If some lands and a spell
are all that you seek
your eyes will find swell
the loosest snap-keep
A while ago, the American Legacy metagame was such that any opener had to be able to deal with a Goblin Lackey. No Force of Will, blocker, or Swords to Plowshares? Ship it back. That's not the case these days, but individual matchups still have their focus, and often a specific card is worth more than a grip of others. The clearest example comes in hate cards, like Rest in Peace against Reanimator or Engineered Plague against Goblins. After all, that's what those cards are for.
It's harder when a card's purpose isn't so definitive. Would you keep six lands and an Oblivion Ring against Sneak and Show? What about Force of Will but no blue card against TEPS? These questions depend on the context of the rest of your deck, but matches are still swayed by individual trump cards, and knowing where and when they're important is a matter of data mining.
I felt this effect during the last Modern GP, just as I do every time I play Modern. Against Jund, I won if I had a Lingering Souls. Against Storm, I won if I found a hate card. In that matchup, Zealous Persecution became another hate card post board when the opponent invariably brought in Empty the Warrens. Often, knowing what cards are important and finding them is more important than lauded skills like sequencing and evaluating complex board states.
PV once defined mulliganing as, “I believe that I have a higher chance of winning the game with a new hand of one card less than with this current hand.” I like this definition a lot, as it's clear and removes emotion.
Once we have a definition, we can start looking at fallacies. You've probably heard someone say, “Well, it's a mulligan already,” before sending a hand back. “I might as well ship this hand because it has a dead card” is great reasoning if the goal of mulliganing is to have no dead cards, but that's a poor goal (see PV's definition). If the hand is an effective six, it should be graded against the average six you would be drawing to, but the dead card shouldn't be used as the decider that some people make it.
After all, many cards are weaker (more dead) than we think. Maybe you'll never get a chance to flashback that Think Twice, or maybe that Pillar of Flame doesn't actually kill anything, or maybe you won't have the mana for that finisher until turn way-too-late. Similarly, a dead card in the opener isn't necessarily dead over the course of a game. Perhaps someone ticks up a Liliana of the Veil, for example.
Most know that openers are evaluated based on a mix of mana, gas, curve, and whether or not the hand actually does anything. It is this last part that is most vague and trips people up. In general, I try to form a rough plan for the first several turns, and match that up with how well the rest of my deck can support it. If whatever plan I reach doesn't fit my deck, the hand does nothing and I'll ship it back. Consider the following hand:
This is both a snap-keep for 5cc and a snap mull for 4c Zoo. In an aggressive deck, the hand does nothing because aggro decks typically lack inevitability, and need a threat to apply pressure. The same hand might be ideal for a control deck that needs to simply buy time.
There are other considerations, of course. In general, the slower the format, the more time your hand has to draw out of a loose keep. In some Standard formats, it's common to keep hands where the first play isn't until turn three, giving slow hands time to draw business. In others, you might be dead by then.
The pacing of the matchup matters too, of course. In some control mirrors pressure isn't applied until both players have made plenty of land drops, and in those situations a grip of seven lands might be a strong hand. Mirror matchups with cheap threats and even cheaper answers, like the Mono-Red mirror, become wars of attrition, and it's more reasonable to keep a hand that's light on lands.
Individual Cards Matter
The more Magic I play, the more I realize how the identification of specific power cards impacts the outcome of a game. Many players keep solid-looking hands when they should be searching, which is one of the more subtle ways to lose before playing a land.
I first encountered this back when Bitterblossom dominated Standard. I didn't play Faeries, but I remember hearing that, in the mirror, you mulliganed to Bitterblossom or you lost to your opponent's. Later, when I was playing UG Survival, I realized my win rate went up dramatically if I mulled to the deck's namesake. Sure, there were other nut hands, but if it didn't make a cheap Vengevine or four, I generally shipped it back.
These days, if I mulligan low, I'm almost always looking for specific cards. Since I'm searching for a trump, winning on a mull to four isn't especially rare. In most of those cases, I had one card that blanked an opponent's entire strategy, such as a Trygon Predator against an Aether Vial. Once, I won on a mull to one against Dredge. I was looking for a Leyline of the Void, which had won game two, but instead found a single Verdant Catacombs. My opponent bricked on Narcomoeba for the first few turns, and I ripped the perfect land, land, Green Sun's Zenith (for Scavenging Ooze). My opponent Dread Returned an Iona, but it was too late, and the Ooze won the race.
Let's look at some examples.
I remember the following hand from a team draft where I was playing infect with a splash:
On the plus side, the hand featured terrific mana and a removal spell. On the other hand, it contained none of the four Plague Stingers that I'd built my deck around. In the end, my teammates convinced me to keep, but I didn't like it. I'd drafted with a plan to apply pressure and, while the hand was good in the abstract, my deck couldn't support a control role.
