"In seats contested" matters more than statewide share for small parties: a party polling low statewide but high where it actually stands is locally decisive even though it looks negligible overall.

In these seats a third party polled more votes than the winning margin, so it decided the result without taking the seat. This is arithmetic over the real vote counts, not a prediction about where those votes would otherwise have gone.

Select opposition parties to recount all 90 seats as though they had contested as a single bloc, using their real 2023 per-seat votes.

This is a ceiling, not a forecast. It assumes every alliance partner's 2023 voters transfer perfectly to a single candidate, which never fully happens in practice. Treat it as the outer bound of the risk, and read it alongside the vote-split analysis above.