Tarkir: The Dragon Scroll | Second Main Phase

The Record

What actually happened 17Lands archive

Every other chapter here is expert judgment, recorded before or during the format. This one is the account of what followed. Card tiles throughout carry a win-rate chip: the games-in-hand win rate, meaning how often the player won when that card was in hand or on the battlefield. Set median is 55.5 percent, so anything above that beat the average card.

Games analyzed

552,428

Premier draft, whole format

On the play

58.0%

vs 52.1 on the draw

Play advantage

+5.9

points. Tempo matters here.

Average game

9.1

turns. Middling speed.

Colors	Deck	Win rate	Games
WR	White-red aggroTwo colors	58.4%	50,443
WBR	Mardu mobilizeClan	55.7%	75,233
URG	Temur harmonizeClan	55.4%	92,612
BG	Golgari midrange	55.1%	14,196
WB	Orzhov attrition	55.0%	12,252
UR	Izzet spells	54.9%	10,622
UG	Simic ramp	54.8%	20,977
WUR	Jeskai flurryClan	54.8%	69,224
UBG	Sultai renewClan	54.4%	79,588
WBG	Abzan endureClan	53.4%	46,781
WUBRG	Five color dragons	52.9%	18,362

Read the top row again. This is a set built entirely around three color wedges. Its five clans have their own mechanics, their own signpost uncommons, their own lands, and its own name is a clan name. The best deck in the format was two colors. White-red finished at 58.4 percent across fifty thousand games, nearly three full points clear of the best wedge, and no video in this collection names it.

The Two Color Heresy

The finding nobody called 108,490 two color games

Weighted across every deck with at least five thousand games, two color decks won 56.6 percent and three color decks won 54.9 percent. The wedges are not a trap exactly, because three of them still beat the format average, but the premium you pay for the third color is real and it is roughly one and a half points.

White-red is the sharp end of it. It is the only pair in the set with no clan of its own, which means nothing in the packs is trying to pull you into it, which in turn means it is chronically underdrafted relative to its power. It plays the mobilize creatures and the cheap aggressive commons that the Mardu drafters want but leaves behind the black cards and, crucially, the third color's tapped lands. In a format where the play advantage is nearly six points, the deck that curves out without stumbling on mana is doing something the others cannot.

How to use this. If white and red are both flowing at the end of pack one, you do not need to find the black cards to justify staying. Two color white-red is not the fallback when Mardu fails. On this evidence it is the better deck, and the tapped tri-land you did not play is part of why.

**Mardu Devotee**58.4%The best of the devotee cycle. A one drop that scries and fixes.

**Salt Road Packbeast**57.5%Cheap, aggressive, and the meta update called it underrated. It was.

**Fortress Kin-Guard**57.1%Top tier white common for the aggressive shell.

**Molten Exhale**57.9%Red removal at common, above the 78th percentile.

**Coordinated Maneuver**55.4%Cast it after mobilize tokens exist and it is two mana removal.

**Rebellious Strike**55.6%Plus three and a card. Best on a blocked token that was dying anyway.

The Experts Got This Right

The honest scorecard Seven videos vs the archive

Across the eight guides in this series, preview season has usually been shaky and post-release commentary reliable. Tarkir breaks the pattern in the good direction. This is the best calibrated set of predictions measured so far, and the overrated and underrated lists in particular are close to flawless.

10 of 10 correct

The overrated list

Every single card called overrated landed in the bottom quarter of the set. Median percentile: 10th. Hundred-Battle Veteran, the last name on the list, is the 4th percentile.

10 of 10 correct

The underrated list

Every card called underrated finished above the set median, from the 57th to the 86th percentile. Mardu Monument at the 86th was the standout call.

4 of 5 in order

The clan ranking

The meta update ranked Temur, Mardu, Jeskai, Sultai, Abzan. Actual order swaps only the top two, which finished 0.27 points apart. Abzan last was called exactly.

Median 95th

The bomb list

Eleven prerelease bombs, ten of them above the 83rd percentile. Ugin, Eye of the Storms is the single best card in the set at 68.5 percent.

Nearly exact

The color ranking

Predicted blue over green over red over white over black. Actual is blue, green, white, red, black. Only the middle two swapped, and by a quarter of a point.

The blind spot

The best deck

Seven videos, hours of coverage, and not one names white-red. Every archetype discussion is organized by clan, so the pair with no clan was never a category.

The lesson is about structure, not skill. These creators evaluated cards and clans with real precision. What they missed was the deck the set's own framing made invisible. When a set hands you five named archetypes, the edge is in the sixth one nobody is counting.

Seven Mechanics

Learn these before you sit down

Behold a Dragon

Control a dragon, or reveal one from your hand, and the spell gets its better mode.

Above rate when you behold, only slightly below rate when you cannot. Run them freely. Four dragons is enough to behold often.

