MTG Deckbuilding Reference

Reputable, cited deckbuilding math — the standing reference our deck-build pipeline applies. Sources: Frank Karsten, Reid Duke, WotC, EDHREC/edhtop16, MTG Arena Zone.

Distilled from reputable sources (Frank Karsten, Reid Duke, WotC, EDHREC/edhtop16, MTG Arena Zone) 2026-07-27. Builder + critic agents: read this and apply the numbers mechanically — do not vibe a mana base. Load-bearing hard numbers are marked ★.

1. MANA BASE (Karsten math, ~90% cast-on-curve)

Land count — use the formula, then round

• ★ 60-card: lands = 19.6 + 1.9 × avgMV − 0.28 × (cheap ramp/draw pieces) (avgMV = avg mana value of NONLAND cards; “cheap” = MV ≤ 2).
• ★ 99-card Commander: lands = 31.4 + 3.1 × avgMV − 0.28 × (cheap ramp/draw).
• Sanity bands — 60-card: aggro 22–23 · midrange 24–25 · control/ramp 25–26. 99-card: low-curve/cEDH 30–33 · midrange ~36–38 · high-curve 40–42.
• Rules: cut ~1 land per 3–4 cheap (MV≤2) ramp/draw pieces; each +1 avgMV ≈ +1.9 lands (60) / +3.1 (99); MDFC land/spell counts as ~0.38 land (0.74 if untapped). Bias UP when unsure — screw loses harder than flood.

Colored-source requirements — the actionable core

★ 60-card, sources of a color needed to cast on-curve (~90%):

• ★ Quick heuristic (60-card): single pip ≈ 14 · double pip ≈ 18–20 · triple pip ≈ 21–23. Requirement drops ~1 per turn you delay the cast.
• 99-card Commander ≈ 1.6 × the 60-card number. Single pip by turn: T1 23 · T2 21 · T3 19 · T4 18 · T5 16. Double pip ≈ 30–32; triple ≈ 37+. In 3+ colors you can’t hit 30+ off-color, so prioritize the commander’s colors / early plays and accept greedy splashes cast late; universal fixers (Command Tower, Arcane Signet, Signets) count toward every color.
• If you can’t hit the source count for a card, add sources or cut the card — this kills greedy manabases.

Source-counting weights (when tallying colored sources)

• Land producing the color = 1 · Fetchland = 1 per color it can fetch · Mana rock (Signet/Talisman) = ~0.75 · Mana dork (Birds/Llanowar) = ~0.5 · land-fetch ramp (Cultivate) ≈ 1. Count cheap rocks/dorks/land-ramp in the land-reduction (−0.28) term too.

Taplands / enters-tapped — ★ minimize

• Each tapland = a real tempo cost (a missed on-curve turn). Afford by archetype: aggro/tempo 0–3 (basically none) · midrange ~4–6 · control/ramp 8+ (racing matters less).
• In 2-color decks, basics are the best untapped fixing — don’t reach for “value taplands” (Temple scry, Thornwood gain-1) when you barely need the fixing. Count UNTAPPED sources for on-curve plays.

Sources: Karsten “How Many Lands”/“How Many Sources 2022” (TCGPlayer/CFB); reproductions peasant-magic.com, metamages.com/mana, scrollvault.net/guides/mana-bases.html; WotC magic.wizards.com/en/news/feature/building-mana-base-2014-11-24.

2. DECKBUILDING FUNDAMENTALS (Reid Duke / WotC)

• Pick a role and build to it: aggro (fast clock, cheap threats, low curve) · midrange (efficient grinding threats) · control (interaction + card adv + finisher) · combo (assemble a winning interaction). A pile of “good cards” with no shared plan is the #1 bad-deck signature.
• Who’s the beatdown: each matchup has an aggressor + a defender; the faster/lower-curve deck is usually the beatdown. Misassigning your role loses games.
• 60-card baseline split: ~24 lands / 36 spells. Aggro 20–22 land · midrange 23–25 · control/ramp 25–27.
• Curve: peak at 2–3, taper up; more cheap than expensive. Target deploying a relevant spell every turn 2→5 (“curve out”). When two cards are close, play the cheaper.
• Consistency: 60-card → run 4-ofs of your plan’s core; get redundancy via functionally-similar effects + tutors. Singleton (Commander/Brawl) can’t rely on drawing one card → lean on redundancy of effects, tutors, card selection, and higher land counts (99 ≈ 35–38 lands). Never port 4-of assumptions to singleton.
• Interaction density (60-card): control 10+ · midrange ~6–10 flexible answers · aggro least (burn doubling as removal). Prefer flexible/catch-all answers over narrow hate. Commander ref: ~8–10 removal + ~7–10 win-cons in a 99.
• Avoid (named mistakes): too many good cards/no plan · greedy mana (pips without the 14/18/21 sources) · too many taplands/slow mana · no turn 1–3 plays · win-more cards · too few lands · singleton math error.

