How much ramp, draw and removal does a Commander deck carry? We counted 165 precons

Ramp, card draw, removal, board wipes and lands, counted across 165 Commander precon lists with the classifier behind our deck tiles: medians, ranges, and how little commander cost moves them.

Aymeric Lamontagne

Nov 10, 2026

Here is the honest headline: the numbers everyone argues about are easy to hit, one at a time. We ran the 165 distinct Commander precon lists in our catalog through the classifier behind Manamancer’s deck tiles, and on November 4, 2026 the median list read 12 ramp, 13 card draw, 10 removal, 3 board wipes and 38 lands. Every one of those meets or passes the starting targets the tiles grade against: 10, 10, 9, 3 and 36–38.

That should temper the celebration when your own deck reaches every target. Wizards’ Brackets announcement of February 11, 2025 places the average current precon at Bracket 2, so a deck that matches these medians has matched a box product. Meeting everything at once is rarer: only 38 of the 165 lists do it.

The second finding deflates a popular question. Precon ramp barely moves with the commander’s cost: the median is 12 under a commander of three mana or less and 12.5 under one of seven or more. This piece goes through each role, gives the odds behind each count, and models what ramp does for a five-, six- or seven-mana commander. It then takes up 37 lands and the so-called 75% rule.

The counts: medians and ranges

The sample is the Commander Decks line of our precon catalog: 189 published lists from June 17, 2011 to October 2, 2026, cut to 165 by dropping 24 Commander Anthology reprints and Collector’s Edition twins that share at least 90% of their cards with a list already counted. The 80 lists in the last column were released from 2023 on. As a check, we ran all 189 lists through the function behind the tiles, and every count matched.

Those five medians do not describe one typical deck. Only 62 of the 165 lists meet all four role targets, and 38 also land inside 36–38 lands. Among the 80 recent lists, those figures are 38 and 28.

How many ramp spells in a Commander deck?

Precons run 12; the app’s starting target is 10. The middle half of lists holds 9–14, and the extremes are far apart. Commander 2013’s Mind Seize has 3 (Obelisk of Grixis, Sol Ring, Wayfarer's Bauble), while Bloomburrow’s Animated Army has 28. Two slots are nearly fixed before any deckbuilding happens: Sol Ring is in 164 of the 165 lists and Arcane Signet in 123.

Colour sets the shape. Lists with green have a median of 12.5, split evenly between 4 land-ramp spells and 4 mana rocks. Lists without it have 10, with 7 rocks and 1 land-ramp spell. And 12 ramp cards are not twelve two-drops: the median list has 6 at mana value two or less and 9 at three or less.

The mechanism is simple arithmetic. In a 99-card library, seeing seven cards plus one a turn, the chance of having found at least one ramp card by turn two is 63% with 10, 70% with 12 and 79% with 15. That is the odds of holding one, not of casting it, which still takes the lands.

How much card draw in a Commander deck?

Precons run 13, and recent ones 14.5, against a target of 10. The range runs from 1 in Commander 2011’s Heavenly Inferno to 31 in Murders at Karlov Manor’s Deep Clue Sea, where every Clue-maker counts. No role has climbed further: the median was 9 in 2011–16 and is 15 in 2025–26.

Read that label with care. The three commonest “draw” cards in the sample are Commander's Sphere (68 lists), Mind Stone (41) and Solemn Simulacrum (29), which are mana first and a card later. Cheap draw is much scarcer: the median list has 3 draw cards at mana value two or less and 7 at three or less. Blue shows up as you would expect, at 15 with blue and 11.5 without.

Draw matters because it sets the pace of everything else. Without it you see one new card a turn, or 15 of your 99 by turn eight, and every other role’s odds in this piece assume exactly that. The chance of finding a draw card by turn four is 71% with 10, 81% with 13 and 87% with 16.

How much removal in a Commander deck?

Precons run 10; the target is 9. The middle half holds 8–13, from 3 in Deep Clue Sea to 22 in Commander 2019’s Merciless Rage. It has drifted down, from a median of 12 in 2011–16 to 10 since 2023, while draw went the other way. Black lists carry 12 and the rest 10.

