This is now tangental to the point, but O’Quigley (2024) journals.sagepub.com/doi/10.1177/... provides a good summary of how such arguments often go wrong.

This may sound like abstract mathematical pedantry, but a similar argument was used in a recent high-profile criminal case in the UK. I don’t understand the medical evidence, so I have no opinion whether she was guilty or not, but I can tell you that the statistical evidence didn't prove her guilt.

Unfortunately, that’s not helpful in this case, because it only tells you the probability is less than 99.9999%. No new information.

However, here a miscalculation has led to a claim that the probability must be less than 1.84%. I’d also bet a beer that it is, but this argument doesn’t prove it.

Yes. If you replaced the words “these two cables” with “some two cables”, and “Yi Peng 3” with “some ship”, then I'd agree.

It’s difficult to estimate that probability, but it’s perfectly legitimate to say, “It must be less than the probability of any two cables breaking within 18 hours."

4/ The probability of at least one such occurrence isn’t quite 1 - (1 - p)^99, where p = 18.6%, although that’s a rough-and-ready approximation. The analytical expression for this probability is surprisingly complicated and unpleasant, but numerical simulation gave my >99.9999% result.

3/ However, that’s just the probability of one particular consecutive pair of breaks being close together. Over the year, there will be an average of 99 consecutive pairs, each having a probability of 18.6% of occurring within 18 hours.

2/ If events occur as a Poisson process, the waiting time from any event to the next follows an exponential distribution. The probability of having to wait no more than 18 hours from one cable break to the next is 1 - exp(-18 / (24 * 365) * 100), approximately 18.6%.

1/ Under those assumptions, the probability of two cable breaks occurring within 18 hours of each other in any calendar year is greater than 99.9999%.

There are certainly plenty of valid arguments that Yi Peng 3 cut the cables, but unfortunately, this can't be one of them.

By estimating the angular acceleration from the AIS heading data, the maximum torque acting on the ship at this time corresponds to a maximum lateral force on the bow of about 2 tonnes.

That would also be consistent with a chain pulling in a mostly fore-and-aft direction rather than transversely.

If so, you'd expect the windlass forces to turn the ship to face towards the anchor's position, so the arrows should (approximately) converge to a single focal point on my plot.

Maybe about (55.69, 15.798) towards the end of her maneuver?

The dataset is indeed at 10 second resolution although there are randomly some records missing, presumably due to reception.

And there's nothing at all on the 17th November, and nothing before 06:01:41Z on the 18th.

5658 records (excluding duplicates) are present for Yi Peng 3 on the 18th.

Does anyone with more maritime knowledge know if the ship has a conventional propulsion system with the thrust alway pointing stern to bow? Or would they be using bow thrusters, or anything which would result in lateral thrust in order to position themselves to an accuracy of 120m in the open sea?

I'd be curious to know if they were also making speed corrections to aim for that suspiciously round longitude coordinate.

AIS doesn't report the engine speed, but given the heading data and GPS track it might be possible to infer the propulsion thrust during this period.

It becomes more interesting with the heading vectors overlayed.

I think the ship is being very carefully steered towards (55.6900, 15.800).

It's making appropriate heading corrections to the north and south, and slowing down when near the target.

Yes, it does.

The discontinuity in the "Course over Ground" is just a quirk of wrapping around 360º rather than evidence of rapid acceleration.

The first record in that dataset for MMSI 414270000 (Yi Peng 3) on Nov 18th was at 06:01:41 (most likely UTC, but not specified).

The only pings within 50km of your estimate of the cut site (55.8695, 17.0048) on the 18th are from MMSIs 210388000, 219031605, 414270000, and 423520100.