How good are the chances?

Every chance on ActRanker is an estimate. This page shows how often those estimates came true in earlier years, per festival and per class.

How the test works

For each edition in the test, the model was rebuilt without that edition. A chance was then calculated for every name with what was known at that moment, including spring announcements by other festivals. Afterwards we checked who actually played.

Editions in the test: 2017, 2018, 2019, 2022, 2023, 2024, 2025, 2026. 2020 and 2021 are missing because the festivals did not take place. Last run: 04/11/2026.

What to keep in mind

The classes were tuned on the same test. That the share per class sits close to the chance of that class follows from this and proves little on its own. What does tell you something: how many names from the top class actually played, and how much of the real line-up came from the low classes.

The rules for tours and radio are not in the test yet, because the figures from earlier years are missing. Ticketmaster only shows upcoming concerts, and we have only been saving them since late October 2026. The radio figures do not go back far enough for every edition in the test. On the site both rules do count.

Most of a line-up consists of smaller names. Artists with a smaller digital footprint give fewer signals that point to a bigger or smaller chance.

Rock Werchter

Of the 101 names rated very likely beforehand, 70 actually played (69%). Of all 757 names that played, 21% had been rated very likely or likely. 49% had a class below possible: the model did not see those names coming.

Class	Names in the test	Actually played	Share
Very likely	101	70	69%
Likely	201	87	43%
Possible	906	226	25%
Unlikely	804	84	10%
Very unlikely	2.012	95	5%
Next to impossible	76.444	195	0%

Pukkelpop

Of the 101 names rated very likely beforehand, 59 actually played (58%). Of all 1.351 names that played, 16% had been rated very likely or likely. 66% had a class below possible: the model did not see those names coming.

Class	Names in the test	Actually played	Share
Very likely	101	59	58%
Likely	401	159	40%
Possible	1.104	243	22%
Unlikely	1.523	184	12%
Very unlikely	4.900	222	5%
Next to impossible	72.258	484	1%

Best Kept Secret

Of the 100 names rated likely beforehand, 36 actually played (36%). Of all 697 names that played, 5% had been rated very likely or likely. 82% had a class below possible: the model did not see those names coming.

Class	Names in the test	Actually played	Share
Likely	100	36	36%
Possible	400	89	22%
Unlikely	700	74	11%
Very unlikely	2.000	100	5%
Next to impossible	76.800	398	1%

Pinkpop

Of the 101 names rated likely beforehand, 41 actually played (41%). Of all 450 names that played, 9% had been rated very likely or likely. 84% had a class below possible: the model did not see those names coming.

Class	Names in the test	Actually played	Share
Likely	101	41	41%
Possible	100	31	31%
Unlikely	1.002	114	11%
Very unlikely	1.203	64	5%
Next to impossible	77.763	200	0%

More about the figures in the methodology.