1) Does the large gap in the implied probabilities of a Labor win in the upcoming Australian Federal Election between overall-result betting (62% at the latest update) versus seat-by-seat betting (23% at the latest update) imply some sort of arbitrage opportunity? Is the overround really that large?

2) Why the gap in the first place? I assume it’s got something to do with the particulars of just how marginal each seat is (see below), but why do the two markets disagree so much? Is one of them massively incorrect?

For those that don’t know, I’d recommend looking at the tabular and graphical representations of the marginal seats here. Here are the guts of it (there are 150 seats in the House of Representatives):

Uniform swing |
Labor gains |
Labor total |
Coalition total |
Result |

2.92% | (+13) | 73 | 75 | Clear Coalition win |

3.27% | (+14) | 74 | 74 | Sort-of-hung parliment (depending on the two independents) |

3.27% | (+15) | 75 | 73 | Sort-of-hung parliment (depending on the two independents) |

4.85% | (+16) | 76 | 72 | Clear Labor win |

The first 13 seats for Labor only need a 2.92% swing, or 0.22% per seat. The 14th seat will take an extra 0.35%, the 15th a further 0.90% again and the 16th another 0.71% on top. Remember that each extra percent of swing to Labor takes an increasing amount of goodwill from the electorate (i.e. increasing your swing from 1% to 2% is doable, but increasing it from 25% to 26% is effectively impossible).