One keep as an experiment

A series of hits on one city is the closest thing battlefield data has to a controlled run: same target, ordered time, the wall and the reins changing while your recipe repeats or shifts.

By hand this is the nightmare job: open every report that touched that keep, write down the clock, the attack type, whether they healed, who was inside, what layers were left, then try to see the night as one story. Teams do not do this. They remember the last hit and a couple of callouts.

The Analyst can sit on that keep and answer questions across the whole imported sequence, in minutes.

Focus

Say the keep and the vantage:

You may get a hit-sequence timeline on the way in. That is a map of the experiment. It is not the writeup of every question below.

Why this used to die as a daydream

You cannot hold twelve layer snapshots and six rein swaps in your head. You cannot see whether the third siege wave was the one that made the later ranged “free” without lining the reports up. Opening them one at a time loses the gaps — the heal, the reloc, the reins that left.

Questions (any wording)

What the sequence was

What paid

What was wasted

Defender-side (teammate city)

What a good answer looks like

Order. Clocks or battle index. A change named in the gap between two hits, not a vibe. Separate “this rally scored points” from “this rally made the next one possible” — that second sentence is a reading, not a sum. Do not treat a missing report in the middle as “nothing happened.” Numbers, and the reading.

Common miss

Asking for “the keep analysis” with no city and no vantage. Or taking the timeline screenshot to another tool to narrate it. Stay on the keep and ask the next question. Near-replicates and rein economics are the usual follow-on families.