Update log
What changed,
and when
Two things get logged here. The conditions readout, every time it is recomputed. And the published methodology, every time it changes, including the changes that removed something.
Readout last refreshed
2026-08-17
47 days ago
Method last changed
2026-08-14
Trend persistence window fixed at 60 days
Refresh cadence
Weekly
Recomputed on the same day each week
Entries
- 2026-08-17Readout refresh
Conditions readout recomputed
Weekly recompute across all seven inputs. Data completeness 100%. No change to the input set or the thresholds.
- 2026-08-14Methodology
Trend persistence window fixed at 60 days
The window was previously described as 'roughly two months', which is not a specification. It is now stated as 60 days everywhere, including on the regime page.
- Published measure changed from '~2 months' to '60 days'
- Regime page input table updated to match
- No change to how the number is calculated
- 2026-08-10Readout refresh
Conditions readout recomputed
Weekly recompute. One venue returned partial hourly depth samples, so liquidity depth used the reduced sample and completeness printed 98%.
- 2026-08-07Methodology
Blank readout below 70% completeness
The detector now prints nothing when data coverage falls under 70% for the window, instead of labelling the window from a partial sample.
- Coverage threshold introduced at 70%
- 'Insufficient data' added as a valid published state
- Example output D added to the regime page to show it
- 2026-08-03Readout refresh
Conditions readout recomputed
Weekly recompute. Data completeness 100%.
- 2026-07-29Methodology
Confidence band removed
An earlier draft of the readout carried a 0 to 100 band. It was removed. A number like that reads as a strength-of-signal reading, which is the exact thing this site says it does not publish.
- 0-100 band deleted from the readout and from all copy
- Replaced with plain condition labels: low, moderate, high, rising, falling
- Nothing in the archive was edited; older entries note the removal
- 2026-07-24Site
Regime page published
Full input list, measurement definitions and worked example outputs made public, so the detector can be rebuilt without a subscription.
Education about trading process. Not investment advice. Nothing here is a recommendation to buy or sell anything. Most retail traders lose money.