Ship-to-Ship Transfer Has Retaken the 30-Day Lead. 95.5% of Dark Manipulation Alerts Are Sub-Max Confidence. The Combined Queue Is 76,315 Alerts With Zero Reviews.

The Setup

Six weeks ago, AIS_manipulation_dark claimed the title of dominant behavioral motif, surpassing ship-to-ship transfer (STS) for the first time by volume. The June 23 analysis documented 26,246 dark alerts with zero acknowledgments. STS had fallen to second place.

That ranking has reversed. Over the last 30 days, STS filed 24,191 alerts. Dark manipulation filed 23,474. The gap is 717 alerts — narrow, but the direction matters: the motif that was displaced has clawed back to first.

The Chain

The two queues now sit at 36,822 (STS) and 39,493 (dark) lifetime alerts. Combined: 76,315. Acknowledged: 0.

The composition tells the more interesting story. STS operates at 1.000 average confidence and 1.000 robustness across all 36,822 alerts. Every firing is treated identically: maximum certainty on both signal dimension and structural path quality.

Dark manipulation runs at 0.820 average confidence. Looked at another way: 37,724 of 39,493 dark alerts — 95.5% — fired below maximum confidence. The motif produces high volume but with degraded signal certainty on nearly every alert.

In the prior seven days, dark filed 4,486 alerts; STS filed 4,864. Both are in week-over-week decline from their prior windows (dark was at 5,468; STS at 5,959), but STS is pacing ahead on current volume while running at higher per-alert signal quality.

The Implication

The compliance queue now holds two distinct tiers of alert quality in the same pile. STS alerts carry maximum confidence and maximum robustness. Dark manipulation alerts are largely sub-maximum on confidence while matching STS on robustness — the structural path quality is equivalent; the evidentiary certainty is not.

This matters for any prioritization scheme. A first-in-first-out queue treats a 0.82-confidence dark alert identically to a 1.0-confidence STS alert. A confidence-weighted queue would process roughly 2,100 STS alerts per day at full weight while discounting 95.5% of dark alerts. The queue has accumulated for long enough that this distinction is no longer academic — it is the difference between a prioritization pass that takes days versus weeks.

What to Watch

STS has returned to the lead on recent volume. If the 30-day trend holds, STS will close the ~2,671-alert lifetime deficit to dark manipulation within six weeks. The more actionable signal: whether dark confidence stabilizes above or below 0.820. A move toward 1.000 narrows the quality gap; a further decline deepens it.

The week-over-week decline in both motifs (-18% for dark, -18% for STS) suggests the raw filing rate may be plateauing. Whether that is a genuine behavioral shift or a data-ingestion artifact is not resolvable from alert counts alone.

Limitations

Confidence and robustness scores are model outputs, not ground truth. A 0.820-confidence dark manipulation alert that has been operationally validated is more actionable than a 1.000-confidence STS alert on a vessel that has since changed flag. No analyst has touched either queue. Confidence tiers are a triage scaffold, not a verdict.

Overwatch tracks motif_alerts across behavioral classes. Signal counts and confidence values are drawn from the live queue as of 2026-07-10. All figures are trailing 30-day and all-time unless noted.