1.0 INTRODUCTION & METHOD
This analysis is based on predictions for the knockout stages of the UEFA 2022 (Women’s) European Championship Finals, which took place between 6-31 July 2022 (the tournament was postponed from 2021). The objective of this report was to have a bit of fun and to see how accurately we could predict the tournament. Two academics (Dr Alex Gillett and Dr Kevin D Tennent, both editors of Soccermadboffins website) recorded their predictions at each stage (Group, Quarter-Final, Semi-Final and then Final) as just prior to each stage, once the competing teams were known. This was done using UEFAs own tournament prediction tool on its website.
A third set of predictions was undertaken at the same time by rolling 2x6 sided dice. A roll of 1-3 meant that was how many goals were scored, 4-5 = nil, whilst a roll of 6 meant roll again and whatever number 1-6 is the total number of goals (this was to allow for the possibility of particularly high scoring games). The idea of the third set of predictions by rolling two ‘random dice’ was for comparison, to see if we could perform better than random chance. We acknowledge that the six-sided dice somewhat limits the ‘random’ score but it gives an approximation, and it is the same method we have used for previous tournaments so it allows comparison.
We thus followed the ‘Analysis Z’ method of our previous tournament production reports, which predicted results and goals based only on actual fixtures. We did not simulate the entire tournament before it began with hypothetical knockout stage fixtures. This was due to the lack of time availability and the lack of a formulated spreadsheet that we could download ‘off the shelf’ as with previous tournaments. We hope that in future the Women’s Euros will be popular enough to attract the attention of spreadsheet prediction architects, or that we have more time with which to create our own.
Results show that predicting game outcomes (which team wins, or a draw) was for us a lot easier to do than to predict how many goals either side will score. Alex was the overall best predictor (outcomes and goals) and the only ‘player’ to get over 50%, all because of his 64% accuracy of predicting game outcomes. Kevin did quite well at predicting the Group stages but was less successful in later rounds and also found it hard to predict score-lines. The ‘random’ dice roles scored only around 19% at game outcomes and overall and was a noticeably less effective means to predict games than was applying common sense.
2.0 FINDINGS
Firstly, we predicted the group stage, the first phase of the tournament comprising four (4) groups, A-D.
The groups each contained four teams and the top two finishers in each group (50%) automatically progressed to the next stage, the ‘Quarter Final’, whilst the two lowest ranked from each group (also 50%) were eliminated from the tournament.
Thereafter, knockout Stages involved teams that qualified from the Group Stages:
- Quarter Finals (8 teams)
- Semi-Finals (4 teams)
- The Final (2 teams)
2.1 Group Stage
Here’s how we calculated the points, a maximum of 5 points per game were up for grabs (or 180 points overall in this stage)
· Correct outcome: 3 points (108 overall)
· Home goals: 1 pt (36 overall)
· Away goals; 1 pt (36 overall)
According to the analysis in the table on next page, Alex was most effective at predicting outcomes (which team won, or if they finished with a ‘draw’) with an impressive score of 51 (70.833% accuracy), compared to Kevin who scored 45 (=62.5%). The dice were not a very useful predictor managing only 31 (43%).
Regarding the goals scored per home team, Kevin had superior predictive power getting it correct in 8 instances (33.333%) and Alex scored 6 points (equating to 25%).
Regarding ‘away’ goals scored, Alex again scored 6 (25%) and Kevin again did slightly better with 7 (29.166%).
The dice’ random predictions did as well as Alex and not far behind Kevin for away goals (6, or 25%) but worse than either real player for home goals, predicting just 4 correctly (16.66%).
So, real brains did substantially better at predicting game outcomes and the exact number of goals scored by ‘away’ teams, but only Kevin was more accurate than random luck at guessing ‘home’ goals.
Both Alex and Kevin did quitte well with overall totals of 46% (Alex) and 43% (Kevin) meaning that they were about as good at predicting. However given the complexity of doing so in terms of outcome, and goals per team, these scores do not look low. As usual the random dice scored 22% suggesting that the Pareto 80:20 rule could be important here!