In the end, my keep didn't matter because Yuya casually finished 3-0ing before I was done with my second round. Fortunately, he was still willing to game the winnings.
UW Flash (Standard)
UW Flash is an engine deck. That is, it wants to chain a pile of cantrips together every game. It can operate on four lands, making the fifth land in the above hand almost redundant, though there's usually at least one Sphinx's Revelation.
Now imagine the same hand without the two Islands. Still very close to the seven, yet it's an amazing five. In fact, you could probably win a softer event starting with that five every game.
Softer event—That's a subject of its own, but I do adjust my mulligan decisions depending on who I'm playing against. I usually don't mention the idea because most people that do take their opponent's skill into account do it too often, and end up mulling to oblivion more against stronger opponents. That's no way to live.
But it is a strategy I use. In a softer field, I believe it's important to extend the range of your opening hands, similar to how you might in poker.
Take a second to examine these simplified graphs.
Against weaker opposition, your cards are worth more, and you'll win more often when you mull low, and more often when you keep a loose one with danger of flooding out. However, if you're keeping the same hands, you're going to mull to oblivion the same amount. As you can see in the above graphs, mulligans constitute a larger percentage of overall games lost against soft opposition than strong. In order to make that number go down, you need to broaden your range.
This is where it gets tricky. Part of the reason your win rate improves against a softer field is because you are, on average, keeping stronger hands than your opposition. This is why most don't consider changing their range, as it can do more harm than good.
Even if you master this skill, your win percentage won't skyrocket because of it. Your mull to oblivion rate might drop to, say, 3%, but you might pick up a 1% loss to having weaker hands. That's a 1% gain. If you play around 25 games in ten rounds of swiss, over four tournaments you're likely to pick up an extra game. It doesn't seem like much, but a single game is often the difference between Top 8 or not.
Now that we've gone over the “why” and “why nots” of expanding your range, the “how” is much trickier. In the first six to seven rounds of an Open or a PTQ, I'll keep any opener that's better than the average five, even if it's worse than the average six. Again, I can get away with it because 1) My cards do more work than they would against a stronger opponent and 2) My opponent's keep criteria are still looser than mine.
As the tournament progresses, your keeps should get more traditional. The right time to adapt is when you notice your opponents getting strong, which could happen earlier or later depending on the event. The important part is that as the opponent's range of possible hands tightens, yours has to as well in order to stay competitive.
Let's continue with the examples, but see how analysis changes post-board.
RUG vs. High Tide
First off, I'm not sure why we have Wasteland post-board against a deck with mono-basics. Even a Forked Bolt would be better than an actual do-nothing. Secondly, pressure is the name of the game against High Tide. If you can make the opponent go off early, a piece of countermagic might do something. If you allow breathing room, countermagic won't get there.
I'm looking for a Delver, a Tarmogoyf, or a hand that can find one as soon as possible, and I'll go as low as it takes. Nimble Mongoose loses its charm against a deck with no removal, especially in this hand with no cantrips. A fast Mongoose gets big on turn three, and even that is too slow for this matchup.
Goblins vs. Elves
Against an unknown opponent, this is a fine hand. Against Elves, it's lacking. Both Goblin Piledriver and Wasteland may or may not be cards, and Ringleader's best function is finding more Incinerators or reloading after a Pyrokenisis. Speaking of Pyrokenisis, that's one of those mystical power cards I was mentioning earlier. Even though it costs a card to pitch, Pyro usually takes three or four of the opponent's cards with it; meanwhile it destroys their tempo, buying you time to find a Goblin Sharpshooter. While the above hand has the tutors to find whatever, it can't do much before the Elves' critical turn (3-4).
Note that post-board identifying the power cards is much easier. With RUG, I knew I had to find the right threat. With Goblins, we now had a sideboard card to go looking for.
Caw-Blade Mirror (Pre-New Phyrexia)
The following hand came up deep in an online PTQ. I won the die roll and sent my opener back. My second contained:
I stared at the hand for a long time before keeping. Craig Wescoe was birding over my shoulder.
“What are you doing?!”
I understood his concern, as the hand can't cast any of its spells. However, with a matchup so dependent on sticking a Stoneforge Mystic, I liked this hand well enough. My odds of ripping lands are better than my odds of mulling and finding both Stoneforge plus lands, after all.
Naturally, I cast Stoneforge on two, Squadron Hawk on three, and Jace, the Mind Sculptor on four. My opponent matched my line, and the second Jace wasn't even dead. It seems like most of my stories end with me ripping well, but you should know that I too have received a bad beat, once. Or at least I assume I have. It's been a while.
More important than my general sackitude is that I've finally penned this article. It's been a long few years of keeping some that most would chuck, tossing others that most would keep, and mulling it all over in silence.