THE DECK LAB / REPEATABLE EXPERIMENTS
Test the idea.
Read the tradeoffs.
Compare deck recipes under different CPU play styles. Look for one-sided matchups, slow games, and cards whose abilities create useful choices. Then take the interesting questions into a human match.
Set up an experiment
Loading the card pool and saved report.
SAVED EXPERIMENT
Results
A saved report will appear here when available.
CPU results test these recipes and policies. They do not prove human balance, fun, strategic variety, or a 10–15 minute playtime.
Matchup matrix
Row deck’s paired score: a win counts as 1, a draw as ½. Bounds are conservative 95% intervals across complete seed pairs. Mirror cells compare starting seats, not deck strength.
Deck results
Starting order & pacing
Pacing below combines all CPU styles in this report.
A turn is one player’s turn. Actions include individual engine decisions. These counts do not estimate minutes at a human table.
What did the cards do?
Reached hand counts unique physical cards. Drawn counts hand arrivals, including redraws. A choice is a unit-turn with both an attack and its ability available. Counts include recorded actions from capped or errored games.
Replay a result
Inspect the sampled match seeds, starting order, and recorded actions. Download the report to preserve the experiment and its recipes.