Designing for Draft

Recipes are for wusses. When I cook, I saddle up my trusty Hundroog and set out into the wilderness in search of rare spices, exotic vegetables, the finest meats and strange earthenware jars of condiment so arcane that they might not even be edible. Then, back within the gothic vastness of my kitchen I chop things with huge cleavers, sear them over roaring flames and add garlic to them whether they need it or not. This is the untamed frontier of the culinary arts. Cooking the same thing twice is the sign of a civilization in decline. Unless it's pad thai noodles. They're always good.

And yet for all my dark arts - if you don't have dark arts you can substitute light arts and a teaspoonful of muscovado sugar - there is a critical flaw in my method. If I one day cook the perfect dish, I will be unable to replicate it.

Wizards of the Coast did that with Ravnica Block draft.

Now at this point I should offer an apology, or perhaps a warning, or possibly a disclaimer. The reason being, I am about to tell WotC how to design Magic sets which draft well. Is this not a risible act of hubris on my part? (If you're too lazy to look up "hubris", just make your serious face and nod gravely.) No, it is not. Allow me to explain. After all, like a supervillain in a videogame cut-scene, you can't actually stop me. Nobody knows more about Magic than WotC. Indeed, in the history of games there has seldom if ever been so much effort devoted to the refinement of a single game. However, in order for it to be useful for me to offer opinions on its design I do not need to know more about the topic than WotC do. Here in the third millennium - quick pause whilst we all put our mirrorshades on - we recognise that crowdsourcing has the potential to crack difficult problems in a way that even the most brilliant individual would find hard to match. So I offer my opinion, hopefully you offer yours and WotC may find things of value in the resulting discussion even if half of what I say is nonsense.

Ravnica block, you see, had enough garlic. And at the time I thought it was the dawn of a new age. I foresaw large chunks of my future disposable income being spent on drafting. I even imagined twenty years from now, late night debates in smoky bars as I slammed my beer onto the table, rose from my chair and pointed at my drinking companion, decrying his choice of first pick. But I know now that will never happen. They've banned smoking in bars.

Perhaps more importantly, it has since become clear to me that WotC got lucky with Ravnica. Certainly a huge amount of skill and thought went into its creation, but the unprecedented playability of the Ravnica-Guildpact-Dissension draft format was, in restrospect, largely down to chance. Partly I began to suspect this as a result of playing later draft formats - more on this later - but my suspicions were confirmed when Aaron Forsythe later wrote an article entitled Making Sealed Work for WotC's "Latest Developments" column. In it, he discussed some of the ways in which WotC manage to avoid making unplayable Sealed formats. But he also said this:

Draft doesn't typically require a high level of precision on our part as it is a "self-correcting" format. If red and black were far and away the best colors in a draft format, many people at each draft table would try to force those colors, leaving the door open for those few that are drafting white, blue, and green to get decent decks.

Aaron's basic point here is correct. But in making it, he has revealed an implicit assumption: that a draft format should be considered healthy if it is about as playable as Sealed. I respectfully disagree. Certainly many historical draft formats have fallen into this category, but at its best draft is something far greater and my assumption in writing is that WotC are leaving much of the potential of draft untapped. Why should they care? Because since the end of RGD I've been drafting far less. I'd like to be drafting more. Or to put it another way: Dear WotC, I would like to give you lots more cash money. Please help me to do so.

Known Elements of a Good Draft Format

I don't like coffee. The smell is great, but sadly it tastes nothing like it smells. I sometimes wonder if coffee drinkers have some genetic variation which stops them from tasting that awful spit-this-out-it's-poison bitterness? And some people don't like olives. Some people don't like anchovies. Some very weird people - who by the way are objectively wrong - don't like chocolate. The point being that tastes vary and the same is true of draft formats. Those people who don't like chocolate absolutely loved triple-Coldsnap draft. Couldn't get enough of it.

But hopefully you all got the need to burble about subjectivity out of your system as undergraduates and will just shut up whilst I get on with making my list? Those members of the audience too young to have done degrees yet, trust me when I say that the ability to burble about subjectivity will get you through any student debate ever and all the hot chicks will want to get in someone's pants. Not necessarily yours, because your attractiveness is subjective.