Sources: Reid Duke “Level One” (magic.wizards.com/en/news/feature/level-one-full-course-2015-10-05); WotC “How to Build a Mana Curve”; draftsim.com; cardkingdom.com blog.

3. cEDH TEMPLATE (EDHREC / edhtop16 tournament data)

• Lands ~28–31 (low, because avg CMC ~1.5–2.5 + fast mana). Target total mana sources (lands+rocks+dorks+rituals) ≈ 40–45.
• Fast mana ~10–12 pieces: Sol Ring, Mana Vault, the Mox suite (Chrome Mox, Mox Diamond, Mox Opal), Lotus Petal, rituals; dorks (Birds, Esper Sentinel, Noble Hierarch, Deathrite) in green.
• Interaction ~12–15+, efficient (≤3 MV): free counters are the backbone — Force of Will, Fierce Guardianship, Flusterstorm, Mental Misstep, Pact of Negation (all ~84–87% play). Plus a board wipe (Toxic Deluge) + cheap spot removal (Swords, Into the Flood Maw).
• Win-con: compact cheap 2-card combo + backup + protection — model = Thassa’s Oracle + Demonic Consultation/Tainted Pact (3 mana, instant win); back up with Jace WoM; protect with free counters, don’t rely on it resolving unopposed.
• Tutors + draw engines run heavily: Vampiric/Demonic/Mystical Tutor; Rhystic Study, Mystic Remora, Esper Sentinel, Smothering Tithe.
• Archetypes: Turbo (max fast mana, min interaction, fastest) · Midrange (balanced, role-flexible) · Stax (taxing/lockdown + stack interaction, more lands).
• ★ Banlist (current Feb 2026), do NOT include: Mana Crypt, Jeweled Lotus, Dockside Extortionist, Nadu (banned Sept 2024; Feb 2026 kept them out). Validate strict color identity (all mana symbols in cost + rules text). Legality = cards.db commander_legal=1.

Sources: edhrec.com/guides (mana, interaction); edhtop16.com/staples; learncedh.com; WotC B&R Sept 2024 + Feb 2026.

4. CONSTRUCTED META + SIDEBOARD (MTG Arena Zone)

• Sideboard (the 15): reinforce your deck’s identity (proactive vs reactive). Allocate 60–70% of slots to your worst + most common matchups; rest flexible. Copy-count = commitment (1 niche · 2 relevant · 3 regular · 4 core) — avoid a pile of narrow 1-ofs without tutors. Every SB card needs a named job + a card it comes in for. Pre-write bring-in/bring-out plans for the top 5–6 archetypes; swap 3–6 cards per matchup (cap ~7–9).
• Positioning: the 60 beats the field on average; the 15 patches the gaps. Aggro/midrange/control is loose RPS — position vs what’s overrepresented, not the abstract “best” deck. Proactive if faster/linear; reactive if you out-grind.
• Bo1 (Standard Brawl, Arena ladder): hand-smoother draws 2 hands keeps the nearer-to-average → run ~1–2 fewer lands than Bo3 (aggro ~20–23, control/ramp ~26–27 in Bo1). No sideboard → maindeck must answer the whole field (flexible answers, maybe maindeck 1–2 would-be SB cards). Standard Brawl singleton → tutors/selection over 4-of redundancy; free first mulligan lets you mull harder.
• Current (2026-07) Standard build-against list (STALE-CHECK before use — re-pull mtgazone/mtgdecks): Bo3 T1 = Mono-Red Aggro (Screaming Nemesis), Dimir Midrange (Kaito, Enduring Curiosity), Golgari Midrange (Unholy Annex). T2 = Azorius Oculus, Gruul Prowess, Boros Burn, Zur Domain, Jeskai Convoke. Reserve the biggest SB blocks for Mono-Red (cheap lifegain/blockers/sweepers) + graveyard hate for Dimir/Oculus.

Sources: mtgazone.com (sideboard, Bo1 land count, Standard tier lists — July 2026, treat meta shares as time-sensitive); scrollvault.net.

Meta/tier data is time-sensitive — re-pull before relying on exact shares/key cards. Mana math, fundamentals, cEDH template, and sideboard ratios are durable.