Your mana curve is the distribution of your spells across their mana values. Four one drops, twelve two drops, eight three drops, and so on. Every deckbuilding guide tells you to have a good curve. Very few of them tell you what that means or how to check whether yours is any good.

Here is what a curve is actually for, the number that changes how you should build one, and a procedure for getting yours right.

What a curve is really for

The point of a curve is not tidiness. It is that unused mana is gone forever.

Magic gives you a growing amount of mana each turn, and any of it you do not spend disappears at end of turn. A turn where you play a land and pass has converted two mana into nothing. Do that three times in a game and you have thrown away roughly as much mana as a large spell costs, except you got nothing for it and your opponent had three free turns.

So a curve is a plan for spending your mana every turn from the moment the game starts. A deck curves out when it has something worth casting on each of its early turns. A deck stumbles when it does not.

That is the whole idea. Everything below is about making it happen reliably.

The number that should change how you build

Here is the part that surprises people. Playing four copies of a card does not mean you will have it.

These are the odds of having drawn at least one copy from a group of a given size, in a 60 card deck, on the play:

Cards in the group	By turn 1	By turn 2	By turn 3	By turn 4
4	40%	44%	49%	53%
6	54%	59%	64%	68%
8	65%	71%	75%	79%
10	74%	79%	83%	86%
12	81%	85%	89%	91%
14	86%	90%	93%	95%
16	90%	93%	95%	97%

Read the first row again. Four copies of a two drop gives you a 44% chance of having one to cast on turn two. That is worse than a coin flip, using the legal maximum of the card you most wanted.

Inverting it makes the point even more plainly. To have a spell available at a given cost with reasonable reliability, you need:

• About 11 cards at that cost for an 80% chance by turn two
• About 15 cards at that cost for a 90% chance by turn two

Both of those numbers are far above four, which is the most copies of one card you are allowed to play.

Which means you count effects, not cards

The rules cap you at four copies, so reliability cannot come from copies. It has to come from cards that fill the same slot.

Eight two drops is not eight copies of one creature. It is four of one and four of another, or three plus three plus two, chosen so that any of them is a fine turn two play. That is what experienced deckbuilders mean when they say a deck runs “eight two drops.” They are buying the reliability the table above prices, and the individual card identities matter less than the count.

This is the single most useful reframe in deckbuilding. Stop asking whether you have the best two drop. Start asking how many acceptable turn two plays your deck contains.

The same logic runs through the rest of your deck, and our card draw odds table covers it in more depth.

What a workable curve looks like

There is no single correct curve, because the right shape depends on what your deck is trying to do. Using the archetypes as a frame:

Aggressive decks want the bulk of their spells at one and two mana, tapering fast. Something in the region of eight one drops, twelve two drops, six three drops, and very little above that. The deck is trying to end the game before expensive cards matter, so an expensive card is a liability rather than a payoff.

Midrange decks centre on two through four, usually heaviest at three. Perhaps four one drops, ten two drops, ten three drops, six four drops, and a small top end of two or three cards. The deck wants a play every turn and a strong one in the middle turns.

Control decks skew later and lean on cheap interaction rather than cheap threats. Ten to twelve cards at one and two mana that answer things, a sweeper or two at four, and a genuine top end that wins the game once you get there.

Combo decks are the exception. Their curve follows whatever the combo requires, and the rest of the deck is card selection and protection. Ignore curve advice that conflicts with assembling the pieces.

Treat those as starting shapes rather than targets. The test is always whether the deck has something to do on the turns that matter for its plan.

Your curve and your land count are one decision

People tune these separately and then wonder why the deck feels wrong.

A low curve wants fewer lands, because you stop needing new ones after turn three or four and every extra land is a card that does nothing. A high curve wants more, because missing your fifth land drop when your best cards cost five is how you lose.

Roughly: aggressive low curve decks land around 20 to 22, midrange around 23 to 25, control around 25 to 27. The reasoning behind those numbers, and what they actually buy you, is in how many lands should you play and the land count chart.

The other half of that decision is colours. A curve full of double coloured costs at two mana needs a very different mana base from one with single costs, and the mana tables show how many sources each demands.

Four ways curves go wrong

Too top heavy. The most common failure by a wide margin. Powerful expensive cards are fun to pick and easy to over include. If you are casting your first spell on turn three most games, you have already given away the early turns.

A hole at two. Turn two is the most valuable slot in most decks, because it is the first turn you can do something meaningful and it is where you are least likely to be interrupted. A deck with six two drops will have a play there under 60% of the time.

Counting the curve without counting castability. A three drop you cannot cast on turn three because your mana does not cooperate is not a three drop. Colour requirements are part of the curve.

Forgetting what happens after turn four. A curve that is only about the early game runs out of cards. Card advantage and a reason to keep playing matter too, which is the trade covered in card advantage and tempo.

Adjustments that genuinely change the shape

• Cheap card selection effectively increases every group in your deck, because a cantrip that digs three cards deep makes your eight two drops behave more like ten. This is the whole basis of the Turbo Xerox principle.
• Ramp lets you play a higher curve than your land count suggests, but only when it resolves and survives. Count a mana creature as something less than a full land.
• Flexible cards that can be cast early or late, or that have modes at different costs, occupy more than one slot on your curve and are worth more than their raw power suggests.
• Interaction is part of the curve. A removal spell is a turn two play. Decks that count only their threats end up with far more early plays than they realised, or far fewer.

A procedure that works

1. Decide what your deck does and by when. Aggro needs turn one through four. Control needs to survive to turn six. That decision governs everything else.
2. Count the cards at each cost, not the card names. Write down how many acceptable plays you have at one, two, three, and four mana.
3. Check the counts against the table above. If you want a turn two play most games, you want double figures at two mana. If you have six, you do not have a turn two play, you have a hope.
4. Set your land count from the curve, not the other way round.
5. Check castability. Count coloured sources for your most demanding early costs.
6. Play twenty games before changing anything. Curves feel wrong for reasons that are often about the mana base or the matchup rather than the shape.

The short version

A curve is a plan for spending mana on every turn of the game, and unused mana never comes back. Reliability at any given cost comes from the number of cards that fill the slot, never from copies, because four copies of the perfect card is a coin flip and eleven acceptable ones is an 80% chance. Count effects, tune your lands to the shape, and check that you can actually cast what you built.