stoken·games
APEX OBSERVATORY
the data dashboard
ACT I: HOW THE CARDS CHANGED
Analysis 01 · card pool
KEYWORD PREVALENCE OVER TIME
% of each year's new cards (first printings) carrying the keyword. Toggle keywords below and watch Ward get born in 2021 and Banding flatline.
The story: mechanics are born, live, and die. Ward appeared in 2021 and now rides on new cards every year; Banding has not been printed since the 90s. Every line here is a design era.
Analysis 02 · power creep, quantified
CREATURE STAT EFFICIENCY
Median (power+toughness) ÷ mana value of each year's new creatures. Higher = more stats per mana.
The story: creatures got better. A 1994 creature gave you about 1.3 stats per mana; today it is about 1.7, and green gets the best deals. That is power creep, measured.
Analysis 02 · power creep, quantified
COMPLEXITY & LEGENDS
Median rules-text length (chars) and % of new cards that are legendary. The legend-ification of Magic is dramatic.
The story: in 1994, 1 in 10 new cards was legendary. Today it is more than 1 in 4, because Commander became the way most people play, and legends are what Commander runs on.
Analysis 03 · color pie bleed
WHO GETS THE KEYWORD?
Color share of new creatures printed with a keyword, by era. Pick one and watch mechanics drift across the pie.
Analysis 04 · commander staples
STAPLES BY COLOR IDENTITY
Analysis 05 · archetypes by color combo
WHAT EACH COLOR PAIR ACTUALLY BUILDS
Most common deck plans (EDHREC tags in each commander's top 4) among the year's 100 most-built commanders, grouped by color identity. Full merge-and-spread treatment: the locked Flex methodology at field scale, run on the complete 500-commander pull.
ACT II: WHAT THE NUMBERS WHISPER
Data science · association
KEYWORD × COLOR CORRELATION
Phi correlation between "card has keyword" and "card is this color," whole modern pool. Cyan = belongs to that color, red = actively avoids it. Reach living almost entirely in green is the color pie doing its job.
Data science · honest diagnostics
WHAT THE DATA CAN'T SAY
2026 is a partial year: every "drop" in the rightmost point is calendar, not design. Charts end honestly; don't read the last point.
"First printing" = earliest paper English release. Promo pre-releases can pre-date sets by weeks; year-level analysis is unaffected.
Staples reflect the top-500 commander sample, not all of EDHREC; popular commanders pull the numbers toward what popular decks play. Why does that matter? Players look up staples to copy them → the sample IS the meta they're copying → self-reinforcing, which is exactly why we state N on screen.
Legendary% tripled (9.9→28.3). Five whys, condensed: more legends ← Commander is the growth engine ← every product needs face commanders ← legends sell packs ← WotC design incentives follow the format players actually play. The chart is a picture of Commander eating Magic.
Forecasts are naive linear fits on 2013-2025: trend illustration, not prediction. If legendary% really hit the projected ~31% by 2029, a third of all new cards would be legends.
"First printing" = earliest paper English release. Promo pre-releases can pre-date sets by weeks; year-level analysis is unaffected.
Staples reflect the top-500 commander sample, not all of EDHREC; popular commanders pull the numbers toward what popular decks play. Why does that matter? Players look up staples to copy them → the sample IS the meta they're copying → self-reinforcing, which is exactly why we state N on screen.
Legendary% tripled (9.9→28.3). Five whys, condensed: more legends ← Commander is the growth engine ← every product needs face commanders ← legends sell packs ← WotC design incentives follow the format players actually play. The chart is a picture of Commander eating Magic.
Forecasts are naive linear fits on 2013-2025: trend illustration, not prediction. If legendary% really hit the projected ~31% by 2029, a third of all new cards would be legends.
ACT III: YOUR DECK, YOUR DATA
Data science · look up your commander
CARD INSPECTOR
Type any card that shows up in the top-500 commander ecosystem: who plays it, how much, since when, and how combo-prone it is, with a one-click jump to its atlas neighborhood.
Analysis 06 · clustering
ENDLESS ATLAS
Every playable card mapped across six lenses: what it is, what it does, tribes, overall, deckmates, and combos. Interactive and shareable. This is the clustering analysis, grown into a standalone tool.
Analysis 07 · archetypes
TOME OF LEGENDS
Every color identity's game plans, ranked, with the 10 cards that define each plan (by lift vs. the identity's baseline) and how much distinctive tech neighboring plans share. Analysis 05's color-pair table, grown into a standalone tool.
Archive maturing
COMMANDER POPULARITY LIFECYCLE
Rank trajectory of new legends: spike → decay → survivors. Needs monthly EDHREC snapshots to compare. First snapshot banked 2026-07-17; monthly task scheduled. Chartable after ~3 snapshots, strong after 6.
Archive maturing
BAN & GAME-CHANGER LIST IMPACT
Deck counts and prices before/after list announcements. Builds from the weekly price archive + monthly EDHREC snapshots, sliced around each announcement date. Ready whenever the next list change lands.
Design ready
PRECON VALUE TRACKING
Precon MSRP vs. TCG-low sum of contents, over time. Needs per-precon decklists (one-time setup per release) + the weekly price archive. The "which precon is secretly worth it" evergreen.