How Pack One works

Start with a Draft Run.

Play ten real decisions from trophy drafts by high-quality, experienced 17Lands drafters. Every question comes from a separate draft, often in a different set. Your first decision is Pack 1, Pick 1. The second is Pack 1, Pick 2, with that drafter’s first card visible. Later questions come from pick-depth buckets: the final question might be Pick 9, 10, or 11.

Read the original drafter’s earlier picks, choose one card, and lock it. The reveal shows what they took and how your choice scored. Matching them earns 100; alternatives receive partial credit from contextual strong-player support.

Two rerolls for the whole run.

New set gives a comparable choice from a different Magic set. New pack gives a comparable choice from another draft of the same set. Each is available once, before you lock a pick. Earlier picks always belong to the current question’s original drafter.

Powered Cube has its own ten-pick game.

Every choice comes from a different high-quality Powered Cube trophy draft. The source omits full P1P1 packs, so Cube starts at P1P2 with the drafter’s real first card visible. Later choices use progressively wider pick buckets. You get two pack rerolls and no set reroll; both replacements stay inside Powered Cube. Cube has its own Daily, leaderboard, and friend challenges.

Come back tomorrow, or challenge a friend.

The Daily gives everyone the same starting ten decisions and resets at midnight Eastern. Your first attempt is saved automatically and can be resumed. Practice is unlimited and needs no account. A friend link opens the sender’s exact final packs; rerolling them continues as practice without a challenge win or loss.

Your Profile tracks games, sets explored, milestones, and Daily streaks. Create an optional account when you want to carry that record across devices. Public profiles are opt-in.

More ways to play.

Every Pack One game starts from a historical Premier Draft seat represented in the replay dataset. The cards in the opening pack and the later packs come from that recorded seat. You are free to make different choices, but Pack One does not pretend those alternate choices would have caused the rest of the draft to change.

That distinction is intentional. Pack One is a decision-comparison game, not a bot-driven draft simulator. Keeping the replay fixed lets two players face the same information and lets the model compare choices without inventing unseen packs.

Top 3 and Full Pack answer different questions.

Top 3 asks you to rank the three starts you most want from the opening pack. It is fast and especially useful for separating a confident first pick from a cluster of cards you consider close. Full Pack follows every decision in the first booster and weights later decisions less as the number of meaningful alternatives shrinks.

The reveal is comparison, not correction.

The strongest use of Pack One is to make the choice before seeing the model. On reveal, ask three questions: Did I identify the same group of plausible cards? Did I rank them differently? Was the model itself confident, or was the top of the pack tightly clustered?

A disagreement in a close pack is a different kind of learning signal from a disagreement where one card receives much more support than everything else. The score compresses the session into a useful number, but the shape of the support is where the study value lives.

Historical drafter choice stays separate.

Draft Run gives full credit for the original trophy drafter’s choice. In Top 3 and Full Pack, the historical choice is supplementary context and the existing consensus-based scoring applies. The consensus model is trained from a broader cohort of experienced, high-win-rate drafts, so the historical pick and the model can disagree.