Each metric row uses the raw values in the table below. For that row only:

share_A = value_A ÷ (value_A + value_B) · share_B = 1 − share_A (50/50 if both are zero)

The headline score is the weighted sum of those shares, scaled to 0–100 and rounded:

score_A = round(100 × (w_1×share_A,1 + w_2×share_A,2 + …))

Weights by interval

• 7 days — Clones (7d) 50%, Views (7d) 30%, Momentum (7d vs previous 7d) 20%
• 30 days — Clones (30d) 50%, Views (30d) 30%, Momentum (30d vs previous 30d) 20%
• All time — Total clones 50%, Total views 50% (no momentum)

Momentum is clone growth: (clones in the window − clones in the prior window) ÷ max(prior window clones, 1). The table shows it as a percentage (×100).

Very high momentum (e.g. thousands of %) is expected, not a calculation bug: the prior window had few or zero clones and the current window has a recent spike. Example: 819 clones this period vs ~0 last period → about (819 ÷ 1) × 100 ≈ 81,900%. That means "traffic jumped from almost nothing," not "819× better than the other repo." After more stable history, momentum usually moves to ordinary ranges (tens of %).

Lead X pts is the gap vs the leader: |score_A − score_B| ÷ max(score) × 100. It is not "repo B has X% more clones than A".