Fixture | Actual | Alex | Kevin | Dice |
England v Austria | 1-0 | 3v1 | 2v1 | 0v0 |
Norway v N ireland | 4-1 | 3v0 | 1v1 | 0v1 |
Austria v N. Ireland | 2-0 | 2v0 | 3v1 | 0v2 |
England v Norway | 8-0 | 2v1 | 2v1 | 3v2 |
Austria v Norway | 1-0 | 1v2 | 1v2 | 1v3 |
Northern Ireland v England | 0-5 | 0v3 | 1v5 | 3v2 |
Spain v Finland | 4-1 | 1v2 | 1v2 | 1v2 |
Germany v Denmark | 4-0 | 2v2 | 3v1 | 1v6 |
Denmark v Finland | 1-0 | 2v1 | 1v1 | 3v1 |
Germany v Spain | 2-0 | 3v0 | 2v3 | 4v4 |
Denmark v Spain | 0-1 | 2v1 | 0v2 | 0v0 |
Finland v Germany | 0-3 | 2v3 | 1v4 | 3v0 |
Portugal v Switzerland | 2v2 | 1v3 | 2v0 | 3v2 |
Netherlands v Sweden | 1v1 | 2v2 | 3v2 | 0v3 |
Sweden v Switzerland | 2v1 | 2v0 | 2v1 | 3v0 |
Netherlands v Portugal | 3v2 | 2v1 | 2v1 | 3v0 |
Sweden v Portugal | 5v0 | 2v1 | 2v0 | 1v0 |
Switzerland v Netherlands | 1v4 | 0v3 | 0v2 | 0v0 |
Belgium v iceland | 1v1 | 1v2 | 0v0 | 1v1 |
France v Italy | 5v1 | 2v1 | 1v1 | 1v0 |
Italy v Iceland | 1v1 | 2v2 | 0v1 | 2v1 |
France v Belgium | 2v1 | 3v3 | 4v1 | 0v1 |
Italy v Belgium | 0v1 | 1v2 | 0v1 | 3v2 |
Iceland v France | 1v3 | 1v3 | 1v5 | 5v1 |
Outcome | 72 | 51 (70.8%) | 45 (62.5%) | 21 (29.2%) |
Goals Home | 24 | 6 (25%) | 8 (33.3%) | 4 (16.7%) |
Goals Away | 24 | 6 (25%) | 7 (29.2%) | 6 (25%) |
TOTAL POINTS | 116 (100%) | 63 (54%) | 60 (52%) | 31 (27%) |
Table A showing group stage results and points according to AnalysIs Z
2.2 Knockout Stages
We then predicted scores and outcomes for the Quarter-Final, Semi-Final and the Final.
2.2.1 Quarter Final
| Actual | AG | KT | Dice |
England v Spain | 2v1 | 2v1 | 3v2 | 0v3 |
Germany v Austria | 3v0 | 3v0 | 4v0 | 3v2 |
Sweden v Belgium | 2v1 | 2v1 | 1v1 (0v0 AET, 4v5 pens) | 3v3 (3v1 AET) |
France v Netherlands | 2v1 | 2v1 | 3v2 | 1v0 |
Outcome | 12 (100%) | 12 (100%) | 12(100%) | 9 (75%) |
Goals Home | 4 (100%) | 4 (100%) | 0 (0%) | 0 (0%) |
Goals Away | 4 (100%) | 4 (100%) | 1 (25%) | 0 (0%) |
TOTAL POINTS | 20 (100%) | 20 (100%) | 13 (65%) | 9 (45%) |
Table B showing results and predictions for the Quarter Final
2.2.2 Semi-Final
At this point there were just two games featuring four teams, so showing % is a bit misleading if comparing accuracy to earlier rounds. However, we show them as part of our overall analysis of comparison when considering predictive power across the tournament. Kevin was the best performer in this round with 9 points (90% accuracy), Alex was just behind with 8 (80%). The dice random guessing yielded 0.
