How much card draw does a Commander deck need?
The usual answer is about ten, counted as if Divination and Phyrexian Arena were two of the same card. They aren't. One draws two cards and is done, and the other draws a card every turn for the rest of the game, which adds up to a lot or very little depending on when you find it. Count those two separately and the number you need comes out of the arithmetic.
An engine is worth the turns left after it lands
Phyrexian Arena draws one extra card in each of your upkeeps. Cast it on turn 3 and, by the end of turn 10, it has drawn seven cards. Cast it on turn 7 and it has drawn three. A one-shot doesn't care when you cast it: Divination always draws two, and Harmonize always draws three.
| Engine cast on | Extra cards by turn 10 | The same as |
|---|---|---|
| Turn 3 | 7 | More than two Harmonizes |
| Turn 5 | 5 | A Harmonize and a Divination |
| Turn 7 | 3 | One Harmonize |
| Turn 9 | 1 | Half a Divination |
Two things follow. First, an engine found late is just a slow one-shot, so the whole case for engines rests on finding one early. Second, an engine that triggers off your opponents rather than your upkeep, like Rhystic Study or Esper Sentinel, can draw more than one card per round in a four-player game. That is why the best of them is on the Game Changers list.
So count the engines that arrive early
Here is the chance of having seen at least one engine by each turn, on the play in a 99-card deck. These are exact hypergeometrics, the same math as this site's calculators:
| Engines in the 99 | ≥1 by turn 3 | ≥1 by turn 5 | None by turn 7 |
|---|---|---|---|
| 2 | 17.4% | 21.1% | 75.3% |
| 4 | 32.1% | 38.1% | 56.4% |
| 6 | 44.4% | 51.6% | 42.0% |
| 8 | 54.7% | 62.5% | 31.0% |
Even eight engines find one by turn 3 in only about half of games, and nobody wants eight cards that do nothing the turn they land. That's the honest limit on engines in the 99. You can't count on one early, you can only improve the odds. It's also why a cheap engine beats an expensive one: a three-mana engine drawn on turn 3 can be cast that turn, and a five-mana one can't.
Your commander may already be the engine
The biggest variable on this page isn't in your 99 at all. A commander that draws cards is an engine you have in every game, from the turn you can cast it, with no probability attached. Tatyova, Benthic Druid draws whenever a land enters under your control, and Toski, Bearer of Secrets draws whenever one of your creatures connects with a player. A deck built around a commander like that has already solved the problem the table above describes, and its own draw count can shrink to a handful of one-shots that refill the hand after a wipe.
Draw and card advantage are different budget lines
“Card draw” gets used for three kinds of card, and only two of them belong in this count.
Cantrips replace themselves. Opt, Preordain and Ponder draw one card for the card you spent, so you end with as many cards as you started. They improve which cards you hold, not how many. Count them at zero here. They matter for your land count, which is further down.
One-shots leave you ahead. Night's Whisper spends one card and draws two, and Harmonize draws three. They're the same value whenever you cast them, and they're what refills a hand after a board wipe.
Engines keep drawing. Phyrexian Arena, Skullclamp, Guardian Project and Esper Sentinel are all engines, and their value is the first table on this page.
Card advantage is a wider idea than any of these. A board wipe that kills four of your opponents' creatures and none of yours is card advantage, and so is bringing a creature back from your graveyard, but neither one draws a card. Those cards earned their slots in another budget, like removal, so count them there. Counting a wipe as card draw as well is how a deck ends up short of both.
The answer, by deck
| Deck | Draw sources | Of which engines |
|---|---|---|
| Commander draws cards itself | 5 to 7 | Few, because the commander is the engine |
| Most decks | 8 to 10 | 4 or more, at 3 mana or less |
| Control or spellslinger | 12 to 15 | 5 or more, plus instant-speed one-shots |
The middle row agrees with the convention, and the engine table is the reason. Four cheap engines find one by turn 5 in 38.1% of games, which is worth having, and the one-shots carry the rest. Low-curve aggressive decks are the exception people argue about. They empty their hand fast, so what they want is a refill, and a wheel like Windfall that hands every player a fresh hand does that job in one card. Count those by what they refill rather than by this table.