Non-Obvious Archetypes
That screeching sound you just heard was the flow of this subsection grinding to a halt as I realised I can't assume you know what a draft archetype is. Nuts.

When you draft 45 cards and pick some of them to play with, that's a deck. A draft archetype (I hear Jung is big in Standard right now, so you should see the etymology) is a way to describe a style of deck draftable in a particular format. In some formats there are archetypes which can be described purely by colour, such as "mono-Red" in triple Zendikar draft. However, more generally an archetype is not just about colour. For example, R/W Samurai in triple Champions of Kamigawa draft did want to run some cards which were not actually Samurai - removal for example - but did not want to run just any old Red or White card for its colour alone. Sometimes the same colour combination supports more than one archetype (such as U/B in Shadowmoor, which could be aggro or control) and sometimes a particular colour combination just isn't good enough to support any viable archetypes at all.

The key set design point is this: A set is better if not all of its archetypes are obvious.

An easy corollary to this is that not all archetypes should simply be colour combinations. This is one reason why multicoloured sets tend to play well (Invasion block, Ravnica block) - because more colour fixing means more flexibility in card selection, which opens up archetypes. It's also a reason why Shards of Alara was less good. With so many multicoloured cards and each Shard more like a colour, options for archetypes are again reduced.

As Few Broken Cards as Possible
Magic players are quite a melodramatic bunch and love nothing better than to label cards as "bombs". Cards significantly above the average power level of a set are fine in Limited and not intrinsically problematic. To quote Mark Rosewater's seminal article When Cards go Bad:

I think the best reason to have a diversity of card power is it increases the skill in the game. A lesser player, for instance, is more likely to draft a sub-optimal card. A more novice player will put questionable spells into his decks. Both occurrences increase the chances that the better player will win.

And the flipside, which he doesn't mention, is that good players are more likely to be able to apply the resources available to find solutions to very strong cards the opponent has. This leads us naturally to the problem with broken cards. A card is broken from a draft perspective if it typically renders the rest of the game in which it is played irrelevant. A broken card is not necessarily good as such. An overcosted card which wins you the game on the spot is broken, but might not be a high pick for an experienced player who would be aware that they would seldom succeed in casting it.

Fortunately, WotC's most broken cards are often rares or, in the modern era, mythic rares. This doesn't make them somehow acceptable, but it at least diminishes the frequency with which they turn up.

Key point: Games should be decided by a single card as seldom as possible

Note that's not the same as ended by a single card. If I deal fifteen damage to you and then cast Lava Axe you have no grounds for complaint.

Nor are broken cards immediately obvious. For example, Mirrodin's Loxodon Warhammer probably seemed as though it could be fair in a set full of artifact removal. As it turned out, it was broken, but it's not obvious WotC condoned this. It's possible they didn't realise. (I am much less forgiving of Behemoth Sledge, because WotC must have known how it would play.)

Incidentally, special hat tip to Zendikar for being the draft format with the lowest frequency of broken cards in history. Long may it continue!

It Must Be Possible to Draft Against Specific Strategies
One of the best designed cards I've ever seen for Limited is Rain of Rust. Why? Because it's too expensive to be good. This is great, because it means that a player who values artifact removal highly can turn this valuation into a strategy. If all the artifact removal in the set was efficient then it would be picked highly and that player's valuation would be irrelevant.

This doesn't just apply to removal. It also applies to evasion, to cheap attackers, to fat creatures, to mass removal, to card advantage and - last but not least - to whatever special considerations may arise in a particular block. It's vital to allow each player to let their valuations shape their draft. That's right at the heart of what draft is about.

Key point: A set's weaker cards should support as diverse a range of priorities as possible

Historically this has not been a strong point for WotC. And when I say "historically" I mean throughout the history of Magic. Indeed, it is also Zendikar's number one shortcoming and has been a source of frustration for many of the players who don't like the set. Beaten to a delicate mousse (with a raspberry leaf on top and a little swirl of coulis) by very fast aggressive decks, they desperately want to draft cards which punish this strategy. Unfortunately they can't, because WotC failed to print any that weren't first picks.

This again expands upon Mark Rosewater's point about bad cards. A bad card should not be randomly bad. It should invite players to value it more highly than it merits by virtue of having a purpose. There is no one card in Zendikar we could point to and say "this one is the problem". It is collectively that they fail to provide sufficiently diverse options.

