Skip to content
Methodology

How often is the baseline right?

Open procedures on AmendEU carry a statistical baseline: how each political group has historically voted on similar files, projected onto today’s seats. That is a description of the past, not a forecast — and it should be checked against reality. This page does exactly that.

The test

For every recorded main roll-call votewith per-group tallies, the method is applied as if the vote had not happened: each group’s FOR-rate is computed from all the otherrecorded main votes (leave-one-out), applied to the group’s actual number of votes cast, and the projected call (pass / fail on expressed votes) is compared with the real outcome. The tested vote never contributes to the rates used to predict it.

92%
Correct calls
175 of 190 recorded main votes
190
Votes tested
main roll-calls with group tallies

Where the method is solid — and where it is a coin flip

Accuracy by projected margin: the further the projection sits from 50/50, the more it can be trusted. Close calls are close calls — no baseline decides them.

0–2.5 pts from 50%n = 0
2.5–5 ptsn = 0
5–10 ptsn = 0
10–20 pts92%n = 190
20+ ptsn = 0

By committee

Committees with at least 8 tested votes; smaller samples are withheld rather than shown as noise.

AFET100%n = 16
ENVI / ITRE / TRAN80%n = 10

What this is not

Not a forecast of any individual vote, not a rating of Members or groups, and not adjusted for the politics of the day — amendments negotiated after the baseline window, group discipline on a specific file, or absences. The population is the recorded main votes this site serves (190); the baselines shown on dossier pages draw on a larger vote history and may differ. Read the methodology for sources and counting rules.