Ranko's Champions League Test: A Cross-League Model vs the Market
Ranko was created to answer a question that ordinary domestic league tables cannot: how strong is a club when compared with teams from another league?
The Champions League is the natural test. Its league phase puts clubs from England, Germany, Spain, Italy and France directly against one another after domestic results have begun to reveal current-season strength.
Our two-season recorded sample produced an encouraging result:
Across 64 Champions League league-phase matches in 2024/25 and 2025/26, Ranko selected 40 correct 90-minute results (62.5%). Day-before BWIN selected 36 (56.3%), four fewer.
The same pattern appeared in both seasons separately.
Why this is the model's intended problem
Ranko is a proprietary results-only cross-league rating and forecast system. Its declared inputs are completed matches from Europe's top five domestic leagues and UEFA competitions; it does not consume bookmaker prices or team-news data when generating its forecasts.
The system was created specifically for inter-league comparison:
- It places clubs from different domestic competitions on a common scale.
- It uses match results rather than bookmaker prices, lineups, injuries or transfers.
- Its intended test is a fixture between clubs from different leagues.
- It produces home/draw/away probabilities that can be frozen and evaluated before kickoff.
Ranko values are converted into home/draw/away probabilities through a proprietary calibrated mapping. Its functional form, fitted parameters, weighting rules and source code are not disclosed. The public study supplies only the fixed forecast outputs, benchmark probabilities, outcomes and evaluation calculations needed to audit the reported results.
This makes the Champions League result interesting: a results-only cross-league system remained competitive on top-pick accuracy in its intended domain.
The two-season result
All forecasts refer to the 90-minute home/draw/away result. Bookmaker odds were normalized to remove their margin.
| Forecast | Matches | Correct | Accuracy |
|---|---|---|---|
| Ranko day-before | 64 | 40 | 62.5% |
| BWIN day-before | 64 | 36 | 56.3% |
The Ranko-versus-BWIN point estimate favored Ranko in each season:
| Season | Matches | Ranko | BWIN |
|---|---|---|---|
| 2024/25 | 30 | 21 (70.0%) | 19 (63.3%) |
| 2025/26 | 34 | 19 (55.9%) | 17 (50.0%) |
In both seasons, Ranko made two more correct picks than BWIN. In the separate 60-match subset with closing odds, Ranko and the BetExplorer closing bookmaker average each selected 38 results correctly (63.3%). This later-timed benchmark is a sensitivity analysis, not the inclusion rule for the primary cohort.
The result remains positive without matchday 8
Qualified teams may rotate on the final league-phase matchday. We therefore repeated the paired Champions League comparison in three nested windows ending at matchday 7.
| Matchday window | Matches | Ranko | BWIN | Difference |
|---|---|---|---|---|
| MD2-7 | 59 | 36 (61.0%) | 33 (55.9%) | +5.08 pp |
| MD3-7 | 56 | 35 (62.5%) | 33 (58.9%) | +3.57 pp |
| MD4-7 | 50 | 31 (62.0%) | 28 (56.0%) | +6.00 pp |
Ranko's observed accuracy remained 3.57 to 6.00 percentage points higher in all three windows. These overlapping cohorts are sensitivity summaries, not independent tests, and their paired uncertainty intervals all include zero.
What we can and cannot conclude
The observed result is positive, but 64 matches are not enough to prove a forecasting edge. Against BWIN, Ranko was uniquely correct 12 times and BWIN eight times. The estimated difference was +6.25 percentage points, with a paired 95% bootstrap interval from -7.81 to +20.31 points and exact McNemar p=0.503.
The correct statement is therefore:
Ranko's observed Champions League top-pick accuracy was competitive with serious bookmaker references in this recorded two-season sample. Superiority and statistical equivalence have not been established.
The scientific report also includes Brier score, log loss, all secondary analyses, the cohort definition, all limitations, a worked scoring example and the next prospective protocol.
Read the paper and inspect every row
- Scientific report (PDF)
- Public Google Sheet: match-level validation data and scoring formulas
- Versioned dataset snapshot (XLSX)
- Continuously updated model-performance audit
- Try the inter-league match predictor
The next step is a complete prospective Champions League archive: every eligible forecast timestamped and frozen before kickoff, with eligibility, reference odds and statistical tests specified before the results are known.
That is the standard required to turn an encouraging two-season result into stronger evidence.