The label is broad: spot removal, counterspells, bounce, fight spells and edicts all count. The usual suspects lead, with Swords to Plowshares in 37 lists, Chaos Warp in 29, Beast Within in 24 and Path to Exile in 21. So does an impostor: Bojuka Bog, also in 29, counts because a text rule looks for “exile target”, and the Bog’s target is a graveyard.

The odds show how thin that is at a four-player table. With 9, you have found one by turn five 70% of the time, 78% with 11 and 86% with 14. That is one answer, against three opponents who are each playing threats.

How many board wipes in a Commander deck?

Three, and here the precons and the target agree: 62 of the 165 lists have exactly 3, and 134 have at least that. But the tile counts two kinds. Board wipes clear a board whatever is on it, while soft wipes are gated by size or condition, one-sided or repeating. The median list has 2 of the first and 1 of the second, only 76 lists reach three hard wipes, and 11 have none at all. The commonest are Blasphemous Act (25 lists), Chain Reaction (15), Austere Command (13) and Steel Hellkite (12).

Three wipes are not a plan to hold one. The chance of having found one by turn eight is 39%, so about six games in ten reach that turn with none seen, even in a deck that meets the target. Four gets you to 49% and six to 64%.

As for how many is too many, the sample offers only a description: 17 lists, about one in ten, run six or more, topped by Commander 2015’s Wade into Battle with 9, seven of them soft. Those seven are damage sweeps such as Earthquake, in a deck whose big creatures can often survive them. Whether a sweeper also kills your own plan is the real test, and no count can see that.

How much ramp for a 5, 6 or 7 mana commander?

If you were hoping precons would answer this, they mostly shrug.

The rank correlation with commander mana value is −0.01 for ramp and 0.08 for lands; even the average cost of the other spells only reaches 0.24. Green versus non-green moves the ramp median by 2.5 cards, more than any step on this table.

Two caveats before anyone reads that as a rule. First, the top buckets are tiny. In the 2023–26 lists alone a faint tilt appears (0.13, with medians of 12, 13, 12, 16 and 14), but its last two buckets hold three lists each. Second, a printed mana value can overstate the cost: of the six commanders at seven or more, Millicent, Restless Revenant, Yuma, Proud Protector and Ghalta, Primal Hunger all discount themselves. Millicent’s list even runs 35 lands, the second-lowest count in the sample.

A flat line describes what a precon does; it does not say what your deck should do. The tile’s ramp target of 10 is just as flat, the same for a three-drop as for Ghalta, Primal Hunger. So we modelled the question directly: 37 lands, a given number of ramp pieces, and the chance of having the mana on a given turn.

If the bar is about four games in five with the commander castable on time, the table says roughly 10 ramp pieces for a five-drop, 15 for a six-drop and 20 for a seven-drop. That bar is our choice, and it is not free. Twenty ramp pieces plus 37 lands leave 42 cards for everything else. Waiting a turn is the cheaper alternative: with 10 pieces a six-drop goes from 69% on turn six to 78% on turn seven, and with 12 a seven-drop goes from 62% on turn seven to 72% on turn eight.

The model is deliberately plain, from five million shuffles per row: a 99-card library, one land drop a turn, a redraw of any opening hand with 0, 1, 6 or 7 lands, and every ramp piece costing two and adding one mana from the next turn. It has no card draw, which makes it pessimistic. It also has no colours, tapped lands, three-mana ramp or wipes, which makes it optimistic. Treat the gaps between rows as the finding, not the percentages.

Is 37 lands enough for Commander?

It is common, not automatically enough. Of the 165 lists, 45 run exactly 37 and 47 run 38, while 63 run more and only 10 run fewer. Among the 80 recent lists, 27 run 37, 30 run 38 and 4 run fewer. These are plain lands: no list in the sample holds a modal double-faced land.

Here is what one land buys, with no mulligans, seven cards plus one a turn, and the first draw that rule 103.8c gives every player in a multiplayer game:

The step from 37 to 38 adds about three points to hitting your fourth and fifth land on time, and about one to the flood column. In the ramp model, spending two slots on lands (39 lands, 10 ramp) and two on ramp (37 lands, 12 ramp) both put six mana on turn six at 74%. Ramp gets there a turn early more often, 54% against 50%. It also needs a two-mana window and, for rocks and creatures, needs them to survive, which the model does not punish.

