Core objects
CardRecord = (name, qty, role, behavior_tag, mode_tags, tags, notes)
Deck = multiset of CardRecord
Run = (deck, config, scores, tensors, timestamps)Offline-first scoring uses the fields you provide. Oracle text is optional enrichment (Phase 2+).
Composite score
DD = ( Ī£_i w_i Ā· D_i ) / ( Ī£_i w_i )
Intel = Ī£_i p^I_i Ā· D_i
Wis = Ī£_i p^W_i Ā· D_iWeights w_i and projections p are configurable.
Deck shape statistics
Entropy(counter) = - Ī£ p(x) log2 p(x)
Gini(qty) = dispersion of copy counts (singleton āThunderdomeā vs redundancy)Shape stats are important: a deckās identity is often more about distribution than power.
Overcommit tensor (v0.2-inspired)
A general tensor for āpaying a drain and converting it into advantageā. This isnāt limited to Mercy effects: it models any planned tempo debt.
Drain(L,r) = (1-r)L
U(mode,L,r) = -Drain(L,r) + Ī±Ā·Gain(mode)
Asset iff U > 0