Every custom card designer eventually hears the same review: "cool, but broken." Balance is the craft half of card design — the part that turns a flavor idea into something your playgroup actually lets you play. The good news: Magic has thirty years of public data about what fair looks like. This guide compresses it into a working method.
Start from a real-card baseline
Never cost a card from scratch. Find the closest printed card and diff against it.
A 2/2 creature for {1}{G} with a small upside is the modern baseline ("Grizzly Bears plus"). A 3/3 for {3} is unexciting; a 3/3 for {2} needs a color with aggressive identity or a small drawback. Removal tells the same story: instant-speed unconditional creature destruction is worth around {2}{B} to {3}{B} depending on era; conditional removal ("destroy target creature with power 4 or greater") earns a discount.
The diff method: take the printed card, list what your version adds and removes, and price each delta. Added a keyword? That's roughly half a mana to a full mana depending on the keyword (deathtouch is cheap on a small body, expensive on a big one). Added a triggered ability that draws cards? Card advantage is the most expensive thing in the game — a full mana or more.
When you import a real card via Scryfall into PipGlyph as a remix starting point, you're doing the diff method with the baseline already on the canvas.
Respect the color pie
Each color pays different prices for the same effect, and some effects are simply off-menu:
- White gets efficient small creatures, board wipes, taxing effects, and removal-by-exile. It does NOT get raw card draw — white card advantage looks like "when a creature dies, draw."
- Blue gets counterspells, card draw, bounce, and copy effects. Blue creature stats are deliberately below rate.
- Black gets the best single-target removal and recursion, paying life and sacrificing creatures as currency. "Anything, at a price" is the brand.
- Red gets damage, haste, impulse draw (exile the top card, play it this turn), and artifact destruction. Permanent answers to enchantments are off-pie.
- Green gets the biggest creatures per mana, ramp, fight-based removal, and naturalize effects. Green should NOT get unconditional "destroy target creature."
The fastest way to make a custom card feel wrong is a color-pie break: a green counterspell, a red enchantment exile. Breaks can be deliberate — but they should be rare, loud, and mythic.
Downsides, drawbacks, and the Necropotence rule
Drawback design is where customs most often go broken, because of an asymmetry: a drawback that doesn't matter is just flavor text. Magic history's most banned cards are "drawback" cards where the drawback turned out to be a benefit in the right shell (Necropotence's "skip your draw step" reads disastrous; with the engine attached it was the best draw engine ever printed).
Stress-test every drawback with: which deck WANTS this cost? "Sacrifice a creature" is a cost in a vacuum and an enabler in an aristocrats deck. "Discard a card" powers madness and reanimator. If the answer is "lots of decks want to pay this," price the card as if the drawback were an upside.
Rarity is a balance tool
Rarity isn't prestige — it's a complexity and impact budget:
- Common: one ability, French-vanilla territory, fair rate. Commons define limited play; keep them honest.
- Uncommon: two abilities or one build-around hook. Signpost designs live here.
- Rare: build-arounds, engines, and rate-pushers with real constraints.
- Mythic: splashy, game-warping, narrow. Mythic should be the card the table talks about, not the card every deck plays four of.
If your design needs three paragraphs, it isn't a common. PipGlyph's editor lets you set rarity per card, so a custom expansion can hold a realistic curve (roughly 10 commons per 3 uncommons per rare in printed set skeletons).
The five playtest questions
Before publishing, run the card through these:
- The four-of test — would a competitive deck run four copies? If it's an auto-four-of in multiple archetypes, it's pushed.
- The turn-sequence test — write the actual turns: if your card hits the table on turn 3, what does the opponent's turn 3 look like across from it?
- The blank test — if the opponent has no answer, how fast does this win? Cards that win unanswered within two turns need answers to exist at the same speed.
- The floor/ceiling test — what's the card's worst reasonable mode? Cards with a high floor AND high ceiling are the broken ones; great designs buy ceiling by lowering the floor.
- The fun test — does the loser of the exchange get to make a decision? Effects that remove the opponent's agency (mass land destruction, extra-turn chains) need to be priced like mythics even when their rate looks fair.
PipGlyph's AI assistant includes a balance check that reads your card and returns a risk assessment with specific cost suggestions — useful as a second opinion before you publish to the gallery.
Balance is a conversation
Publish the card, watch how it plays, and iterate — every card stays editable after publishing, and the community's likes and comments on the gallery are a live playtest panel. The printed game ships mistakes too; the difference for a custom designer is you can fix yours the same evening.
FAQ
Is there a formula for creature costs?
A rough modern baseline: power + toughness ≈ 2 × converted cost, adjusted by color and keywords. A {2}{G} creature can be a 3/3; add trample and it's a 3/2 or costs {3}{G}. Formulas get you to the neighborhood; playtesting parks the car.
Should custom cards be balanced against modern power level or older eras?
Pick a deliberate reference point and state it in the card's tags or set description. A card that's fair in 2003 Magic is unplayable next to 2024 designs — neither is wrong, but mixing eras in one set without intent reads as inconsistency.
How do I balance a card for Commander specifically?
Multiply every "once per game" effect by three (you'll see it more often than in 1v1), discount aggro stats, and price card advantage up — 40-life games are attrition games. The fun test matters triple at a four-player table.