Choose a card
Draft complete
How to use it
Click a card to draft it. Seven bots pick at the same time, packs pass left in packs one and three and right in pack two, exactly as in a real draft.
Your pool, your color counts, and your curve update after every pick. At the end you get a read on whether you assembled enough playables.
The panel on the right is the part to watch. If one color is climbing and the others are not, you have found your lane. If two are level at pick eight, you are fighting someone and should commit.
What the bots are doing
The bot logic is deliberately simple and worth stating, because knowing it makes the signals readable rather than mysterious.
Each bot scores every card in the pack as its raw quality, plus a bonus for each card it already has in that color. It takes the highest score.
That is it. No hate drafting, no archetype knowledge, no rare drafting. And it is enough to produce the behaviour that matters:
- Early on, bots take the best card regardless of color, so early packs are noisy
- After three or four picks, each bot has committed, and it starts taking its colors over slightly better cards elsewhere
- By pack two, the bots to your left and right have settled, and what they pass you is consistent
That last point is what makes the exercise real. The signals are not scripted, they emerge from the bots doing something simple, which is roughly what happens at a real table.
The wheel, and the arithmetic behind it
A card wheels when it returns to you after going all the way around the table.
With eight seats and fifteen card packs:
| Your pick | Cards left in the pack | What it means |
|---|---|---|
| 1 | 15 | Everyone sees this pack |
| 2 | 14 | You are seeing the pack passed by your neighbour |
| 8 | 8 | Every player has taken one card |
| 9 | 7 | This is your pick one pack, wheeled |
| 15 | 1 | The last card, take whatever |
So at pick nine you see the same pack you saw at pick one, minus seven cards. Seven of the fifteen are gone and eight remain, and you get to compare directly against what you remembered.
This is the single most reliable signal in draft. If you saw two good black cards at pick one and one of them wheels, black is open. Not maybe open, actually open, because seven players declined to take it.
Practical habit: at pick one, note the second and third best cards in the pack. At pick nine, check which are still there. That comparison is worth more than any amount of general theory.
Pick order that transfers between sets
The specific cards change every set. The framework does not.
1. Bombs. A card that wins the game on its own if unanswered. Take it, worry about colors later.
2. Cheap unconditional removal. Answers anything, including their bomb, and fits in any deck in that color. This is the pick that keeps you open.
3. Efficient creatures on curve. The two, three, and four drops that make up the body of every limited deck.
4. Conditional removal. Kills most things, misses some. Still excellent.
5. Card advantage. Draw and value engines. Better in slow formats, worse in fast ones.
6. Combat tricks. One or two is right. More than that and you have a deck that does nothing on its own turn.
7. Situational cards. Sideboard material.
The two most common mistakes against this list: taking a creature over removal because creatures feel like the deck, and taking too many tricks because they feel skillful. A limited deck with six tricks and no removal loses to the first large creature it meets.
The numbers a limited deck needs
Forty cards, and the breakdown barely varies:
- 17 lands. Sometimes 16 in a very low curve deck, 18 with expensive cards.
- 23 spells. Of which roughly 15 to 17 creatures.
- Two colors. Splashing a third is possible with fixing and costs consistency.
- Curve: somewhere around 5 two drops, 5 three drops, 4 four drops, 2 to 3 five drops, and 1 to 2 above that.
The curve is the part people ignore and it decides more games than card quality. A deck of powerful five drops loses to a deck of adequate two drops, because it is dead before it casts anything.
Staying open versus committing
The tension every draft turns on.
Stay open in the first four or five picks. Take the best card, prefer removal, avoid gold cards that lock two colors. You are gathering information, not building a deck.
Commit by around pick five to eight of pack one. At that point you should have a main color and a strong idea of the second. Staying open past this is not flexibility, it is indecision, and it produces a pool with nine cards in each of four colors and no deck.
Reassess between packs. Pack two comes from the other direction, meaning a different set of neighbours. A color that was closed in pack one can be wide open in pack two, and if you are short on playables that is when to move.
The failure this simulator will show you fastest: finishing with fewer than twenty-three playables. When that happens, look back at when you committed. It is almost always too late.
What this version does not do
Stated plainly, because a tool that overstates itself is worse than a limited one.
It uses a generic limited pool, not a specific Magic set. The cards are archetypes, rated on the standard limited scale, chosen so the pack dynamics are realistic. What it teaches is signal reading, curve discipline, and pick order priorities, and those transfer to any set.
What it cannot teach is set specific knowledge: which archetypes a given expansion supports, which uncommons overperform, how fast the format is. That requires the actual card data for the actual set, and it arrives here once the card database is wired in.
Until then, this is deliberately a trainer rather than a replacement for drafting the real set.