02 — Research note
Changepoints.
Two related tests on VLCC tanker AIS and energy benchmarks: FOCuS changepoints on USO, and a Mahalanobis stall signal on listed tanker names against BWET.
5 October 2026
Research question
The note tests whether changepoints in daily AIS summaries on the Persian Gulf–Singapore route line up with tradable moves in crude and listed tanker equities, without assuming a parametric prior on how trade regimes shift.
Strategy 1 asks whether FOCuS/Page-CUSUM changepoints on route partitions predict USO swings. Strategy 2 asks whether stalled-tanker anomalies at named chokepoints predict relative performance of FRO, STNG, TNK, and DHT versus BWET.
02 — Method
Method and data
- AIS every 30 min
- Grouped by day
- FOCuS
- Page-CUSUM
- O(log n)
- Mahalanobis
- Z > 1.8
01
Ras Tanura
02
Strait of Hormuz
03
Malacca Strait
04
Singapore
Waypoints named in the Changepoints note. Not a map.
AIS positions are sampled every 30 minutes and aggregated by day. Prices are USO for Strategy 1 and BWET as benchmark for Strategy 2.
FOCuS implements Page-CUSUM over all window sizes with pruning for O(log n) average runtime. Strategy 1 enters USO on coordinated regional changepoints with direction from simple moving averages and exits on rule-based swing completion.
Strategy 2 uses Mahalanobis distance on stalled ships in a 500-hour window with trigger Z > 1.8. Waypoints: Ras Tanura, Strait of Hormuz, Malacca Strait, Singapore. Universe: FRO, STNG, TNK, DHT. Exits: 2.5× ATR trailing stop and 48-hour time stop, with a 94-hour lag for AIS reporting delay.
03 — Results
Documented strategy results
Strategy 1 (USO): 43 trades, 62.8% win rate, 65.23% cumulative return over two years. MCMC summary: 99.1% probability of profit, 1.69× expected multiplier, 5% floor 1.15×.
Strategy 2 (stalled tankers vs BWET): 26.30% net return, 61.54% win rate, 0.94% annual alpha versus BWET with the 94-hour AIS lag applied. No second MCMC block is reported for this leg.
Two-year simulation · Backtest
Return over two years, two separate tests
65.23%65.23%
Changepoints on USOil
Two-year simulation
26.30%26.30%
Ship Movement Tracking
Backtest
% · Two years of historical dataChangepoints research note. 5 October 2026. Strategy 1 is a simulation. Strategy 2 is labelled a backtest.
| Series | Kind | Value |
|---|---|---|
| Changepoints on USOil | Two-year simulation | 65.23% |
| Ship Movement Tracking | Backtest | 26.30% |
- Win rate
- 62.8%
- Strategy 1, 43 trades
- Backtest
- 61.54%
- Strategy 2 win rate
- Backtest
- 0.94%
- Strategy 2 annualized alpha
MCMC check · Strategy 1
Three printed figures
- Probability of profit
- 99.1%
- Expected multiplier
- 1.69×
- 5% risk floor
- 1.15×
Strategy 1, two-year simulation. The note gives no distribution, so none is drawn. Source: Changepoints, Strategy 1 results.
Scope and limits
Results are tied to the documented windows, universes, and lag assumptions above; they are not presented as out-of-sample guarantees.
Strategy 2 explicitly models AIS reporting delay with a 94-hour lag.
Strategy 1’s MCMC table reports probability, expected multiplier, and a 5% floor only; the full posterior is not reproduced on this page.