Research
The scorelines football fans over- and under-predict
By Denver (@Denveloper), founder of CalledItMate · Published October 2026
We compared 22,221 score predictions made by 971 players on CalledItMate, a free football prediction game, with the final scores of the 5,617 completed matches they were made on, all kicking off between 22 March and 30 September 2026.
0-0 ended 6.1% of matches but got 1.2% of predictions, about a fifth of its share. 2-1 got 17.3% of predictions, more than any other score, and ended 8.5% of matches.

Aggregate data: download the CSV. The chart and data are free to use with attribution to CalledItMate. Cite the dataset as doi.org/10.5281/zenodo.23154456.
Fans back decisive scores and skip the tight ones
2-1, 1-2, 3-1 took 39.7% of all predictions between them, but ended only 19.3% of matches. The three tightest results went the other way: 1-0, 0-1, 0-0 ended 21.8% of matches between them and got 7.3% of predictions.
1-1 was the most common final score of all (542 matches, 9.6%), and got 5.8% of predictions. The pattern describes what fans predict, not why: the data cannot say whether it is optimism, habit or something else. For draws in detail, see Football fans predict draws half as often as they happen; for goalscorers, see The goalscorers football fans back.
Every common scoreline
Sorted from most over-predicted to most under-predicted. The gap is the share of predictions minus the share of matches, in percentage points. Scorelines are home-away, so 1-2 is an away win by one goal.
| Score | Predictions | Matches | Gap |
|---|
| 2-1 | 3,847 (17.3%) | 478 (8.5%) | +8.8 |
| 1-2 | 2,715 (12.2%) | 360 (6.4%) | +5.8 |
| 3-1 | 2,251 (10.1%) | 245 (4.4%) | +5.8 |
| 2-0 | 2,510 (11.3%) | 387 (6.9%) | +4.4 |
| 1-3 | 1,566 (7.0%) | 167 (3.0%) | +4.1 |
| 3-0 | 1,517 (6.8%) | 272 (4.8%) | +2.0 |
| 0-2 | 1,317 (5.9%) | 244 (4.3%) | +1.6 |
| 0-3 | 700 (3.2%) | 139 (2.5%) | +0.7 |
| 0-4 | 201 (0.9%) | 54 (1.0%) | -0.1 |
| 6-0 | 19 (0.1%) | 23 (0.4%) | -0.3 |
| 0-5 | 63 (0.3%) | 37 (0.7%) | -0.4 |
| 1-5 | 22 (0.1%) | 27 (0.5%) | -0.4 |
| 3-4 | 4 (0.0%) | 23 (0.4%) | -0.4 |
| 4-3 | 2 (0.0%) | 26 (0.5%) | -0.5 |
| 1-4 | 203 (0.9%) | 78 (1.4%) | -0.5 |
| 5-2 | 9 (0.0%) | 31 (0.6%) | -0.5 |
| 5-0 | 77 (0.3%) | 51 (0.9%) | -0.6 |
| 4-0 | 405 (1.8%) | 135 (2.4%) | -0.6 |
| 2-4 | 46 (0.2%) | 47 (0.8%) | -0.6 |
| 4-2 | 71 (0.3%) | 56 (1.0%) | -0.7 |
| 4-1 | 284 (1.3%) | 110 (2.0%) | -0.7 |
| 5-1 | 34 (0.2%) | 53 (0.9%) | -0.8 |
| 3-3 | 15 (0.1%) | 55 (1.0%) | -0.9 |
| 2-2 | 824 (3.7%) | 260 (4.6%) | -0.9 |
| 2-3 | 247 (1.1%) | 119 (2.1%) | -1.0 |
| 3-2 | 324 (1.5%) | 141 (2.5%) | -1.1 |
| 1-1 | 1,283 (5.8%) | 542 (9.6%) | -3.9 |
| 0-1 | 516 (2.3%) | 376 (6.7%) | -4.4 |
| 0-0 | 266 (1.2%) | 341 (6.1%) | -4.9 |
| 1-0 | 831 (3.7%) | 506 (9.0%) | -5.3 |
Scorelines with fewer than 20 matches and fewer than 100 predictions are left out of the table and the CSV.
Methodology
- Window: matches that kicked off from 22 March 2026 to 30 September 2026 inclusive (UTC), finished, and have a recorded final score. Data read on 1 October 2026 and frozen.
- Score prediction: a player's predicted final score for a match, at most one per player per match, as held at kick-off.
- Shares: a scoreline's share of predictions counts each prediction once; its share of matches counts each match once, among matches that received at least one prediction.
- Excluded: house accounts run by CalledItMate, internal test matches, and duplicate accounts merged into another account.
- Same window as the draws study, read later the same day: a few more results had settled and accounts had been merged, so totals here (22,221 predictions, 5,617 matches) are slightly higher than that study's. Quote each study with its own totals.
- Limitations: CalledItMate players are a self-selected group of football fans, not a random sample of the public, and the competition mix follows what they chose to predict.