Omen

Dragons carry a cheaper spell half. Casting the omen side shuffles the card back into your deck rather than exiling it.

Not a two for one. Treat these as flexible split cards, not card advantage.

Harmonize

Flashback with a discount: tap a creature as an extra cost and reduce the harmonize cost by that creature's power. All natural harmonize spells are sorceries.

Flashback level power, with the printed costs inflated to pay for the discount.

Renew

Pay a cost and exile the creature from your graveyard for an effect. Sorcery speed.

Attached to bodies opponents hate to remove. Extra value any time you self mill.

Endure

Endure N: either put N counters on the creature, or create an N/N Spirit token.

Go tall to break a board stall. Go wide to play around a removal spell.

Mobilize

Attacking creates that many tapped attacking 1/1 Warriors, sacrificed at end of turn.

The tokens expire, so the value is in the payoffs. Only take mobilize creatures highly once you have pump, anthems, or death triggers.

Flurry

Triggers when you cast your second spell in a turn.

Needs real fuel: cheap spells, cantrips, cost reducers, harmonize, or self bouncing dragonstorms. Without them it is a dead ability.

Hybrid generic mana

Symbols like two-or-green let you pay two generic instead of that color.

Quietly the most important text in a wedge set. It lets colorful cards slot into decks that cannot reliably cast them.

The Five Khans

Every clan, with its real record Ordered by win rate

Attack wide every turn with warrior tokens, back them with anthems, deathtouch, and sacrifice value. Hold your tricks for combat so blocks get punished, then sacrifice the leftovers after damage.

The most drafted deck in the set. Deploy big creatures, then tap their power to recast harmonize sorceries out of the graveyard. The blue-red half also feeds spell counts for the flurry cards.

Aggressive tempo fuelled by casting a second spell every turn, via cantrips, cost reducers, and self bouncing dragonstorms. Close with a pumped threat before the slower decks stabilize.

Sticky creatures with graveyard renew abilities, with self mill turning into value rather than loss. It is also the best base for a five color dragon deck, because it has the fixing, the draw, and the interaction.

Counters and tokens: go tall to break stalls, go wide against removal. It finished last of the five and it is the only clan below the format average, so it needs the payoffs to be genuinely open before you commit.

The Dragons Themselves

Take on sight Chips are real win rates

The top of this set is genuinely top heavy. Every card below sits at the 93rd percentile or higher, and the first two are the only cards in the set above 68 percent.

**Ugin, Eye of the Storms**68.5%The single best card in the set. Nothing else is close.

**Jeskai Revelation**68.5%Four colors and worth every one of them.

**Ureni, the Song Unending**68.0%A five color dragon that simply ends the game.

**Dragonback Assault**67.5%Mythic payoff that turns a board into a win.

Marang River Regent // Coil and Catch — **Marang River Regent**67.4%The best of the omen dragons by a wide margin.

**Elspeth, Storm Slayer**66.8%Doubles your tokens. Mardu and Abzan both want it badly.

**Roar of Endless Song**65.4%Rare, and the meta update was right to call it a bomb.

**Sage of the Skies**64.9%A rare that outperforms most mythics in the set.

**Shiko, Paragon of the Way**63.5%Jeskai's marquee legend, and it delivers.

**Betor, Kin to All**62.1%The Abzan spirit dragon, best of the clan's top end.

The Mana Is The Format

A wedge set lives here

More games are lost in this format to mana than to card quality. The fixing is abundant enough that a bomb in a closed wedge is still worth keeping, because you can build a two color base and add a light touch of the missing colors. That only works if you take the fixing when it comes.

The pick order for fixing

Tri-lands deserve the same pick slot as premium removal when you are in those three colors. They enter tapped and that is the cost of doing business.
Gain lands beat any filler card that would make your deck, but not premium removal, premium commons, or genuine value cards.
Evolving Wilds, Dragonstorm Globe, and Sagu Wildling are the common fixers worth prioritizing. The Globe belongs mainly in late game dragon decks, where it also buffs them.
The Devotee cycle supplements the mana base, and the bodies are fine on their own, so play one even when you do not need the off color mana.
Embermouth Sentinel and Roamer's Routine are the last resort. A deck with good fixing should simply cut them.

Splashing, and when not to

Splash less when you are aggressive. Missing a curve turn punishes an aggro deck far more than it punishes a slow one, so only splash the genuinely powerful cards.
Worth it: Inevitable Defeat off a red-white deck. Not worth it: a weaker card like Bone-Cairn Butcher over an on color four drop.
Sequence your lands before your spells. Counter-unless spells see real play here, and an extra untapped land often just wins the exchange.
Bait the counterspell with a threat you care about less before you jam your best card.
Tapping carelessly loses games outright with this many multicolor lands. Plan the whole turn before you tap anything.