A Set's Themes and Mechanics Must Create Interesting Choices
I'm going to cheat a bit here and assume you've read WotC's two articles on the subject, again by Mark Rosewater. If not, suffice to say that WotC are well aware that choice is a key aspect of Magic's gameplay.

In the second of these articles, Mark Rosewater is defending the M10 rules changes from a horde of peasants with pitchforks and burning torches when he provides an excellent example:

Pre-M10 rules changes, you blocked, put damage on the stack, and then sacrificed your Sakura Tribe-Elder for a land. Sure, you had other things you could do, but all of them were eclipsed strategically by this play. Under M10 rules, the same thing happens. What do you do? It's not so clear.

I'm not interested in talking about M10 here. Why this example is so good is that it perfectly sums up what goes wrong with choices in Magic. If, most of the time, it's obvious which choice to make then the presence of the choice is not interesting. Take a look at this list of cards: Errant Ephemeron, Shaper Parasite, Sprout Swarm, Inner-Flame Acolyte, Earthbrawn, Burn Trail, Oona's Grace, Court Archers, Drag Down, Deadshot Minotaur, Island.

If that last one didn't seem out of place, I expect you spotted the pattern. The list contains one highly playable card from each set since Ravnica. In each case the card uses a mechanic from the set intended to provide the player with a choice. And in each case, the choice almost always turns out to be obvious in practice. (In case you're still puzzled by the Island, it's the decision card for Zendikar's landfall mechanic.)

Now probably some of you leapt out of your seats at this point to say something like "But I both suspended and hardcast Errant Ephemeron! That's a legitimate choice!". And probably the rest of you are laughing at them. Not only because they're wrong, but also because people leaping out of their seats is inherently funny. Particularly if they spill their coffee or hit their head on something as a result. (Did that happen to you? If so, we want to hear about it in the forums so we can point and laugh.) It's wrong because in each individual situation in play it is almost always obvious which you want. There is a sweet spot somewhere in the middle where you topdeck your Ephemeron at five lands or so, but it happens too rarely to matter much.

Key point: As many cards in a set as possible should offer non-obvious choices in typical play.

This doesn't mean choices have to be explicitly on the card, of course. A high proportion of the choices in any badly designed set come from nothing more complex than the spot removal cards. In particular, whether to remove your opponent's best creature or wait for something better. And between that and mulligan decisions, that's already enough for good players to win a decent amount. Which refutes the oft-repeated argument that a format is good just because good players win.

Incidentally, WotC have designed some superb cards for choices. Dredge, Convoke, Bloodthirst, Replicate, Graft and Forecast were all absolutely fantastic for... wait a second... were those all in Ravnica block? Yes, yes they were.

Actually Channel and Ki Counters were great too, but I didn't want to spoil my own kick-ass rhetorical momentum, OK?

Trying to Improve Draft Format Design

None of the points above are unknown to WotC as such. Where shortcomings occur in these respects they are usually due to these matters being too low priority within the many and complex requirements that compete within the world of set design. What I do think probably remains unexplored is the idea of carrying out an explicit draft design phase during set design. I'm told if you want something done you should do it yourself, but I'm way too busy and in any case don't want to move to the US because all they do over there is drive 4x4s, eat whole cows and try to find clever ways to destroy their own economy. So I'm going to do the next best thing and try to come up with a rough outline of how I think things should be done.

Design for draft should happen after the set's concepts and key mechanics are devised, but before sealed pointing. The bulk of the cards in the set (the 75%+ which are not going to see serious Constructed play) should be designed as part of draft format design. Yes, that's a fairly radical change to the way things are currently done. And yes, I'm completely serious. And yes, I do think it's practical based on the limited amount we know about what happens inside R&D.

Just as Constructed formats are not made by intentionally forming tier one decks a card at a time, so draft formats should not be made by intentionally forming archetypes. The task is instead to avoid various traps likely to harm the archetype diversity in the set.