| Actual | AG | KT | Dice |
England v Sweden | 4-0 | 3-1 | 2-0 | 0-2 |
Germany v France | 2-1 | 2-1 | 2-1 | 0-2 |
Outcome | 6 (100%) | 6 (100%) | 6 (100%) | 0 (0%) |
Goals Home | 2 (100%) | 1 (50%) | 1 (50%) | 0 (0%) |
Goals Away | 2 (100%) | 1 (50%) | 2 (100%) | 0 (0%) |
TOTAL POINTS | 10 (100%) | 8 (80%) | 9 (90%) | 0 (0%) |
Table C showing results and predictions for the Semi Final
2.2.3 Final
The most famous fixture in Europe? Certainly one of them. Full 5 points (100%) to Alex, a tremendous 4 (80%) to Kevin and a hapless 0 to the dice.
| Actual Result | AG | KT | Dice |
England v Germany | 2v1 (100%) | 2-1 (100%) | 3-1 | 0-0, 0-3 AET |
Outcome | 3 (100%) | 3 (100%) | 3 (100%) | 0 (0%) |
Goals Home | 1 (100%) | 1 (100%) | 0 (0%) | 0 (0%) |
Goals Away | 1 (100%) | 1 (100%) | 1 (100%) | 0 (0%) |
TOTAL POINTS | 5 (100%) | 5 (100%) | 4 (80%) | 0 (0%) |
Table D showing results and predictions for the Final
3.0 DISCUSSION
We now summarise our predictions and provide concluding remarks.
3.1 Summary of Predictions
In the Group stages, Alex was the best predictor of results based on Analysis Z (three points for predicting the outcome, then a point each for correctly predicting goals scored for each team) with 64% accuracy compared with Kevin’s 44% and dice’ 20%. This follows a very close for with pattern and scores of the last tournament Kevin and the dice predicted, although Alex had a noticeably better accuracy.
The knockout stages were interesting. Alex again did well predicting 100% of the outcomes and the scores in the Quarter Final. Kevin did almost as well for outcomes but was less accurate about the scores. The dice did surprisingly well with the outcomes gaining 9 pts from a possible 12, but got no points for actual goals scored.
In the semi-finals Alex and Kevin both got 100% of the outcomes correct but the Kevin was slightly better at predicting actual score-lines. The dice though picked up zero points.
The Final was another successful round for Alex who predicted the correct outcome and goals for each team, Kevin was nearly as successful but was overoptimistic about the number fo goals England would score. Again, the dice were way off.
Viewing the tournament in its entirety it is possible to rate the overall predictive power of Alex, Kevin and the dice by compiling all of the games and predictions into a single set of calculations, by adding together scores at each round.
Total number of games = | Alex | Kevin | Dice |
Outcome (151 pts possible) 116 + 20 + 10 + 5 | 63 + 20 + 8 + 5 = 96 (63.6%) | 45 + 12 + 6 + 3 = 66 (43.7%) | 21 + 9 + 0 + 0 = 30 (19.9%) |
Goals Home (31 pts possible) 24 + 4 + 2 + 1 | 6 + 4 + 1 + 1 = 12 (38.7%) | 8 + 0 + 1 + 0 = 9 (29%) | 4 + 0 + 0 + 0 = 4 (12.9%) |
Goals Away (31 pts possible) 24 + 4 + 2 + 1 | 6 + 4 + 1 + 1 = 12 (38.7%) | 7 + 1 + 2 + 1 = 11 (35.5%) | 6 + 0 + 0 +0 = 6 (19.4%) |
151 + 31 + 31 = 213 TOTAL POINTS (213 pts possible) | 96 + 12 + 12 = 120 (56.338%) | 66 + 9 + 11 = 86 (40.376%) | 30 + 4 + 6 = 40 (18.779%) |
Table E showing overall tournament prediction accuracy
We have used our 'Analysis Z' method used in previous tournament prediction reports.
From the table we see that Alex was most accurate at predicting the outcomes and goals, but with Kevin just when it came to predicting goals. Both of us were wrong more than we were right about the exact scores, but in terms of outcomes Alex the evens by scoring approximately 64%, in other words he was ‘right’ about two thirds of the time. As with past tournaments there were clearly insufficient games for the laws of averages to help the random dice scores, with the dice only predicting outcomes around 20% of the time. So better to consult a soccer mad boffin about a game prediction than rolling a dice!