For a formula, Frank Karsten’s land-count analysis on TCGplayer, a page last updated February 13, 2025, gives, for 99 cards, 31.42 + 3.13 × average mana value − 0.28 × cheap draw or ramp spells. He defines cheap as mana value two or less and calls the Commander version an imprecise back-of-the-envelope estimate, ported from 60-card data. It reaches 37 lands at an average of 2.7 with ten cheap pieces, or at 3.5 with about nineteen. The median precon averages 3.69.

Using our ramp and draw labels as a stand-in for his definition, the formula asks for a median of 40.6 lands where precons run 38, and 153 of the 165 lists sit below it. For the recent lists that is 39.6, with 73 of 80 below. Either precons run light or the port runs heavy, and with no game data this piece cannot settle which. Opinions range widely as well: a CoolStuffInc guide dated April 27, 2026 reports EDHREC’s average at about 36 and creator Rebell Lily’s safe number at 43, with its author around 37.

A useful cross-check is lands plus ramp together, which has a median of 50 in the sample and 51 since 2023. That is about the 50 mana sources SaffronOlive suggested on MTGGoldfish on May 5, 2017.

What is the 75% rule in MTG?

It is a community rule of thumb, not a rule of the game, and it has two meanings.

The power meaning is the older one in our sources. In a CoolStuffInc article of February 13, 2014, Jason Alt describes a deck built “75% of the way to an optimal build”: able to compete with a tuned deck on a good day, without running over a casual table. Nothing measures that percentage. It is a statement of intent. Wizards’ own vocabulary for that conversation, since 2025, is Commander Brackets, and the announcement does not use the term.

The theme meaning comes from the same 2017 SaffronOlive piece: a deck reads as a theme deck when about three quarters of its nonland cards are on theme, while half can still work. That one can be set against our count. The median list has 62 nonland cards, and 35 of those carry at least one role label, which leaves 27 free of them. Reaching 75% means about 46 on-theme cards, so roughly 20 role cards would also have to be on theme. That is the arithmetic case for a ramp or draw card that also fits the theme, and a reason to read a role count alongside the theme rather than instead of it.

What changed since 2011

Old precons make poor benchmarks, because the product has moved.

Each column tells part of the same story: more ramp, more draw, a cheaper curve, a land fewer and slightly less removal. One measurement caveat: the Tagger role rules were calibrated on the 3,000 most-played cards as of October 2026, so older lists may be undercounted a little. And nothing here says why Wizards made those choices. It only says the choices were made.

Where this count is wrong

Five limits, because the table above looks more certain than it is.

First, the labels overlap. The median list has 40 role labels on 36 cards. Commander's Sphere is ramp and draw, Arcane Denial is removal and draw, and Vandalblast is removal and a wipe.

Second, the classifier makes mistakes in both directions. It combines rules-text patterns with Scryfall Tagger’s community labels. Bojuka Bog is a text false positive, and Sun Titan, in 16 lists, reads as ramp through Tagger’s land-ramp label. Going the other way, a tutor that finds removal does not count as removal.

Third, it reads the card, not your deck. A modal card counts even if you always choose the other mode, and a ramp creature does not stop counting because your table kills creatures on sight.

Fourth, the sample is one publisher’s product line, built around themes and reprints and not offered as a set of strong decks. It contains no game results, so nothing here can say that one count performs better than another.

Fifth, it is a snapshot. The lists come from the catalog’s September 26, 2026 data, the card text from an index refreshed on November 3, and the Tagger labels from October 30. The four Star Trek decks, due November 13, have no lists yet and are not counted. Neither are the catalog’s separate Secret Lair and Brawl lines.

How to use the counts

Compare your own numbers with the ranges rather than the medians. By definition half the lists sit outside the middle-half column, so 15 ramp or 8 is not a diagnosis. Then look at the cards behind the count. How many of the 12 ramp pieces cost two? How many of the 13 draw cards are really rocks you sacrifice late? Are the three wipes hard, and do they kill your own plan?

On a Commander deck in the builder, the Overview tiles show these same counts against the same targets, and an account is needed to save decks. To see what a real list looks like at these numbers, browse the precons.

Read how to read your deck stats — The guide to what the panel measures, what it estimates, and what a count can’t tell you.

May your fifth land arrive on turn five, and may one of your three wipes turn up on the turn the table needs it.