Any linear mechanic within the set - a term used by Mark Rosewater to describe one end of the synergy spectrum - will unavoidably form the basis of at least one archetype if the number of Commons which feature it is large enough. There are two ways to avoid this becoming a problem. One way is to make the linear mechanic large enough that it allows for a lot of flexibility in drafting, supporting multiple archetypes. Mirrodin's Affinity might have achieved this if designed differently, but as it was ended up being really a single archetype. Onslaught failed spectacularly, with the same decks played again and again. Lorwyn did a bit better by diversifying the colours in each tribe and adding cross-tribe synergies, but pure linear strategies were dominant, with the subtlety only coming in when things went wrong. You could often go 2-1 with a clever deck, but the 3-0 decks were all the same.

The other solution to linear mechanics is to print too few at Common to make an archetype viable, so that collecting such cards adds to a deck's power but does not define it. Planar Chaos did well here with its Slivers. The same block's Rebels missed the mark a little, ending up as a single archetype. Zendikar's Allies missed the mark in the same way, although it's harder to spot since the Allies span a range of colours. An argument could be made that G/R and G/W Allies are different decks. Possibly true, but each is in any case an inflexible single archetype in terms of the cards it wants.

The above may seem like a criticism of linear mechanics, but modular mechanics are just as potentially problematic. Here, the trick is to make sure that a card doesn't go in too many archetypes. A card that every deck values highly is bad for the draft environment, because you just have to be lucky to open it. Similarly, a card which is good in every deck that can cast it is not adding to the drafting experience. Some such cards are inevitable, but all other things being equal they should be avoided. Too many such cards and all archetypes are just colour combinations.

What makes a good card when designing for draft? There are too many factors to list. It would be a pointless exercise. Fortunately, WotC already have many talented designers who are quite able to assess cards for themselves. As such, I recommend scoring on a scale of 0-5 exactly as one would for Sealed pointing. Except that instead of assessing the card's power it is being assessed for its contribution to the draft environment.

Our 5.0 rating is here pretty much unattainable. The closest any real card would come to this would be Ravnica block's bouncelands (oh look, Ravnica block again), which were pretty much perfect. Each one was very strong in multiple archetypes, less strong but still very playable in others (when only half on-colour) and yet never broken because they only make mana. Synergy with the block themes was amazingly high... OK, I'll shut up now, you get the idea.

Our 0.0 rating is, however, similarly unattainable. It could be given only to a card which singlehandedly made the entire format significantly less playable. Accidents like Umezawa's Jitte don't qualify because it's a Rare and so comes up too seldom. And Sparksmith, despite being playable in only one archetype and overpowered did still lead to some interesting games, so definitely not a zero. I think the closest any card so far comes is Rolling Thunder. We can be fairly confident WotC will never make such an error again!

Grab yourself a tasty beverage. It's time for a quick test run! Using the above scale, evaluate the list of cards below. I'll give my answers (not necessarily better than yours!) inside the spoiler below, together with a bit of commentary. Aim to give castable filler that doesn't interact with the set much, like Cylian Elf, a score of 2.0 or so. A score of 4.0 should be given only to a card which makes you want to play the format it's in. You can hopefully work out where everything else belongs from there.

1) Caterwauling Boggart in Lorwyn.
2) Oona's Gatewarden in Shadowmoor.
3) Battlewand Oak in Lorwyn.
4) Hideous End in Zendikar.
5) Lose Hope in Fifth Dawn.
6) Grinning Ignus in Future Sight.
7) Blessing of Leeches in Betrayers of Kamigawa.
8) Caravan Hurda in Zendikar.
9) Herd Gnarr in Time Spiral.
10) Esper Stormblade in Alara Reborn.

