Wednesday, December 17, 2014

Drafting Lands in Khans of Tarkir

This will look like a strategy article at first blush, but the strategy is one everyone who's drafted Khans of Tarkir more than a few times should already know, and that I assume has already been written about a dozen times; this post is more about the design that led to the strategy.

It is correct to take lands that produce at least two of your deck's colors highly, and not because the Tarkir's three-color decks need more fixing than the two-color decks we commonly draft in non-gold blocks. They do—and that's the reason we take them—but it's not why we take them highly.

There's a reactive reason, and that is quite simply that our fellow drafters are taking them highly, so if we want any, we have to take them sooner than we'd like. This demonstrates an ability to adapt and is a vital skill to foster both in Magic and in life.

But it doesn't tell us why anyone started taking the lands highly to set this chain reaction in motion. If we can understand that spark, then we can do better than react next time, we can lead. Fortunately, it's not a huge mystery. WotC increased the number of playable cards per booster quite a bit in Khans of Tarkir, so that we can spend picks on the spells we need to fill out our decks as well as the lands to help them run smoothly.

If you recall your first couple drafts, you quite likely found yourself with 30 or more highly playable cards. While it can sometimes be frustrating to choose what to cut, having options in deck-building and for side-boarding is a very positive thing. It's just not nearly as positive as having reliable mana. If you could trade the half-dozen cards you didn't end up main-decking for dual lands, your deck would be quite improved. The trick is that you have to actually draft those lands.

Going into your next drafts, you're aware that you don't need as many spells, and so rather than tell yourself you'll pick up some spare lands floating around late in the pack when there's nothing better for you, you can start taking those lands 2-3 picks higher than that. Except you're still taking them too late. Several reasons: The tri-lands are uncommon and will never table; Of the 10 common dual lands, three are relevant to your clan, but two are relevant to another clan, and there are likely two players in most clans anyhow; and there are one or two 5-color drafters at your table taking nothing but lands and the very best spells, regardless of color.

So just how early do you need to take lands? Depends on your deck, obviously, but the average three-color deck probably wants to take bombs first, lands second, removal and staples third. Sometimes you can skew that by the pack's direction; I might take lands lower in my first pack, hoping that sending strong signals will send more of my preferred land back in pack two, with the back-up plan of first-picking land throughout the third pack. Alternately, you could take land highest in the first pack, watching signals and keeping your options open so that you can pick a clan (or commit to 5c) in the second pack and perhaps pick up some strong spells while other players try to catch up on land. I'm not the expert and you should be reading Star City Games for the keenest drafting advice; I'm discussing all this to talk about its ramifications to Magic Design.

First, is to appreciate that percentage-of-playables for Limited isn't just a number that R&D has been fine-tuning for Magic Draft at large over the years, but a knob that they are keenly aware of and tweak when a given format shows a need to deviate from the norm. Avacyn Restored, for example, turned this number down, making drafters adjust to a new evaluation schema where bad removal was as good as you could expect, which trickled down to creatures mattering more, which meant stalemates became more common, and evasion hit an even higher premium than usual.

Innistrad saw more playables than average, which wasn't a fix for reduced card choices like it was in Return to Ravnica; Innistrad's large playable count increased variance in decks from draft to draft and enabled less obvious archetypes like Spider Spawning, but made side-by-side card evaluation more difficult. Because there were fewer bad cards, new players would make fewer terrible decks, while skilled players had to focus more on deck synergy. I'm completely undecided whether Innistrad widened or narrowed the skill gap because I see a number of ways it pushed in both directions. Would love to hear your thoughts on the subject.

Next, I'd like to talk about the interesting choice to provide the set's color fixing via lands rather than spells. Players are reticent to cut spells for lands (because spells are exciting and productive and lands just mess you up if you don't draw the right number), and they are almost as reticent to cut action spells for color fixing spells. Cards like Azorius Signet are strong enough, but I'm told R&D has moved away from 2cc mana rocks. Firewild Borderpost and kin made the cut because you could play them on turn 1. The Khans cycle including Mardu Banner is weak, despite producing a third color and being cycle-able; Morph fights with the banners both for the 3-drop slot and for late-game mana. So, it's important that they gave us 15 lands to help smooth our mana without necessarily eating up spell slots. Why include the banners when they're so badly positioned relative to lands? I think they knew the lands would become high picks and that unprepared players wouldn't have much recourse unless they did. They might also have been aimed at other formats, like Commander.

