Magic Arena offers two forms of Constructed Magic. You can play best-of-one matches, or you can play best-of-three matches.

There are three key differences between these formats. Of these three key differences, the least important is that the best-of-three matches allow sideboards!

A second is that the players of best-of-three matches are much stronger players, using much stronger decks. Often I’ll face a sea of red in the best-of-ones, then a sea of Dimir in best-of-three. This makes sense. The matches are more expensive to enter, they require building a sideboard, which is expensive, and they give much better payouts. It is the natural equilibrium for better players to compete in the format with better prizes, driving the less capable players to the place with lesser prizes.

Despite the worse payout structure, my experience says that it is much, much easier to turn a gold-based profit in the best-of-one Constructed event, rather than Competitive Constructed, despite a break-even point that’s over 10% higher.

The more interesting difference, the one I want to focus on here, is that best-of-one games adopt a new opening hand rule: You draw two hands, and it selects the hand with the amount of lands closest to the average for your deck.

I will focus on the question of what changes for players, and ignore for now the question of whether this is a good idea.

This is a dramatic difference in the distribution of opening hands.

For a 24-land deck, by default you draw a two-land hand 26.9% of the time, a three-land hand 30.8% of the time, and a four-land hand 19.6% of the time. That means you’ll get an “acceptable” land count 77.4% of the time and usually keep, and most of the remaining 22.6% of the time you’ll mulligan.

With the choice of two hands for land count, the chance of a 0-, 1-, or 5+ land hand declines from 22.6% to 5.1%. The chance of exactly a 3-land hand increases from 30.8% to 52.2%. Realistically, you go from a 25%-35% mulligan rate to a 7%-20% rate.

It also means that 21 and 22 lands become very different amounts, since 21 or fewer land decks will favor 2-land hands, and 22+ land decks will favor 3-land hands. Decks that are hoping to stop at 2 lands can cheat on their land counts—if you run 20 lands, you’ll get 2-4 lands 92% of the time and 2-3 lands 84% of the time. Even cutting down to 16 lands you’re still 84% to draw 2-4 lands, as opposed to 59% under the old rule. New possibilities abound.

This also punishes decks that use spells that double as lands. If you play Llanowar Elves and Druid of the Cowl, or especially cards like Adventurous Impulse or Flower // Flourish, or even spells like Opt and Anticipate, they all count as spells, causing the algorithm to choose the “wrong” hand. On top of that, the “insurance” you are buying has decreased in value. You’re doing lots of work to set up a deck that can find the mana it needs when it doesn’t have enough, or turn extra lands and mana sources into valuable spells. Opponents not doing that work now get a lot of that work for free.

The live experience of playing changes dramatically.

In a regular game, you must plan to operate on six cards, or face six cards, often. Keeping a solid opening hand feels really good under normal rules. With the new rule, it is expected that both players will keep most of the time, and most games will involve both decks operating largely as intended. This also means that hands that punish poor starts, or can’t respond well to fast starts, are much worse.

Mulligans are so rare that they are highly informative of the opponent’s deck choice and skill. Better players mulligan hands with good land-spell mixes far more often, better players play control more often especially in Arena, and control decks have far more draws that can’t be kept despite a reasonable number of lands. They have more colors, they are more exposed if they don’t interact in the early turns, and they recover well from mulligans. With opponents operating more reliably and efficiently, keeping slow but playable hands becomes much worse, increasing mulligan rates further.

An interesting corner-case is that the percentage of dual lands in your deck becomes helpful, punishing basic lands in 2-color decks in a new way. If I add another basic Island to my deck, that actively decreases how often I’ll have white mana, because I’ll be selected into hands with that Island that lack other lands. Replacing an Island and a Plains with a Meandering River and a spell substantially improves my colors at lower risk than before to my mana count.

At one point, I was playing a deck that was highly disadvantaged against true control decks in a way that being up a card would not fix, and I was sad when my opponent went to six cards. Knowing they had taken a mulligan actively lowered my win rate. As I tuned the deck, my control matchup got less awful and now I’m indifferent. Either way, if they are going first and don’t keep, I mulligan assuming they are on control.

The rule’s wording is ambiguous, so I don’t know for sure if they do the same thing with six-card hands, but given how rare it is that anyone goes to five cards, I’m going to assume that they do use it. This means that going to six cards is much, much safer and stronger, since the chance of a mulligan into oblivion shrinks so much. With a 24-land deck, the chance of a 0-, 1-, 5-, or 6-land, 6-card hand goes from 25.2% to 6.4%, and the chance of not getting two lands or more on 5 cards declines from 32.7% to 10.7%.

Introducing this rule makes not only mana but the rest of everyone’s decks vastly more consistent. It is unclear whether this is more or less generous than a free mulligan to seven. It is less generous than a free mulligan into a hand guaranteed to have 2-4 lands in it, but has almost the same effect.

Play is more reliable and game play speeds up a lot, and players build their decks assuming they have 3 lands and 4 spells at their disposal—and their opponents have the same. Beating good draws is the new order of the day.

Win rates of good players against bad players, and for players with favorable matchups, dramatically increase. It’s rare that I lose to an obviously poor player, or to an unfavorable matchup, in these best-of-one games. It’s not as obvious as you might think that the best-of-one is less skill testing than a full best-of-three would have been.

I continued to discover new implications throughout writing this, so I’m doubtless missing many others, in addition to the ones I’m excluding for compactness. What other changes have you noticed or discovered?

Then there’s the other question. Are these changes a good idea? I look forward to a lively discussion. These questions aren’t getting the attention they deserve.