1) Caterwauling Boggart 3.0 - This is a typical example of a reasonably suitable card for draft. The power is high enough to be playable, but not high enough to make the card a first pick. The ability is good in two ways. First, you do not necessarily want to play it on the first turn that you have the mana. Second, it interacts with the opponent's board position in a non-trivial way. If your opponent knows the format they can adjust their play to take this card into account. The tribal nature of the card is not as limiting as it might have been since the designer had the good idea of making it span two tribes. The card does not score higher for two reasons. First, the tribal mechanic still holds the card back here; maybe unavoidable, but we score cards on results not effort. Second, the card really only has one application.
2) Oona's Gatewarden 3.7 - This card is very clever! Its flexible cost makes it potentially castable for a lot of decks. It comes down early and has a high impact, but because it cannot attack this does not risk speeding up the format too much. The low cost also helps to smooth out mana problems, giving the caster time to draw a third land. Wither works perfectly here, allowing the card to have a high impact against far more expensive cards. However, whilst this is a strong control card it is never unfair since the opponent will always be able to attack into it with any creature, risking at worst a 1-for-1. Even for a control deck the card is not a top pick, which means that players can prioritise it higher or lower according to how they want their decks to function. This card's design doesn't really have any flaws, it fails to score higher because as a simple card it provides few options. That's fine. Some cards need to be like this to stop a set from becoming too complex.
3) Battlewand Oak 2.2 - There are interesting decisions that can arise with this card such as the possibility of holding back either a Forest or a Treefolk to squeeze more damage through. However, these will not arise very often. Some credit should also be given for the fact that this card is not a high pick and so helps to make a weaker tribe (Treefolk) into a viable archetype. Where this card fails is by being much too inflexible. It's only good in one archetype (G/B Treefolk) and just about makes the cut as filler in a couple of others. Worse, it only plays one way: you drop it as early as you can and swing with it. For its complexity level it does too little and does not add much to non-draft formats either.
4) Hideous End 1.2 - This card would score even lower were it not for the fact that Zendikar is light on removal. It fails in three ways. First, spot removal of this quality at Common is bad for draft because it's always first pickable and you always want to draw it. Second, this card offers an additional bonus on top of the kill. The last thing that needs doing to overpowered Commons is to make them better! Thirdly, the bonus given shows a certain blindness to the needs of the format. The format is too fast, so adding damage to removal is making the problem worse. If the card had to give a bonus is should at least have been +2 life for the caster (which could be given a vampires-draining-blood flavour, for example). Now don't get the wrong idea here; there are lots of reasons to like this card design, it's just that none of them relate to the health of the draft format.
5) Lose Hope 3.5 - This kind of card gets printed far too seldom. Instant speed interactivity is provided as it would be by a removal spell, but here the card power is much lower which (this key point again) allows players to decide for themselves how much they value the effect and prioritise accordingly. Because the targetted effect is weak, the addition of Scry is a welcome boost to the card's power. It also adds meaningful choice as well as a tension between wanting to wait for a better target or playing the card early to Scry your way to a better hand.
6) Grinning Ignus 3.3 Works well with the Storm archetype whilst being playable in other decks as a mana accelerator and providing some utility as a creature even when it's not useful for anything else. My one reservation is that I'm a little unsure whether the Sorcery speed only nerf was warranted. With mana burn still in the rules at that time, wouldn't it be fine for Red to get a reusable blocker at the cost of burning its controller? Adding complexity to remove a non-broken strategy seems wrong to me.
7) Blessing of Leeches 2.8 This is a good example of the correct way to design a weak card; and yes, WotC do need to design weak cards. It's obvious that it isn't very strong, but it's not obvious exactly how weak it is. Can you maindeck this? Again, each player is free to decide how much they value this, even if their answer turns out to be that they won't play it. The one change I'd like to see to this card is that it should be worded so that the enchanted creature gains the regeneration ability and its controller loses the life. That way you could also profitably play it on an opponent's creature. It is not always correct to add this kind of flexibility, but in the case of a weak card it can make the difference between an unplayable and a marginal card. The latter is preferable.
8) Caravan Hurda 0.8 This card scores very badly because it shows very poor awareness of the format. With Makindi Shieldmate at 2W and Pillarfield Ox at 3W any White player who likes big butts is already going to be satisfied. (If you found your way to this paragraph via Google I'm sorry to disappoint you but this is a Magic the Gathering article. That's the trouble with keywords.) Now a bit of redundancy might be appropriate if these kinds of cards were really the kind of thing the format's control decks needed, but in fact they're not. And even if they were, Caravan Hurda would still be a bad design since it costs an extra mana, way too much for the format, for the loss of a point of power. This is the wrong way to design bad cards. It adds nothing to the format at all, even for bad players. Personally I don't think the format needs a third creature of this type, but if I had to print a card like this I'd opt for 3+ power and Defender. It would still be quite bad, but at least there would be valid reasons why a player might run it in an emergency. Better by far would be to give it a defensive landfall ability, of which the set contains too few, but I don't wish to start designing my own cards here.
9) Herd Gnarr 2.4 Reasonable attempt, but this is an example of an category of error which turns up a lot. Here is a card which is not easy to evaluate and encourages interesting play decisions. It isn't very strong, but is flexible enough to go in a number of archetypes. However, given that Time Spiral contains an unusual number of ways to trigger it, why does this card not have Trample? Note that I'm not saying it should have Trample because it would become stronger. I'm saying it should have Trample because it would lead to deeper interaction with the set. Playable in several archetypes whilst being a potential high pick for a specific one is a very good thing. And it creates interesting gameplay when the opponent has to think about using removal on your Herd Gnarr in case it's going to be a 10/10 trampler next turn, whereas having a chump blocker is something which would often happen by accident.
10) Esper Stormblade 0.5 Horrible, horrible design. This card comes down early and easily and deals damage much too fast. There is almost never a reason to adopt any line of play other than playing it as soon as you can and swinging every turn. It usually flies. The cycle of which this card is a part caused huge damage to the format they were in. The resulting games were swingy due to being so fast and often involved little to no interaction or choices, just praying for a good opening draw or the right card off the top.

