Champions League Opening Round: Our Model vs the Market
We ran our model against bookmaker prices on five Champions League league-phase openers. Fifteen selections, one rated better than 'no edge' — and the draw was mispriced in every single game.
The 2026-27 league phase is under way, and we did the boring thing: took our model's probabilities for five of the opening fixtures, put them next to the bookmaker's implied probabilities, and looked at the gaps.
The headline is not the one betting sites usually publish. Across 15 selections, exactly one earned a rating above "no edge". That's it. One.
The gaps, in full
| Fixture | Selection | Odds | Model | Market | Gap |
|---|---|---|---|---|---|
| Bayern – Bodø/Glimt | Bayern | 1.07 | 87.7% | 88.7% | −1.0 |
| Bodø/Glimt | 24.0 | 5.3% | 4.0% | +1.3 | |
| Draw | 13.0 | 7.0% | 7.3% | −0.3 | |
| Slavia Prague – Lens | Slavia | 2.69 | 38.2% | 35.3% | +2.9 |
| Lens | 2.55 | 36.2% | 37.2% | −1.1 | |
| Draw | 3.46 | 25.6% | 27.4% | −1.8 | |
| PSV – Shakhtar | PSV | 1.49 | 64.7% | 63.6% | +1.0 |
| Shakhtar | 5.80 | 16.7% | 16.3% | +0.3 | |
| Draw | 4.74 | 18.7% | 20.0% | −1.3 | |
| Como – RB Leipzig | Como | 1.77 | 54.1% | 53.8% | +0.3 |
| Leipzig | 4.16 | 23.4% | 22.8% | +0.6 | |
| Draw | 4.06 | 22.5% | 23.4% | −0.9 | |
| Fenerbahçe – Roma | Roma | 2.02 | 48.8% | 47.0% | +1.9 |
| Fenerbahçe | 3.54 | 27.0% | 26.8% | +0.2 | |
| Draw | 3.62 | 24.2% | 26.2% | −2.0 |
Gaps are in percentage points. Market figures are de-vigged, so they sum to 100 and are comparable to ours.
Every draw is priced above what we think it's worth
Look down the draw rows. Five fixtures, five negative gaps, from −0.3 on Bayern to −2.0 on Fenerbahçe–Roma. Not one exception.
That is a directional disagreement, not noise in one game. Our model reads these matchups as more decisive than the market does. Whether the market is wrong or our model underrates stalemates in a new competition format is genuinely open, and we'll know more in a couple of months once results start settling. For now it's the clearest signal in the dataset, and it points at avoiding the draw rather than backing it.
The one gap worth naming is also our least confident call
Slavia Prague at 2.69 is the only selection rated as carrying any edge: 38.2% from us against 35.3% implied, a 2.9-point gap.
Now the awkward part. Our confidence score on that fixture is 0.30, labelled Low, and it's the lowest of the five. The reason is on the team sheets: Slavia are missing nine named players, Lens seven. When a squad list is that shredded, form data from previous weeks stops describing the team that will actually walk out.
So the single place our model argues with the market is the single place it knows least. That's not a coincidence worth ignoring. Thin information produces wide estimates, and wide estimates drift from a price the market has set with better team news than we have.
Compare PSV–Shakhtar: confidence 0.97, our highest, and the gap on PSV is +1.0. Where we're sure, we agree.
Favourite shading is visible on Bayern
Bayern are 1.07, which de-vigs to 88.7%. We have them at 87.7%. Bodø/Glimt at 24.0 implies 4.0%; we say 5.3%.
Both gaps run the same way: the market pushes the favourite up and the outsider down. That's the textbook favourite–longshot pattern, and at 1.0 and 1.3 points it's small, but it showed up exactly where you'd expect it, in the most lopsided game on the slate.
Goals: nothing unusual
Over 2.5 goals sits between 47.4% and 50.8% across all five. The biggest mismatch on the slate, Bayern–Bodø/Glimt, tops it at 50.8%; Fenerbahçe–Roma comes in lowest at 47.4%. A four-point spread across five games is nothing. No goal-fest, no lockdown. Both teams to score ranges from 17.9% in Munich to 52.7% in Prague, which tracks the mismatch gradient and tells you little you couldn't guess.
What we'd actually do
Nothing, mostly. A 2.9-point gap attached to a Low confidence rating is not a bet, it's a note to check the Slavia team news an hour before kick-off and see whether the price has moved toward us or away.
The useful takeaway is the draw pattern. Five out of five is worth watching across the next rounds, because if it holds it's a structural read on how this format is being priced rather than a single-week quirk. We'll revisit it once the ledger has enough settled results to say something with a confidence interval attached.
Our model's calls are recorded before kick-off and graded automatically as matches finish, including the ones that miss. You can see the running record on the track record page, and the method behind these numbers is written up in the methodology.
18+. Odds quoted were captured pre-match and move; a gap this small disappears with one price change.