We published six predictions for the midweek round of 24 to 26 August, covering the Premier League, Serie A, La Liga and the Champions League play-off second legs. All six won. Three of them were sub-50% calls, including a 35% pick on LASK Linz that knocked Celtic out of Europe. The season record now stands at 25 wins and 14 losses from 39 graded picks, 64% overall. A perfect round is a small sample, not a signal, and the rest of this page explains why.
Every graded pick, 24 to 26 August 2026
Each row shows the pick exactly as it was published on the predictions board before kick-off, the confidence score attached to it at that moment, and the official result.
| Competition | Match and result | Our pick | Conf. | Graded |
|---|---|---|---|---|
| Serie A | AS Roma 4–0 Fiorentina | Roma win | 61% | WIN |
| Champions League | Bodo/Glimt 3–0 NEC Nijmegen | Bodo/Glimt win | 59% | WIN |
| Champions League | Sabah 5–2 Hapoel Beer Sheva | Sabah win | 57% | WIN |
| Premier League | Fulham 2–3 Chelsea | Chelsea win | 49% | WIN |
| La Liga | Valencia 0–1 Real Betis | Real Betis win | 39% | WIN |
| Champions League | LASK Linz 5–1 Celtic | LASK win | 35% | WIN |
| Midweek total | 6 picks | 6 won, 0 lost | 100% |
The 35% pick that came in: LASK Linz 5–1 Celtic
This was the lowest confidence score we have ever attached to a published pick, and it is the one that landed. LASK were behind in the tie, at home, against the Scottish champions. Our model gave them a 35% chance of winning the leg, which is another way of saying we expected them to lose it roughly two times in three.
They won it 5–1 and put Celtic out. That is the correct way to read a 35% number: it is not a prediction that something will not happen, it is a statement that it happens about a third of the time. Sites that only publish their winners would frame this as a masterstroke. It was a long shot that came in, and we are logging it as exactly that.
Six from six does not mean the model got better
The honest framing matters more than the headline. Six picks is a tiny sample. A round like this moves the season number from 58% to 64% on volume alone, and a bad weekend will move it back down just as fast. Three of the six were rated under 50%, which means that on our own numbers we expected roughly two or three of the six to lose.
What this round genuinely tells us is narrower and more useful: our low-confidence band is not junk. Before this week the close-call band was returning 38%, which is close to what you would expect from picks we ourselves labelled as coin flips. Three wins from three has pulled that to 50%. Neither figure is the truth. The truth is somewhere in the middle and we will only know it after a few hundred picks, not 39.
Season record after 39 graded picks
| Competition | Won | Lost | Graded | Strike rate |
|---|---|---|---|---|
| Serie A | 5 | 1 | 6 | 83% |
| Champions League | 5 | 2 | 7 | 71% |
| La Liga | 6 | 3 | 9 | 67% |
| Premier League | 4 | 2 | 6 | 67% |
| MLS | 5 | 5 | 10 | 50% |
| UFC | 0 | 1 | 1 | 0% |
| All competitions | 25 | 14 | 39 | 64% |
The Champions League jumped from 50% to 71% in one night because all three play-off second legs we called came in. That is the same small-sample effect described above, running in our favour this time.
What the confidence bands look like now
This is the check that matters. If the confidence score is worth anything, higher-rated picks should win more often than lower-rated ones. Across 39 graded picks they still do, but the gap has narrowed sharply this week.
| Confidence band | Picks | Won | Lost | Strike rate | Change |
|---|---|---|---|---|---|
| Strong, 70% and above | 7 | 6 | 1 | 86% | No change |
| Good, 55% to 69% | 16 | 11 | 5 | 69% | Up from 62% |
| Close call, under 55% | 16 | 8 | 8 | 50% | Up from 38% |
The bands are still in the right order, which is the minimum test a confidence score has to pass. The spread between top and bottom has fallen from 48 points to 36. If the close-call band keeps climbing towards the good band, that is not good news, it means our scores are not separating anything and the labels need recalibrating rather than celebrating.
The draw problem did not go away
None of the six midweek matches finished level, which is why the round graded clean. Across the whole season, nine of our 14 losses have been draws rather than picks that named the wrong winner. That ratio is unchanged. One clean round does not fix a structural pattern, and it is still the single biggest thing standing between this record and a better one. We took that apart in detail here: Draw No Bet Meaning: What 33 Graded Predictions Show About the Draw Trap.
What is live now
- Wednesday 26 August: a full Betway Premiership midweek round, with Richards Bay v Kaizer Chiefs and Mamelodi Sundowns v AmaZulu both at 19:30 SAST, plus the last Champions League play-off second legs.
- Thursday 27 August: the Champions League league phase draw in Monaco at 16:00 GMT. No picks, the fixture list follows on 29 August.
- Friday 28 August: Crystal Palace v Manchester City opens Premier League Matchweek 3, and Bayern Munich v VfB Stuttgart opens the Bundesliga at 75%, our highest-confidence pick of the week.
- Saturday 29 to Monday 31 August: 31 fixtures called across the Premier League, La Liga, Serie A, the Bundesliga and MLS.
Follow the picks this week
- Today’s football betting tips, refreshed each morning
- Bet of the day, one pick only, and no-bet days are declared
- Tomorrow’s soccer predictions
- Correct score predictions, with the honest hit rate attached
- How accurate are football predictions? Our real, graded numbers
How we grade, in full
- Every pick is graded against the official full-time result, wins and losses both.
- A pick wins only if the named side wins the match. A draw counts as a loss. We do not reclassify draws as pushes or half-wins.
- Extra time and penalties are ignored. Two-legged ties are graded on the individual leg we picked, not on the aggregate.
- A fixture that cannot be verified as played is voided and excluded from the record. Eleven picks have been voided this season.
- Picks are never edited or removed after kick-off. The board you see before the match is the board that gets graded.