Overall, the dice scored just under 20% total predictability using the Analysis Z formula of outcome (3 pts) + goals (2 pts), whereas Alex and Kevin were noticeably better (approx. 56% and 40%, respectively).
3.2 Comparison with other predictors
The analyst.com had England and France as even chance winners (19%) and Germany was considered as having a 15% chance of winning the tournament (https://theanalyst.com/eu/2022/06/womens-euro-2022-prediction-england-france-germany-spain-sweden-netherlands/)
The Guardian newspaper’s analysts seemed keen on England buut some favoured Spain and the Netherlands ( https://www.theguardian.com/football/2022/jul/04/womens-euro-2022-predictions-winners)
We can also compare our performance with the UEFA Women’s Euros 2022 with our performance of the previous international women’s football tournament that we predicted, the 2019 FIFA Women’s World Cup. For reasons of brevity we have limited our comparison to just ‘analysis z’ and overall accuracy (outcomes and goals scored by each team in Group and knockout stages). Alex’s score improved by 3.1% whilst Kevin’s fell by the same amount. Most dramatically the dice were 20% less effective in 2022 than in 2019, explained by the random nature of this method and its lack of informed logic. However, perhaps next time this randomness could work in the other direction?
Overall % accuracy of predictions (outcomes and goals scored by each team in Group and knockout stages) | Alex | Kevin | Dice |
2019 FIFA Women’s World Cup | 53.5 % | 43.5 % | 33.8% |
2022 UEFA Women’s Euros | 56.4% | 40.4% | 18.8% |
Table F comparing overall performance between 2019 and 2022
4.0 Conclusions
Our judgement, although not too accurate a method for predicting the tournament, was better than random chance as measured by the ‘random dice’ method. Alex in particular was correct at predicting outcomes about 64% of the time but predicting exact score lines was much more difficult.
It seems that other predictions of ‘the experts’ were probably not much better than we were at correctly predicting how the tournament would unfold, as we both thought that England would at least reach the final. Once the teams in the Final were known we both chose England, with Alex predicting the exact score and Kevin only slightly too optimistic about England’s goal haul.
In conclusion this was a great tournament, not entirely predictable, and it was good to see England ‘bring it home’.
To cite this publication:
· Gillett, A.G., Tennent, K.D. (2022) ‘Predicting the 2022 Women’s UEFA European Championship: A report by the Soccer Mad Boffins’. 21st September 2022. Available online at: http://soccermadboffins.blogspot.com/2022/09/predicting-2022-uefa-womens-european.html
Also in this series about predicting major international football tournament scores:
· Gillett, A.G., Tennent, K.D. (2021) ‘Predicting the 2020(21) UEFA European Championship: A report by the Soccer Mad Boffins’. 21st July 2021. Available online at: http://soccermadboffins.blogspot.com/2021/07/euro-20202021-predictions-how-did-we-do.html
· Gillett, A.G., Tennent, K.D. and Fanning, J. (2019). ‘Predicting the 2019 FIFA Women’s World Cup Finals Part 2: Knockout Stages & Overall Analysis - A report by Soccer Mad Boffins’. 12th July 2019. Available online at: http://soccermadboffins.blogspot.com/2019/07/predicting-2019-fifa-womens-world-cup.html
· Gillett, A.G., Tennent, K.D. and Fanning, J. (2019). ‘Predicting the 2019 FIFA Women’s World Cup Finals Group Stages: A report by Soccer Mad Boffins’. 21st June 2019. Available online at: http://http://soccermadboffins.blogspot.com/2019/06/the-womens-world-cup-group-stage-predictions.html
· Gillett, A.G., and Tennent, K.D. (2018) ‘World cup predictions: human brain, advanced statistical modelling, or completely random?’. 2nd July 2018. Available online at: http://soccermadboffins.blogspot.com/2018/07/world-cup-predictions-human-brain-or.html
Blog:
· http://soccermadboffins.blogspot.com/