Finally, we haven't seen common lands be important picks since Zendikar (okay, the gates of Ravnica mattered, but they never got picked this highly). I dare say this situation shows that R&D learned the value of making land matter from that set, and are using that knowledge to add more variety to new sets.


  1. Good thoughts here. One more point that bears addressing:

    Khans is set up to have more defensive cards than usual to allow decks time to find their mana. That weakens aggressive decks, which were the only real reason to play a two color deck. You don't just want to make the cards more aggressive because that undermines the original intent, but if the aggressive decks don't win before their opponents start casting tri-colored spells, they're way behind on card quality.

    Except that having good lands means other drafters have to spend early picks on them while two-color drafters can pick up a disproportionate number of powerful but not game-breaking cards like Feat of Resistance, Tuskguard Captain, and Arc Lightning.

  2. For the record, I think gates were a higher priority in DGR (by far) than they are in KTK.

    That said, I think this is a very interesting article with a lot of design implications. I think a lot about the "power level profile" of each set/color/rarity, and they often tell you a lot about formats.

    If you imagine a graph of the power level of, say, the twenty commons of a given color in a typical set, it will vary quite a bit from set to set. White, for example, usually has an excellent common or two, followed by a very large drop. By comparison, the best Green common is usually much worse than the best White common, but the Green cards are more steady, with a large number of Green commons being average/playable/interchangeable. Note that I'm not saying that, in a vacuum, the power is unfairly distributed between White and Green, just that the power is distributed differently. [Long Mathy aside (feel free to skip): on one hand, the total power level is really the area under these curves, which suggests that balanced means that these two curves have equal area under them (assuming you can make sense of that), but the balancing problem is actually far more complicated than that, because it is only the cards that people actually play that matter. Thinking about this makes clear why so many sets end up with colors being lopsided, it is a very, very tough problem.]

    This differing pattern between power level of White and Green does weird things to the color dynamics at the draft table (at least until people start to figure a format out). The best common in the pack you open stands a much, much higher chance of being White than being Green, meaning that after a few picks, it is likely many people are in White and a small number of people are in Green (this is ignoring other rarities and colors for simplicity). This becomes especially problematic for the White players, as almost immediately the really good White cards will dry up. At this point, suddenly, the medium quality Green cards (like my beloved Centaur Courser) start becoming the best cards in the pack, and maybe people start moving into Green. Maybe many people abandon their early White picks, or stick to them to the detriment of everyone playing White.

    This pattern, especially in Green and White, repeats itself set after set, inevitably leading to pros eventually declaring that it is a viable strategy to just start drafting mediocre Green cards with your first pick and accept that many more will come.

    (continued in next post)

    1. WOTC is well aware of this, and when the great Overrun Debate came up after M12, Aaron explained the reason that they put it in the set was they felt they needed a first pickable Green card that wasn't a Rare/Mythic to slightly balance the starting colors of the players (particularly in a core set where many drafters may not know when they're supposed to abandon their first picks).

      The thing that most strikes me about Khans in this regard is actually the comparative power curves of the Commons, Uncommons, and Rares of the set. The commons are so bad compared to the uncommons that the vast majority of the time your first pick will be uncommon/rare (most often uncommon, in my experience). This is exacerbated by the fact that some of the most powerful commons are the 3 color morphs, which typically make poor first picks.

      I haven't heard them say it, but I would bet that WOTC has made this power distribution very deliberately, in order to increase the diversity of drafts. If you're usually first picking a common, a lot of your drafts blur together. If you're usually first picking a rare, I think that infuses a bit too much variance into the format, particularly as far as the power level of what you open. [My first pick was Aegis Angel, great, but yours was Grave Titan, *sad trombone*]. Uncommon seems like the sweet spot.