Sagu Wildling // Roost Seek — **Sagu Wildling**56.4%

Take These

The commons that carried Ranked by real win rate

Commons are what you actually get, so this is the list that decides most of your matches. Six of these ten were on a video's underrated list, which is why that list scored so well.

**Dispelling Exhale**57.9%Counter unless they pay. Play lands first against it.

**Piercing Exhale**57.5%The behold removal cycle is uniformly strong.

**Ainok Wayfarer**57.3%Fixing on a body, which is exactly what a wedge deck wants.

**Temur Tawnyback**57.1%Quietly excellent in the most drafted deck.

**Riverwalk Technique**57.0%Counter or pseudo removal. Flexible and cheap.

Leave These

Called overrated, and correctly All ten in the bottom quarter

This is the overrated list from one video, in the order it was presented, with the real numbers attached. It is the most accurate call in the whole series. Note the caveats below the grid: three of these are genuinely fine in the right deck, and a guide that flattened that into "never play" would give you bad advice.

**Breaching Dragonstorm**52.5%17th percentile. The dragonstorm cycle flatters to deceive.

**Corroding Dragonstorm**50.6%6th percentile. Highly specific situations only.

**Stormbeacon Blade**50.1%5th percentile. Equipment is a trap at this speed.

**Ringing Strike Mastery**52.5%15th percentile. Too slow for what it does.

**Delta Bloodflies**51.1%7th percentile. Looks like an evasive threat, is not.

**Reputable Merchant**52.8%20th percentile. Needs good Abzan mana to be worth it.

**Yathan Tombguard**51.2%7th percentile. A late pickup for a counters deck, not a build around.

**Tempest Hawk**53.3%23rd percentile. The most defensible name on this list.

**Highspire Bell-Ringer**51.8%10th percentile. The flurry payoff that never arrives.

**Hundred-Battle Veteran**49.6%4th percentile. The worst of the ten, and near the worst in the set.

The caveats worth keeping. Stormbeacon Blade is fine in a dedicated go-wide aggressive deck. Yathan Tombguard is a reasonable late pickup once you already have a counters deck. Reputable Merchant is fine with good Abzan mana and the synergies to use it. These are cards you should not first pick, not cards you should never play.

Play Better

Technique, not card evaluation

Combat and the stack

Snakeskin Veil is one mana for a counter and hexproof. If a green opponent holds one mana, expect your removal to be blanked.
Sarkhan's Resolve destroys a flier, which means it kills dragons, with plus three plus three as the fallback.
Cast Coordinated Maneuver after mobilize tokens exist but before blockers, so it counts them. That turns it into efficient two mana removal.
Attack tokens into Unsparing Boltcaster. A chump block still costs the opponent a real creature.
Assume a counter whenever blue leaves up two mana. Spectral Denial and Dispelling Exhale are both common enough to play around by default.

Drafting the wedge

Stick with an early bomb even when its colors look closed. The fixing in this set is good enough to get there.
Pair mobilize with entry and death payoffs. Shocking Sharpshooter and Venerated Stormsinger turn expiring tokens into free damage.
Hold your spells until combat with a mobilize board. The threat of a trick keeps blockers home, which is most of the value.
Two colors is a real destination, not a failure state. See the heresy section above.

The Numbers

General limited math Not set specific

These come from the one video in this collection that is not about Tarkir at all. They are hypergeometric facts about a 40 card deck with 17 lands and a nine-eight color split, and they hold in every limited format you will ever play. The splash number is the one that matters most in a wedge set.

Lands in a 40 card deck

Split nine to your main color, eight to your second

71%

Hit your third land on the play

Keeping a two lander. 85 percent on the draw

Sources to splash one card

The standard baseline

25%

Of games the splash is dead

By roughly turn seven with three sources

95%

Cast a single pip on turn four

Nine sources, on the play

73%

Cast a double pip on turn four

Same nine sources. This gap is why pips matter

66%

Double pip in your second color

Eight sources. A third of games you miss

58%

Find your second color by turn four

On the play, with none in your opener

The one to internalize. A single pip card is 95 percent to be castable on curve. The same card with a double pip is 73 percent. That 22 point gap is larger than the gap between most good and bad cards, and it is the real reason double colored top end cannot be splashed.

The Short Version

If you read nothing else

White-red is the best deck and nobody is fighting you for it. It has no clan, so nothing in the packs points at it.
Take the fixing. Tri-lands pick alongside premium removal, gain lands over filler, Mardu Devotee highest of its cycle.
Abzan finished last and is the only clan below the format average. Need the payoffs open before committing.
Mobilize tokens expire. They are worth nothing without pump, anthems, or death triggers, so do not take the creatures early on faith.
Flurry needs real fuel. Without cheap spells and cost reducers it is blank text.
Behold freely. Four dragons is enough, and the cards are close to rate even when you cannot turn it on.
Play lands before spells, and bait the counter with your second best threat.