Does drawing more let you play fewer lands?
Seeing more cards does find more lands, and the size of the effect is easy to compute. Here is the chance of making each land drop on time, on the play with 37 or 36 lands, and what one, two or three extra cards seen by that turn do for the 36-land deck:
| Land drop | 37 lands | 36 lands | 36, +1 card | 36, +2 cards | 36, +3 cards |
|---|---|---|---|---|---|
| 3rd, on turn 3 | 72.7% | 70.4% | 78.0% | 84.0% | 88.6% |
| 4th, on turn 4 | 55.5% | 52.7% | 62.1% | 70.3% | 77.3% |
| 5th, on turn 5 | 39.1% | 36.2% | 45.6% | 54.7% | 63.1% |
One extra card seen by turn 4 more than pays for the land you cut: 62.1% against 55.5%. The catch is getting that card in time. Only draw cast by turn 3 adds a card before turn 4's land drop, which in practice means one-mana cantrips, and with eight of them in the deck you've seen even one by turn 3 in 54.7% of games.
Put the two together exactly, with every cantrip cast on the turn it's drawn, and four one-mana cantrips at 36 lands make the fourth land drop on time in 56.4% of games, against 55.5% for 37 lands and none. That is the most generous version of the trade. It assumes the mana for each cantrip is always spare, and that you never want to keep one for later. So the honest rule is at most one land for every four one-mana cantrips, and less for two-mana draw spells, which can't be cast on turn 1.
The second catch is the one that builds bad decks. A cantrip finds a land. It doesn't make mana. Every cantrip you cast spent mana and a card to dig, so a deck that leans on draw to replace lands spends its early turns digging instead of developing. That's why the ramp and land count page stops at 36 lands even for a deck running ten-plus cheap ramp alongside cheap draw, and no amount of draw takes you lower. The calculator below runs the same land math on your own counts.
Check your deck's own numbers
Enter your engines, your lands and your one-mana cantrips. It opens on four engines and four cantrips at 36 lands.
With 4 engines you find one by turn 5 in about 38% of games. At 36 lands with 4 one-mana cantrips, your fourth land drop arrives on time in 56.4% of games, against 55.5% for a plain 37-land deck.
On the play, no mulligans. Every cantrip seen by turn 3 is assumed cast and drawn from by turn 4, the generous case, and a card a cantrip draws is never another cantrip.
FAQ
How much card draw should a Commander deck run?
Most decks want 8 to 10 sources of card draw, and at least four of them should be engines that keep drawing and cost three mana or less. The engines are what matter, because an engine cast early draws far more than any single spell, while one found late is barely worth a Divination. Control decks want 12 to 15, and a deck whose commander draws cards on its own can get by with 5 to 7.
Do cantrips count as card draw?
Not toward your card draw count. A cantrip like Opt or Ponder draws one card for the card you spent on it, so you finish with the same number of cards you started with. It improves which cards you hold rather than how many. Cantrips do help your land drops, since they let you see more cards early, but that is a separate question from card advantage.
Can I run fewer lands if my deck has a lot of card draw?
One fewer at most, and only with cheap draw. One extra card seen by turn 4 more than pays for a cut land, but only draw cast by turn 3 can deliver it, which in practice means one-mana cantrips. Even counting every cantrip cast the turn it is drawn, four of them at 36 lands make the fourth land drop on time in 56.4% of games against 55.5% for 37 lands, so the trade is roughly one land per four cantrips and no further.
What is the difference between card draw and card advantage?
Card draw is cards entering your hand, and card advantage is ending up with more useful cards than your opponents by any route. A board wipe that kills four opposing creatures and none of yours is card advantage, and so is bringing a creature back from the graveyard, but neither draws a card. Count those in the budget you bought them for, like removal or recursion, rather than in your draw count.
Can I play Rhystic Study in every bracket?
No. Rhystic Study is on the official Game Changers list, so a Bracket 2 deck cannot run it, and in Bracket 3 it uses one of the three Game Changer slots. The One Ring, Necropotence and Consecrated Sphinx are on the list too. Phyrexian Arena, Esper Sentinel and Mystic Remora are not, so they fit any bracket.