Mulligan Strategy Calculator

How the Mulligan Strategy Calculator works

Every Commander game starts with the same decision, made before you've cast a single spell: is this seven-card hand good enough to keep? Most players answer that question on gut feeling — "I've got three lands and a two-drop, that's probably fine" — because actually working out the odds of a hand like this developing into a real turn-three or turn-four play requires math most people don't want to do at the table. The Mulligan Strategy Calculator does that math for you, ahead of time, so you walk into every game with an actual number behind your mulligan decisions instead of a hunch.

The tool works by running a genuine Monte Carlo simulation — 10,000 independent, simulated games of your deck — rather than a lookup table or a simplified formula. Each simulated game shuffles a virtual copy of your deck (built from the category counts you enter, or from a decklist you paste in), draws a real seven-card opening hand, and then decides whether to keep or mulligan that hand using the same category targets you set. If it mulligans, it redraws exactly the way Commander actually works: your free mulligan first, if you have it enabled, which lets you draw a fresh seven with no hand-size penalty at all, and then every mulligan after that follows the London mulligan rule — draw seven again, then choose that many cards to keep minus one, down to a hard floor of four cards. Running the actual shuffle and the actual keep/mulligan decision ten thousand times over is what makes the output a real probability rather than a guess dressed up as a percentage.

What counts as a category, and why three

The calculator tracks three fixed categories — Lands, Ramp, and Interaction — because those are the three things almost every Commander mulligan decision actually turns on. Does this hand have enough lands to function at all? Does it have a way to accelerate past a slow start? Does it have an answer if the table plays something that has to die right now? You set two numbers for each category: how many copies of that category exist in your deck (its "copies in deck"), and what you want to have seen — and by which turn — for the hand to count as a success. A default preset is loaded for you (36 lands wanting 3 by turn 3, 10 ramp pieces wanting 1 by turn 2, 10 interaction spells wanting 1 by turn 3) as a reasonable starting point for a midrange 99-card deck, but every one of those six numbers is yours to change.

If you'd rather not count your own decklist by hand, use the Import decklist tab: paste your list in directly (the same plain-text format — "1 Sol Ring", one card per line — that the main deck analyzer accepts), or paste a Moxfield, Archidekt, or TappedOut deck link and the importer will fetch the list for you. Either way, it looks up every card, reads its type line and oracle text, and buckets it: any land goes to Lands, anything that fetches a land, taps for mana, or otherwise accelerates you goes to Ramp, and anything that destroys, exiles, counters, or otherwise answers an opposing permanent goes to Interaction. This is a fast heuristic built for convenience, not a hand-verified card list — it will occasionally miscategorize an odd card or miss one it doesn't recognize, and it always shows you what it detected before you compute, so you can nudge any number up or down before running the simulation.

The keep-or-mulligan rule

A simulated hand is kept once it contains at least half (rounded up, with a minimum of one) of each category's target count. That threshold is deliberately looser than requiring the full target already sitting in your opening seven — a hand that wants three lands by turn three doesn't need all three lands in the opener, because two of your normal draws happen before turn three arrives anyway. Requiring the entire target in hand on turn zero made almost every simulated hand mulligan all the way down to the four-card floor, which doesn't match how anyone actually plays: real players keep hands that look like they're "on track," not hands that already have the full plan assembled. The half-rounded-up bar is that same practical judgment call, encoded consistently across ten thousand games instead of applied inconsistently the way tilt, fatigue, or optimism sway a real decision at the table.

What "success" actually measures

Success is checked per category, per game, and a game only counts as an overall success if every category hits its target by its own turn. "Seen by turn N" means the cards in your kept hand plus one new draw for every turn from turn one onward — Commander doesn't skip the starting player's draw step the way some other formats do, so turn one already counts as a draw. A hand that kept at six cards after one free mulligan and then drew into a fourth land by turn three is a genuine success for a "3 lands by turn 3" target; the simulation tracks that exactly, card by card, rather than approximating it.

The penalty percentage exists to model something the raw math can't easily capture on its own: a hand that mulliganed twice hasn't just lost a card, it's lost tempo, flexibility, and often color fixing in a way that's hard to reduce to a single number. Every non-free mulligan you take rolls an extra chance of failure equal to the penalty percentage you set, stacking with each additional non-free mulligan in that game. Setting it to 0% tells the calculator to trust the raw card-counting math completely; raising it models a deck (or a playgroup) where falling behind on tempo is unusually costly.

What this tool does and doesn't model

This calculator is deliberately scoped to the mulligan decision itself — the one moment in the game where you have complete information about your hand and nothing else. It does not simulate opponents, so it can't tell you whether your interaction actually resolves against a deck full of counterspells, or whether someone at the table is about to end the game on turn four regardless of your hand quality. It also doesn't model card draw spells, tutors, cost reducers, or free spells triggering off other permanents — every card in the simulation is either drawn naturally or it isn't, with no allowance for a deck finding extra copies of what it needs mid-game. For a full turn-by-turn simulation of an entire game plan, including land sequencing, ramp coming online, and a tipping-point turn calculation, see the separate Deck Tester tool, which goldfishes your actual decklist rather than a three-category abstraction. The Mulligan Calculator is the faster, more focused companion to that: a way to sanity-check "should I keep this hand" against real numbers before a game even starts, using whatever categories matter most to your deck.

Because the simulation is seeded and deterministic in its shuffling logic, running the exact same inputs twice will always produce results within normal Monte Carlo noise of each other (typically well under half a percentage point on the success rate, given ten thousand simulated games) — small enough that you can trust the headline number to reflect your actual inputs rather than random variance in the tool itself.

FAQ