      Super powerful commons also lead to a lot less variety in game play. If I never see another Wingsteed Rider, Seraph of Dawn, Mist Raven,... it will be too soon. I think Khans has perhaps done the best job of any set I've seen at balancing these power curves across colors and rarities in a way that leads to the most varied, interesting, and balanced drafts.

    2. Only the most experienced players don't unconsciously lend more weight to rares than uncommons than commons, so it's fairly standard across all sets for a first-pick to be uncommon or rare. And that's a good thing, for precisely the reason you illuminated.

      I haven't pointed cards for more than a couple sets, so I can't back up your analysis, but my instinct says you're dead-on. Regardless, this is an excellent contribution. Thanks, Tommy!

    3. I'm regularly fascinated when people talk about the 'obvious' strength/weakness/whatever of particular cards and cycles from past Draft formats because so many conclusions that are obvious to one group are completely alien to another. This, DESPITE the theoretically homogeneous experience we derive from Magic Online.

    4. It was only a few years ago where you almost always first picked Doom Blade or Pacifism or the like. I really like the new take on that.

    5. Did I say something was obvious? I was using a lot of hindsight in my analysis to be sure (and none of the particular examples were intended to be dogmatic).

    6. "I think gates were a higher priority in DGR (by far) than they are in KTK."
      I was referring to more than just your post. Not saying you're wrong either.

    7. Yeah, the trick is finding the right green common effect to push to first-pick quality. The good thing about a card like Murder is that it's efficacy is limited by the size of your opponent's largest creature. So a deck with two Doom Blades is good, but it won't just wreck you. But what does a pushed Elvish Mystic, Centaur Courser, or Giant Growth look like? Seems dangerous to increase the reliability of ramp, pump, or cheap beefy creatures. A 4/3 for 1GG at common would be no better than Doom Blade in terms of win percentage, but much more frustrating to play against.

    8. It does help that they've gone away from Doom Blade / Murder.

  3. I've never played a game of Limited in my life, so I might not be qualified to discuss this, but I'm strongly in favor of a flatter power curve, like the one you describe for KTK and INN. With a flatter power curve, it's harder to build a horrible deck (so beginners feel more comfortable) and also harder to build an awesome deck (so experienced players feel more challenged). There's less variance overall, and synergy becomes more important, creating more interesting drafting decisions.

    I'm honestly not sure what the downside is here, unless it's that sets of this type are more difficult to develop.

    1. I think you're mistaken if you assume that newer players will do better in a set with a flat power curve and synergies than in a set with a sloping power curve.

      A new player can certainly decide which is better for their Black deck, Taste of Blood or Grave Titan. New players are not making choices randomly, they may just misestimate the power level of some cards. I think the best thing to do for new players is to make it so the majority of cards are about as good as they look. Then some percent are much better, some are much worse, and some are just sort of weird and people have to figure them out.

      Bonus points if it is easy for a new player to tell a card is better than they thought when it beats them. "Hey, that 3 mana 5/5 flying vigilance creature is kicking my butt, maybe I should pick those higher." Negative points if the card wins you games without the new player opponent being able to see it, e.g. Merfolk Looter. "Man, you get so lucky, you always have the card you need."

      [A certain type of experienced player absolutely loves cards like this, so I think it is okay to throw them a bone every now and then, but winning because your opponent didn't value Merfolk Looter correctly is a lot like winning because you understood how damage on the stack worked.]

      I've put a lot of time into designing an introductory draft experience for friends learning magic, my Core Set Cube as I call it. Here is an enormously outdated article I wrote about it ( ).

      It is outdated both in that I know use card from Tenth Edition through M15, and in a shift in a lot of my philosophies. Also, based on a surprising number of people being keen to exactly reproduce that cube, I've made it so that the rares are almost all super affordable. I actually haven't posted the most recent version of the list because I'm having trouble with White commons for the exact reason discussed above.

