How to read a checkup report: 30 seconds, 3 minutes, 30 minutes
A checkup report is built in three layers: the conclusion in 30 seconds, the evidence in 3 minutes, the raw data in 30 minutes. What each layer shows, how grades and risk scores are computed, and the report's methodology limits.
The one-line conclusion: read the grade and the one-liner first (30 seconds), then the section cards (3 minutes), and only then the raw fields (30 minutes). Do not reverse the order.
30 seconds: grade, one-liner, two risk scores
The top of the report carries three things:
- Grade A–E: mapped from the larger of the two risk scores — A is 0–19 (healthy), B is 20–39 (watch), C is 40–59 (caution), D is 60–79 (high risk), E is 80–100 (danger).
- A one-line verdict: it only restates the stage in plain language, the grade, and the two risk numbers. No extra interpretation.
- Long risk / short risk: both are 0–100 scores; the higher side has the worse risk structure. Once more: a score is not a probability.
How the scores are built (formula version ck-1): start from the current stage's baseline (for example, distribution carries an 85-point long baseline, markup a 70-point short baseline); when a dump score exists, long risk is blended half stage baseline and half dump score; when there is no dump score, that section is marked empty instead of being filled with a zero.
3 minutes: the section cards
A full checkup has a fixed order of 11 sections: verdict, stage, long/short risk, ignition conditions, dump score, funding and positions, predictions and calibration, stage history, anomalies, methodology and engine version, next steps. At the 3-minute layer, focus on:
- Ignition conditions: where the dual-track rules currently stand (see Ignition and dual-track confirmation).
- Dump score: the five-dimension breakdown (see The five dimensions of the distribution signature).
- Anomalies: only pre-declared threshold checks, such as long risk reaching 60 points or the dump score reaching 70 points — no improvised judgments.
Sections without data render as empty; no fake numbers are filled in. When the daily quota runs out, the evidence sections lock while the first three sections, the methodology, and next steps stay visible — a lock is not an error.
30 minutes: pro mode and history
With the pro-mode entitlement, pro=true additionally returns the raw field names actually used and the formula version ck-1. Combined with stage history (the coin's past act changes) and the "predictions and calibration" section (which forwards calibration rows computed by the market module — the checkup never recomputes hit rates), a report can be drilled to the bottom.
Three methodology limits
- An archive is a snapshot: reopening a full checkup for the same coin on the same UTC day returns the archived copy, unchanged no matter how the market moves; quick checks always recompute from the latest state.
- Grade validity is not a win rate: the public validity block reports the maximum adverse excursion within 7 days after a grade (how far price fell at worst). Grades with fewer than 30 samples only say "not enough samples, no conclusion". That is a historical description, not a win rate.
- Data can be stale: when the engine or gateway has not updated for over 10 minutes, the report is still served but marked
stale=true, and the page shows the data timestamp.
Entry points: the site's quick check only forwards you; the full report lives in the web app.