the opening hand or drawn later. The saved public data is aggregated using card-copy counts.
Change the deck context in Today’s Data to see how the result changes. Small samples, splashes and the deck’s wider plan still matter.
Calculated from unrounded values. Default minimum: 200 in-hand copies; you can change it in the filters.
Numbers and commentary answer different questions. Bands and ranks use measured data. Draft Atlas reviews interpret the available creator commentary; they do not claim that three creators agree or assign a consensus grade. The original band framework is credited to Sierkovitz in Guidance Sources.
REALITY FRACTURE / MAGIC NUMBERS #196
Right card. Right deck.
A searchable draft companion to Sierkovitz’s set review. Find a card’s best home, then check the evidence.
Atlas live ranks use this deck context, band and Δ; they are not a published creator pick order.
Data source, freshness & calculations
What “LIVE” means
This tab checks the public files on 17Lands when opened and when you select “Check for updates”. The displayed game date comes from the actual records. 17Lands publishes these reusable files in batches; there is no real-time or same-day feed here.
When a newer release appears, select “Load new release”. The compressed game and draft files are processed in the background and saved in this browser. A refresh replaces the table only after both files finish successfully. Closing the page or cancelling leaves the last complete dataset intact.
Read the numbers
GIH = opening-hand + drawn copies; GP = maindeck copies. OH, GD and GNS mean opening hand, drawn later and not seen. Each win rate weights games by the relevant number of copies. GNS subtracts opening, drawn and tutored copies from the deck, floored at zero. IIH = GIH WR − GNS WR.
Δ = GIH WR − the selected group’s game win rate. The bands follow the episode’s key below. “Unrated” means insufficient sample. Bands use unrounded values; displayed figures are rounded.
Decks & draft picks
Each colour-pair group uses the public file’s main colours, including recorded splashes. “All decks” also includes mono-colour and three-or-more-colour decks. The card colour filter does not change the baseline.
ALSA and ATA always use all draft records. Picks are numbered from 1. ALSA retains the last sighting of each card instance in a pack, deduplicating the eight-seat wheel within each draft. ATA averages the pick at which the card was selected, including recorded secondary picks.
Public-release timing, data cleaning and the selected sample can differ from 17Lands’ own current dashboard. These observations do not measure causal card strength. Copy counts are not independent observations or confidence intervals. This atlas is not affiliated with or endorsed by 17Lands.
GUIDANCE SOURCES
Keep the evidence close.
The original discussions, the data, and how the two fit together.
* BO3 bottom: limited coverage. In the saved check, only Mindseeker Oculus had a reported GIH rate. This figure is not a reliable benchmark for the group.
What these averages mean
GIH WR is games-in-hand win rate. These figures are the equally weighted mean of cards with a published GIH rate in each group, using 17Lands’ All Time filter for this set. Eligible cards differ between groups. They are card averages, not the players’ overall game or match win rates; BO3 card statistics are also measured per game.
Top, middle and bottom are 17Lands’ performance groups, based on results in at least two of the previous three sets. Most users are outside these groups. The panel is a snapshot checked on the date above; it does not represent only games played today. Select a figure to view the latest source. Card-list filters do not change these set-wide averages.
THE METHOD
A card is only as good as its home.
Draft Atlas is an independent Limited project by Jake Peter. The starting idea — comparing a card’s in-hand win rate with its deck’s baseline — comes from Sierkovitz’s Magic Numbers #196. Today’s Data combines separately calculated public statistics with editorial summaries and attributed notes from Sierkovitz, Chord of Calls and Lords of Limited.
Read the bands
Top cards
Δ ≥ +3 pp
Good cards
+1 ≤ Δ < +3 pp
Filler
−1 ≤ Δ < +1 pp
You want very few
−3 ≤ Δ < −1 pp
Almost never
Δ < −3 pp
Unrated
Insufficient episode evidence
The cutoffs follow his 8:08–11:51 explanation. The labels match the source key. Episode ranks follow the full-set best-reported-fit order: band, then reported Δ, then card name. Cards without an episode assessment follow alphabetically; their numbers indicate list position only. Boundary choices are explicit above; rounded spoken values can move a card across a boundary.
Two kinds of evidence
Spoken number means an approximate point estimate was extracted from the episode’s narration. It is not a fresh 17Lands query or a transcription of every chart.
Commentary assessment maps a clear verbal assessment onto the named bands: great → Top cards, good → Good cards, fine → Filler, weak → You want very few, avoid → Almost never. These are interpretations of the discussion, not measured deltas.
Specific numerical results take priority over a general qualitative statement. Unclear figures and unmentioned cards stay unrated. Missing data stays unrated.
Use deck context first
“Best reported fit” shows the highest supported band among the available colour pairs. It is not an overall pick order, and the highest reported home may not be the card’s true best home if other results are missing.
Choose your colour pair for a useful draft shortlist. Supported splashes are labelled separately. A strong splash result does not prove that your mana can support the card.
The episode’s pair groups include light splashes. GIH minus baseline is a comparison, not a causal effect or the “improvement when drawn” metric. Sample counts and confidence intervals were not extracted.
Coverage & downloads
One entry per card name. Basic lands, alternate art, Commander-only and Alchemy-only cards are excluded. Special Guests are indexed separately and were not rated in this extraction.
Lords of Limited ↗ · Ben Werne and Ethan Saks. Only verified card-specific commentary is summarised; a missing note is a coverage gap.
Sierkovitz — Magic Numbers #196 ↗ · source of the original colour-pair analysis and episode notes. Card-specific source links appear below each Draft Atlas review.
17Lands card data ↗ · source of the statistics discussed by Sierkovitz. The episode views preserve the original analysis. The Today’s data tab independently aggregates the reusable 17Lands public datasets and checks for new releases.