It's important to note that this approach to assessing cards for draft-friendly design can only be used once a set's card list is mostly complete, because the card's interaction with the rest of the set is critical. But even for a complete set, assessing each card is not the end of the story. It's equally necessary to check that a decent number of viable (or seemingly viable) archetypes exist. This is tricky because you can only really test it at a late stage but the design work to fix problems would need to happen quite early. And that's why individual card assessment is so important. If, after the first-pass design of the set, your cards are scoring well then there's a good chance the set will support diverse archetypes. All that's needed is a decent scattering of mana fixing in whatever form.


Conclusions in a bit, actually. First I need to talk about low-fat donuts.

For those of you fortunate enough never to have been near a low-fat donut, allow me to summarise: they're revolting. It doesn't take a genius to predict this. After all, donuts are all about fat plus sugar. Taking out the fat is like taking the wheels off a car. No, you don't end up with a flying car. However, if you happen to be on a diet then low-fat donuts won't help, because they're basically solid sugar. As such, for a long time I thought they were pointless. Clearly just a marketing scam aimed at the subset of very stupid people who also want to lose weight. It turns out I was wrong.

Low-fat donuts are perfect for a particular group of people: those who like donuts so much that even a completely broken donut is preferable for them to no donut at all.

And this is the thing about cards like Hideous End. There is, I'm quite sure, a subset of players who love having them in their draft formats despite the harm they do. The euphoric happiness they feel when they have one and blow away their opponent's critter far outweighs any costs of the card's existence. Making Common removal cost five mana or come with drawbacks would just annoy them. Sadly, WotC have to take this into account. But it can be done, which we know because WotC have done it in the past. Last Gasp was satisfyingly strong, but still fair. Oblivion Ring is a potent answer, but also manages to be interactive. Understanding the problem is the biggest step towards implementing solutions.

Similarly, my own biasses colour this article as you might expect. In terms of WotC's player psychographics I am closest to a Johnny/Spike. Although to be honest the main reason I want to win is because drafting is just way too expensive without a decent win rate. So perhaps really I'm just a Johnny. The point being, I am perhaps neglecting the needs of the unwashed Timmy hordes in petitioning for better draft formats. And yet, again there is no problem with satisfying Timmy's requirements whilst still designing good draft sets. I may not empathise, but wouldn't Timmy be happy so long as some archetypes beat down with huge monsters or blast people with massive spells? Surely it's not necessary for the whole format to play that way?

And that, ultimately, is my conclusion: WotC have already proved they have the design skills to make strong draft formats and to resolve the many challenges that entails. But they seldom do. I am therefore forced to conclude that they simply pay the matter too little attention. Whether this is because their internal process makes it impractical or because their awareness of the issues is too low I don't know. My response to the problem is to write this article. And then, having done so, to persuade those of you reading it to join me in asking WotC to make efforts to improve matters.

Please do comment in the forums - I'm sure I have made more errors in my analyses than WotC ever would. And please do contact WotC, point them at this article and ask them politely if they'd like all my money!


Posts Quoted:
Clear All Quotes