Velocity versus distribution: which one actually trips the wire?

Velocity versus distribution: which one actually trips the wire?

When a profile gets flagged, two suspects are usually present at the scene: a skewed anchor distribution and an unnatural acquisition velocity. The literature tends to blame distribution, but the temporal dimension may carry more weight than its share of the discussion implies.

Consider the mechanics. Penguin moved to near-real-time in 2016, processing links granularly rather than in periodic refreshes. A granular system is structurally better positioned to notice that forty exact-match anchors appeared in nine days than to make a holistic judgment about a steady-state ratio.

— Distribution is a snapshot.
— Velocity is a derivative — the rate of change in that snapshot.

On one hand, a 5 percent exact-match profile built slowly over three years reads as plausibly organic. On the other, the same 5 percent acquired in one quarter, all from the same anchor string, presents a sharp temporal signature even though the ending ratio is 'safe.'

Limitation: we cannot observe the system's actual time-windowing, and most public penalty post-mortems lack clean acquisition timestamps, so velocity claims rest on inference.

Open question: do safe-ratio profiles still get devalued when the within-anchor velocity spikes — and if so, is the durable advice 'manage the slope,' not just the share?