Conference League qualifying gave us a useful lesson: the model found a few clean reads, then spent the rest of the round being mugged by late goals and awkward scorelines

The round mostly went against the grain, which is usually how these things choose to be remembered.

Across 60 fixtures, Fipster had only two STAR calls and spent most of the round in TICK-or-MISS territory. That sounds blunt because it is. The model found enough correct outcomes to keep itself respectable, but it repeatedly leaned on tidy home wins or narrow away wins that never quite survived contact with the actual scorers. In this sort of qualifying round, one early goal can make a prediction look clever or foolish very quickly. This time it was mostly the latter, with a few neat exceptions to stop it becoming a complete farce.

The cleanest call was FC Copenhagen at Debreceni VSC.

This was one of the few places where the pre-match logic held up properly. Fipster had flagged Robert as a scorer, and he did score. More importantly, the broader read was that Copenhagen’s attacking edge would travel better than Debreceni’s home comfort would protect them. The post-match data backed that up: Copenhagen scored three without reply from Debreceni after taking control early enough to make the rest feel like administration. The model did not need to be perfect here; it needed to identify which side had more reliable finishing. It did that, even if it also invited a couple of extra names to the party who never arrived.

The miss that stung most was Austria Vienna against Beitar Jerusalem.

This was not just a wrong scoreline. It was a wrong story. Fipster expected Austria Vienna to control it at home and named several possible scorers around that idea. Instead, Beitar scored twice before Austria could really settle, and then the tie went all the way to penalties after full time. That matters because it shows what got over-weighted: confidence in Austria’s attacking pattern, when the decisive detail was Beitar’s ability to land first and keep forcing the issue. Kelvin Boateng did at least repay one call for Austria, but by then the model’s tidy home-win script had already been folded up and put away.

The round’s key names were not always the ones Fipster had circled first.

Kevin Behrens was still one of the better reads on paper and he delivered again for FC Lugano, scoring in both legs across this section of the round’s narrative and finishing with five season goals after this round. That is exactly why he kept showing up in Fipster’s thinking: he remained a live threat even when Lugano’s results were less obedient than expected. Nicklas Røjkjær also justified his pre-match billing for FC Nordsjaelland by scoring in their win over Valur Reykjavik. Kelvin Boateng did likewise for Austria Vienna, even if his side then wandered into penalties like they had left their map in another competition.

Then there were the players who decided matches without being called at all. Andreas Cornelius turned up for Copenhagen as an unforeseen scorer, while Gabriel Pereira Magalhães dos Santos also punished Debreceni from nowhere in model terms. Petur Knudsen and Marius Kryger Lindh dragged NSI Runavik back into their tie against Lugano; Omer Atzili and Johnbosco Kalu did similar work for Beitar; Bendegúz Bolla shaped Rapid Vienna’s win over Paide; and Peter Berthelsen opened things for Nordsjaelland against Valur. That is where this round really bit: not just in missed outcomes, but in how many decisive goals came from names Fipster had not prioritised before kick-off.

So the lesson is fairly plain. The model can still find strong individual scorer angles and occasionally nail a result cleanly, but this round punished overconfidence in structure more than in personnel. The next step is obvious enough: keep trusting proven attacking profiles when they are genuinely supported by form, but be less sentimental about neat scorelines when knockout ties start behaving like knockout ties do — awkwardly, late, and with no regard for anyone’s spreadsheet.