Premodern Mana Sources

Requirement numbers are Frank Karsten's, from his 2022 update. A spell counts as castable on curve when you have at least an 89 + M percent chance of drawing the colored sources for it by turn M, conditional on having drawn M lands. One-drops are held to 90 percent, four-drops to 93. He publishes the numbers at 20, 25 and 30 lands; the column here tracks your list and interpolates between them.

Turn one counts untapped sources only. From turn two everything counts, tapped or not, on the reasoning that a tapland could have been played a turn earlier. That one rule is why a card can miss on turn one and clear on turn two without the deck changing.

Multicolor costs are split, looked up per color, and each requirement raised by one. They report as one row rather than one per color. You need both at once, so the worse color decides the row and the later turn decides when it becomes reliable.

Falling short is a statement about timing rather than a failure. One colored pip wants 14 sources at 25 lands to be reliable on turn one, and 13 by turn two. Anything short is also given the turn your current sources do support it.

Gush is where mana costs stop describing the cost. Free-cast, it wants two Islands in play, and two blue sources is a different quantity: Skycloud Expanse and Tainted Isle both make blue and neither is an Island. Land types written as pips count against lands carrying that type, plus any fetchland that can find one. Snow-Covered Islands carry the type.

There are no dual lands in the format. A fetchland finds a basic, so its colors come from the rest of your list. Flooded Strand in a deck with no Islands is not a blue source.

Mana symbols overstate a lot of lands here. Skycloud Expanse prints white and blue and makes neither alone. Cabal Coffers and Gaea's Cradle make nothing on an empty board. The Tainted cycle is colorless until a Swamp is out. Anything excluded from a count is listed above it with the reason.

The land-drop panel is exact, not an estimate. It reads Karsten's 2017 hypergeometric table for the probability of playing a land on each of your first turns, on the play, under his standard mulligan rule, at whatever land count you pasted. It is deliberately not a single recommendation: he showed that a deck needing three lands to operate wants a different number from a low-curve deck, and noted himself that there is no scientific reason to prefer 83.5 percent over 80 or 87. The numbers let you make that call rather than making it for you. The toggle switches between playing and drawing first; every figure is a few points higher on the draw, since you have seen one more card.

One thing the panel does not do is adjust for cantrips. It reads the table at the land count you typed, and a deck full of cheap card draw effectively runs more lands than that number suggests, since each cantrip digs you toward your next drop. Karsten's own land-count formula accounts for this: lands = 19.59 + 1.90 times your average mana value, minus 0.28 for each cheap card-draw or mana-ramp spell. A four-cantrip deck is therefore about one land richer than it looks. This applies to replacement cantrips like Opt that draw a card and swap themselves out one-for-one; a dig spell like Impulse that looks at several cards and takes one is card selection rather than replacement, so it does not thin your deck the same way and is a looser fit for the formula. The tool cannot sort this out for you, since it means reading what each card does, so run the adjustment yourself and read the panel at the higher number if you want the fuller picture.

Nonland mana is included. Dorks count half, because they die. Artifact accelerants count three quarters. Mox Diamond counts one. Rituals count zero: Dark Ritual multiplies black you already have, and it will not cast a turn-one Duress.