    2. I agree that a flat power level is not great for new or casual players. Tommy hit all the right notes, but one additional point I'll raise, particularly with gold sets like Khans, is that playing a correct mana base is extremely tough and new or casual players don't know how to build their mana base properly. If all the cards are at a flattened power level they are more likely to add colors to their deck. Khans has a flatter than normal power level and in the past few weeks at FNM I have seen many new players try the 5 color deck in draft with 3 of each basic land and 2-3 of the dual/trilands in their deck. Because they didn't know to prioritize the lands in their picks, they pick up Banners, as Jay said, but those are so far less useful then the lands. They rely on morph to hit their curve and never seem able to flip things.

    3. Totally agree with Nich - I think that multicolor sets and particularly wedge sets STRONGLY imply to newer players that playing 3 colors is fine. I think to make up for this mana fixing needs to be made more exciting for newer players. I don't know exactly what this means, but tacking "Sacrifice: Draw a card" onto Manaliths doesn't cut it.

      I have definitely heard newer players verbally getting frustrated with this format that gives them a bunch of cool gold cards, but in their minds doesn't let them play with them (or enough of them simultaneously).

    4. I don't actually consider Khans an especially flat power level. I think new players are getting sucked into playing too many colors because the format is not flat, because they look at a pack and look at the cards in their colors, but then see an Armament Corps or a Villanous Wealth and think "I could just splash that" instead of taking the Highland Game or Witness of the Ages they are supposed to.

      This is particularly exacerbated by the fact that if there are disciplined drafters around them, they'll see these high tier gold cards late in the pack, much later than they'd normally see that quality of card.

      As to the flatness of the power level, I agree that the worst cards are better than normal, and the best cards are worse than normal, but they way they are distributed doesn't feel particularly flat to me, except among the commons (which are not high picks).

      I think Innistrad and M13 are probably the flattest power levels I've seen, at least in recent memory, and given WOTC's overall depowering of commons (the pros and cons of which have already been discussed), I don't think we will see it again for a while.

      I should note, though, that I think experienced drafters tend to prefer flatter power levels because the correct choice is less often clear and more often influenced by subtle things. I don't think it is a coincidence that those sets I named, along with Rise, are fairly flat in power level and among the most popular with the experienced drafting crowd.

    5. As to the mana fixing in Khans, I don't think it is really possible to make a set with more. Then ten dual lands at common and the 5 uncommons is a ton, way more than has been done previously. Compare, for example, with Shards, which had 5 panoramas at common and the 5 tri lands at uncommon (oh and a worse version of the banners, but I think it is best to just ignore those).

      Indeed, there is some evidence they pushed too far, with the period where the dominant strategy seemed to be to draft five color decks. Ultimately it was this period, I think, that pushed lands so high in the pick order.

      If you want to both help new players and stop people from drafting 5cc, you probably need a system like not putting any fixing in the set and then at the end of the draft saying "Okay, now everyone pick three gates in addition to your basic lands."

    6. One of the lessons I've learned only through developing and playtesting is that a set that wants to support two-color decks should have very little color fixing, a three-color set should have two-color fixing, and so-on, with players generally running one more color than their cards explicitly produce. (Surely not a hard-and-fast rule, little is.)

      I know several players who love playing five-color decks in Tarkir, so it's not a bad thing that the tri-lands enable that, but the existence of those decks does cost the single-clan decks both mana consistency and spell power. I can't help but wonder if the set would have been better—or at least more focused—without them.

    7. One thing to note is that while players enjoy playing the 5 color deck, statistical analysis shows that it is actually pretty damn terrible.
      See this Ars Arcanum article

  4. This great discussion reminded me of another impact the priority of lands has caused. People are seeing good spells come around later than usual, which means we need to reevaluate our threshold for signals.

    And speaking of signals, I've observed that people have adjusted slowly to reading signals from the _land_ left in a pack. If someone passes me an Opulent Palace, that's a signal (intentional or not) that she's not in Sultai and probably not in Temur or Abzan either. But we're not used to reading much into the land that's passed in a pack, so we saw a lot of missed signals and unintentional signals early on. And I think we're still seeing some of that.