ETA accuracy benchmark · 🇯🇵 Yokohama / Tokyo Bay

How Port Flow's predicted ETA compares to each vessel's own broadcast ETA at Yokohama / Tokyo Bay, measured on closed voyages over the last 7 days. We publish the number — good or bad. Lower error is better.

Port Flow predicted ETA (RMSE)

9.22 h

MAE 4.79 h · 678 voyages

Broadcast ETA (RMSE)

53.79 h

crew-declared, 444 voyages (same set)

Our advantage

82.9 %

lower error than the broadcast ETA

Broadcast ETAs more than 7 days from the actual arrival are excluded as AIS sentinel / placeholder values (199 excluded). Both figures are computed on the same remaining voyage set — apples to apples.

Closed voyages

error = predicted/broadcast minus actual arrival (hours)


431011509	product	10/12, 05:25 AM UTC	1.02 h	6.58 h
431021282	container	10/12, 05:00 AM UTC	0.70 h	-29.01 h
431200703	container	10/12, 04:48 AM UTC	0.23 h	-142.80 h
431010674	product	10/12, 04:12 AM UTC	0.31 h	8.29 h
431000771	container	10/12, 04:00 AM UTC	0.49 h	9.17 h
431011759	container	10/12, 03:36 AM UTC	0.60 h	-41.43 h
431015069	container	10/12, 03:32 AM UTC	0.47 h	-55.55 h
431014012	product	10/12, 03:15 AM UTC	0.36 h	-128.27 h
431300239	container	10/12, 02:54 AM UTC	6.74 h	-90.90 h
431009717	product	10/12, 02:53 AM UTC	3.03 h	—
431019191	product	10/12, 01:57 AM UTC	6.51 h	9.55 h
431000734	container	10/12, 01:35 AM UTC	1.44 h	9.08 h
241969000	product	10/12, 01:17 AM UTC	2.28 h	-50.91 h
563280900	container	10/12, 01:13 AM UTC	-3.49 h	—
431024341	container	10/12, 01:07 AM UTC	1.66 h	-37.12 h
246122000	product	10/12, 12:56 AM UTC	2.00 h	-1865.94 h
431000639	product	10/12, 12:16 AM UTC	5.93 h	-98.45 h
440864000	container	10/12, 12:09 AM UTC	-0.74 h	-61.66 h
431101177	container	10/11, 11:58 PM UTC	-1.28 h	-1923.97 h
431800699	product	10/11, 11:45 PM UTC	1.48 h	-58.43 h
431013119	container	10/11, 11:45 PM UTC	-0.38 h	-397.76 h
431028356	container	10/11, 11:35 PM UTC	2.98 h	-0.09 h
563111100	general-cargo	10/11, 11:33 PM UTC	-0.61 h	-691.55 h
431402067	product	10/11, 11:32 PM UTC	1.91 h	-55.04 h
431020818	product	10/11, 11:29 PM UTC	-16.75 h	-23.65 h
431000972	product	10/11, 11:25 PM UTC	1.74 h	9.07 h
431501083	container	10/11, 11:13 PM UTC	1.12 h	9.28 h
431009717	product	10/11, 11:09 PM UTC	1.28 h	—
431003614	product	10/11, 11:04 PM UTC	1.27 h	-222.58 h
431000132	container	10/11, 10:58 PM UTC	0.29 h	-875.47 h
431024139	container	10/11, 10:57 PM UTC	0.25 h	—
431001258	container	10/11, 10:57 PM UTC	0.26 h	—
431012675	product	10/11, 10:56 PM UTC	4.30 h	9.06 h
431009162	product	10/11, 10:47 PM UTC	-0.19 h	0.20 h
352003726	container	10/11, 10:46 PM UTC	0.29 h	-1835.78 h
431012847	container	10/11, 10:42 PM UTC	0.11 h	3.54 h
431401857	container	10/11, 10:39 PM UTC	0.51 h	-63.16 h
431001258	container	10/11, 10:28 PM UTC	-13.73 h	-4.97 h
218818000	container	10/11, 10:23 PM UTC	1.20 h	-2.39 h
477302300	general-cargo	10/11, 10:21 PM UTC	4.96 h	-15.36 h
356352000	general-cargo	10/11, 10:20 PM UTC	1.54 h	-24.34 h
431018825	product	10/11, 10:20 PM UTC	2.42 h	-15.59 h
636022007	container	10/11, 10:19 PM UTC	0.69 h	-1.32 h
538010433	general-cargo	10/11, 10:13 PM UTC	0.57 h	-0.72 h
431932000	product	10/11, 10:09 PM UTC	-13.79 h	-182.17 h
431028356	container	10/11, 10:00 PM UTC	5.13 h	-13.00 h
477466300	general-cargo	10/11, 09:55 PM UTC	-0.74 h	-615.42 h
431020932	product	10/11, 09:45 PM UTC	2.87 h	9.25 h
228410700	container	10/11, 09:22 PM UTC	0.40 h	-2.46 h
431000771	container	10/11, 09:21 PM UTC	1.25 h	—

See it on a live desk

Predicted ETAs, congestion and sanctions screening across 51 ports — try it in one click, no signup.

Open dashboard See pricing →

How it's measured

Predicted ETA: distance-to-berth over speed-made-good, refreshed on every AIS position, with a seasonal correction from rolling medians.
Broadcast ETA: the crew-entered ETA in the vessel's AIS message — the de-facto industry default.
Both are compared to the actual arrival timestamp on the same closed voyages. We report RMSE (penalises large misses) and MAE (typical error).
Broadcast ETAs more than 7 days off are dropped as AIS sentinel / placeholder values, and our model is scored on that same set — so the comparison is honest, not cherry-picked.
Coverage is public AIS only; ports with weak coverage show fewer closed voyages. We'd rather show a small honest sample than a big misleading one.