About Draft Sense
What it is
Draft Sense is a real-time draft assistant for Magic: The Gathering – Arena. It watches your draft from the side and tells you which card in the pack a model trained on millions of real Premier Draft picks would take, given the pool you’ve already built — plus a running summary of where the deck is going (colors, curve, removal density, creature count, archetype signaling).
Why we built it
Most existing draft tools score cards in isolation: this card had a 58% win rate when picked, so pick it. That works for pack-1-pick-1 and breaks down at pack-2-pick-7 when the question isn’t “which card is best?” but “which card best fits the deck I’m actually building?” A graph model can reason about pool synergy, color commitment, and signal-passing in a way static per-card averages can’t. Draft Sense is the result of taking that idea seriously.
How it works
Under the hood is a graph neural network — cards are nodes, edges encode mechanical and semantic synergies, and each pick conditions on the running pool. The model is trained across many sets on millions of real Premier Draft picks, so its sense of what works comes from how thousands of drafters actually play the format, not from an LLM guessing at synergies. The overlay reads Arena’s log files locally and never talks to Arena itself — same approach most established Arena tools use.
The alpha-tester program
Draft Sense is in active alpha. We accept testers in exchange for unlimited free access plus a real voice in what we build next. See the alpha-tester program for what you get and what we ask in return.
Who’s behind it
Draft Sense is built by Mentalis AI, a Tennessee LLC. For founder background, company history, and what else we’re working on, head to mentalisai.com/about.