ETA accuracy benchmark · 🇵🇱 Gdansk

How Port Flow's predicted ETA compares to each vessel's own broadcast ETA at Gdansk, measured on closed voyages over the last 30 days. We publish the number — good or bad. Lower error is better.

Port Flow predicted ETA (RMSE)

18.30 h

MAE 8.11 h · 354 voyages

Broadcast ETA (RMSE)

55.43 h

crew-declared, 188 voyages (same set)

Our advantage

67.0 %

lower error than the broadcast ETA

Broadcast ETAs more than 7 days from the actual arrival are excluded as AIS sentinel / placeholder values (158 excluded). Both figures are computed on the same remaining voyage set — apples to apples.

Closed voyages

error = predicted/broadcast minus actual arrival (hours)


255806300	container	10/12, 05:08 AM UTC	0.68 h	-295.14 h
255915602	container	10/12, 04:41 AM UTC	4.64 h	-62.69 h
636023839	container	10/11, 10:31 PM UTC	0.69 h	5.48 h
255806452	container	10/11, 10:26 PM UTC	-16.07 h	-749.44 h
257673000	container	10/11, 10:24 PM UTC	0.06 h	-2008.41 h
245246000	container	10/11, 09:32 PM UTC	-0.58 h	—
230688000	product	10/11, 02:40 PM UTC	3.19 h	-34.67 h
261037120	product	10/11, 01:24 PM UTC	-0.48 h	-243.41 h
563087500	product	10/11, 11:41 AM UTC	-56.21 h	-1149.70 h
636023839	container	10/11, 11:33 AM UTC	-31.11 h	-34.56 h
565502000	product	10/11, 09:32 AM UTC	4.95 h	2.46 h
255916581	product	10/11, 09:10 AM UTC	-4.90 h	-284.18 h
211845000	container	10/11, 08:37 AM UTC	2.23 h	-104.63 h
248369000	chemical	10/11, 08:10 AM UTC	2.61 h	-1.18 h
250002133	container	10/11, 07:17 AM UTC	2.87 h	-1.28 h
311001591	chemical	10/11, 04:58 AM UTC	-0.55 h	-10.98 h
245246000	container	10/10, 10:31 PM UTC	-0.20 h	-3138.53 h
230719000	container	10/10, 09:42 PM UTC	-16.21 h	-97.70 h
255705000	general-cargo	10/10, 08:07 PM UTC	0.38 h	-592.13 h
230714000	container	10/10, 06:56 PM UTC	-5.22 h	-1127.95 h
563330900	product	10/10, 05:31 PM UTC	0.55 h	-3.52 h
311001591	chemical	10/10, 04:27 PM UTC	3.36 h	1.53 h
636026344	container	10/10, 11:07 AM UTC	2.87 h	-0.13 h
255806436	general-cargo	10/10, 08:09 AM UTC	1.43 h	-59.66 h
230631000	container	10/10, 07:34 AM UTC	1.24 h	-28.58 h
230688000	product	10/10, 04:42 AM UTC	2.51 h	-0.72 h
305313000	container	10/10, 02:41 AM UTC	4.12 h	7.30 h
304047000	container	10/10, 02:29 AM UTC	-52.22 h	-3004.48 h
245190000	container	10/09, 07:48 PM UTC	2.49 h	-1.76 h
563326600	product	10/09, 07:15 PM UTC	1.95 h	-474.25 h
230700000	container	10/09, 05:09 PM UTC	0.09 h	-437.17 h
563173900	product	10/09, 04:41 PM UTC	2.28 h	-4.68 h
636022719	container	10/09, 03:25 PM UTC	-100.49 h	-168.42 h
311000820	product	10/09, 03:10 PM UTC	1.10 h	-324.18 h
564306000	container	10/09, 02:54 PM UTC	-31.58 h	-876.90 h
250005692	container	10/09, 12:31 PM UTC	1.71 h	-2871.52 h
305313000	general-cargo	10/09, 11:22 AM UTC	0.08 h	-154.37 h
255806519	product	10/09, 11:09 AM UTC	1.45 h	-1073.16 h
255805594	container	10/09, 10:46 AM UTC	1.67 h	-703.77 h
255915667	container	10/09, 09:28 AM UTC	4.83 h	—
219699000	container	10/09, 07:35 AM UTC	-1.35 h	-813.58 h
246063000	container	10/09, 05:54 AM UTC	0.31 h	-1679.91 h
636025365	container	10/09, 03:37 AM UTC	2.32 h	-295.63 h
209276000	general-cargo	10/08, 11:41 PM UTC	-0.93 h	-6.69 h
209276000	general-cargo	10/08, 08:14 PM UTC	0.12 h	-3.25 h
210350000	container	10/08, 07:46 PM UTC	-10.07 h	-48.77 h
261208000	container	10/08, 03:18 PM UTC	4.48 h	21.69 h
305597000	container	10/08, 02:49 PM UTC	0.47 h	-1975.82 h
255916226	container	10/08, 02:26 PM UTC	1.17 h	-104.45 h
255915667	container	10/08, 12:00 PM UTC	0.95 h	-825.01 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.