An 11-season Footiqo study for Fantasy Premier League managers
Six-match form gave a clearer signal than three-match form, but the season-to-date record proved more useful still. That is the main finding of an analysis of 3,517 Premier League fixtures across 11 completed seasons, using only information available before each match.
The study was conducted using historical Premier League match data from Footiqo. It compares rolling team form with actual results, and looks at two team-level indicators that matter to fantasy managers: scoring and clean sheets.
Which measure of form predicted the next result best?
For an out-of-sample test, we fitted simple match-result models using 2,558 fixtures from 2015/16 to 2022/23. We then evaluated the models on 959 fixtures from 2023/24 to 2025/26, which were not used for training. Each form-based model used the home and away clubs’ points per match over the stated period.
| Pre-match information | Correct picks | 1X2 accuracy | Log loss* |
| Historical result frequencies | 419 / 959 | 43.7% | 1.071 |
| Previous three matches | 451 / 959 | 47.0% | 1.045 |
| Previous six matches | 469 / 959 | 48.9% | 1.035 |
| Season-to-date record | 503 / 959 | 52.5% | 1.002 |
*Log loss measures the quality of predicted 1X2 probabilities, not just the most likely result. Lower is better. The baseline uses historical home/draw/away frequencies with no team-form inputs.
Key findings at a glance
- Use six-match trends as a more stable starting point than a three-game snapshot, but do not treat a short streak as a forecast on its own.
- Check the season-to-date record alongside current form. In this test, it predicted the next 1X2 outcome better than either recent-form window.
- For attacking and defensive selections, consider how often the team has recently scored or kept clean sheets, then adjust for the next opponent and player-specific factors.
Figure 1. Three-match, six-match and season-to-date models were evaluated on exactly the same 959 unseen fixtures.
Six-match form correctly identified the most likely outcome in 48.9% of the test fixtures, against 47.0% for the three-match version. However, the model using each club’s season-to-date points per game reached 52.5%. It also had the lowest log loss, indicating better probability estimates, not merely more correct picks.
The six-match improvement over three matches was modest, and the test does not establish that the difference is statistically conclusive. By contrast, the longer season-to-date record performed better than either short-window version in this particular comparison.
What happens when one side has much better six-match form?
The full 3,517-match sample shows a clear association between the gap in the clubs’ recent points per game and the eventual result. When the home side led by at least one point per game over the previous six matches, it won about two-thirds of its next fixtures. When the away side held a similar advantage, away wins were more common than home wins.
Figure 2. Outcome shares by the home team’s points-per-game advantage over the away side across the preceding six league matches. Sample sizes appear next to each band.
| Previous six-match form gap | Matches | Home win | Draw | Away win |
| Away 1.0+ ahead | 584 | 23.3% | 22.8% | 53.9% |
| Away 0.34-0.99 ahead | 720 | 38.8% | 26.0% | 35.3% |
| Similar form | 1,035 | 46.5% | 23.0% | 30.5% |
| Home 0.34-0.99 ahead | 661 | 53.9% | 26.0% | 20.1% |
| Home 1.0+ ahead | 517 | 64.6% | 18.2% | 17.2% |
These are descriptive associations, not evidence that recent form causes the result. A strong run may simply reflect greater underlying team quality, which helps explain why the season-long record remained the strongest of the simple 1X2 models.
A practical signal for fantasy attacking picks: repeat scoring
Across 7,034 team-match observations, sides that had scored in all six of their previous league fixtures scored again in 80.8% of their next games. Grouping clubs by how often they found the net in the preceding six matches gives a less noisy picture and avoids conclusions from tiny categories.
Figure 3. Share of teams scoring at least once in their next match, grouped by how many of the preceding six league games they scored in.
| Scored in previous six | Team matches | Scored in next match |
| 0–2 of 6 | 459 | 65.1% |
| 3–4 of 6 | 3,053 | 70.0% |
| 5–6 of 6 | 3,522 | 77.0% |
For FPL managers, recurring team scoring can help identify a useful attacking environment. It does not tell us which player will score, assist or earn bonus points. Opponent strength, expected minutes, injuries and individual involvement still matter.
The defensive view: how reliable are clean-sheet streaks?
Clean-sheet form was informative, but the signal was less decisive than some simple streak-based advice suggests. Teams without a clean sheet in their previous six matches kept one in 22.5% of their next fixtures. Those with three or more clean sheets across the same period did so more often.
Figure 4. Next-match clean-sheet frequency by the number of clean sheets over the previous six league games.
| Clean sheets in previous six | Team matches | Next-match clean sheet |
| 0 of 6 | 1,275 | 22.5% |
| 1–2 of 6 | 4,219 | 26.3% |
| 3–6 of 6 | 1,540 | 31.8% |
A defence with several recent clean sheets may still face a difficult next opponent. The data measure team clean sheets, not a goalkeeper’s saves or a defender’s full FPL return, so they are best used as one part of fixture assessment.
Methodology and data source
Source: Footiqo historical Premier League results, 2015/16 to 2025/26. The sample contains 3,517 completed, non-neutral fixtures with valid 90-minute scores and at least six earlier league matches for both teams in the same season. Only prior fixtures were used when calculating each side’s form. Rolling form is across all league venues, not solely home or away.
For the prediction comparison, a multinomial logistic-regression model used the home and away teams’ points per game over the previous three matches, six matches, or entire season to date. The same modelling procedure and 959 held-out fixtures were used in each comparison. The no-form benchmark used historical result frequencies. Accuracy means the highest-probability 1X2 outcome was correct; log loss also assesses the confidence of the predicted probabilities. These are match-result forecasts, not FPL points forecasts or betting returns.
For the descriptive charts, team scoring and clean-sheet outcomes were counted separately for the home and away participants in each match, producing 7,034 team-match observations. The two observations within one fixture are not statistically independent. Small categories were pooled, and no causal claim is made. Changes in clubs and league conditions across seasons may influence the pooled averages.
Footiqo provides free team and league football statistics covering results, goals, corners, cards, home and away performance, recent form and other match indicators.
