Each Gameweek I will be previewing the forthcoming matches by using the ‘Wisdom of the Crowd’ – the aggregate predictions from tens of thousands of football fans playing the Premier League Predictor game on Superbru.
Please do join the FFS league here and add your own predictions to the data – the more there are, the more accurate the crowd should be (in theory).
Last season’s results
Following a comment from ViperStripes in my introductory article earlier this week, I have looked back at the data from last season, where there is around 11m predictions to analyse.
The data showed that the crowd correctly predicted the outcome of 221 of the 380 matches, for a success rate of just over 58% – and 11% of those were exact score predictions. Of the matches predicted incorrectly, over half were draws, suggesting that there were plenty of near misses in the community predictions. A 58% success rate doesn’t feel hugely impressive, but it’s probably not bad, and we’ve added ‘Crowd Wisdom’ as a user in the FFS pool, and we’ll enter the predictions each week to keep track of performance.
For the Expected Goals predictions last year, the average margin of error across the season was +/-0.46 goals.
So if the overall average expected goals for Liverpool is 2.15, the result is likely to be within 0.46 of that prediction – which means two goals are probable, and they are more likely to score three goals than just one.
I also noticed that overall the crowd tends to slightly under-predict goals, so it appears the margin of error is more likely to be upwards than downwards.
Across the season, there was an average of 1.6 goals scored by the home team, compared to 1.2 goals scored by the away team. The crowd predicted an average of 1.43 goals for the home team, and a fairly accurate 1.19 goals for the away team.
Now you can’t read too much into statistics, and obviously a team cannot score 0.1 of a goal, but I find this quite interesting.
Gameweek 1 Predictions
We now have over 25,000 predictions recorded for each match in Gameweek 1, and here’s the updated chart showing Expected Goals:
We can rank these in order of expected goals scored to see where we should focus our attacking resources. We can also look at the expected goals conceded to see where the clean sheets are most likely to come from, according to the crowd.
Those picking one or more of Manchester City trio Kevin de Bruyne, Gabriel Jesus or Sergio Aguero could be in line for a bumper set of attacking points on this basis. Arguably one of these players could even be handed the captaincy.
It is also still worth paying attention to Chelsea, Liverpool and Man United players. Romelu Lukaku is the runaway leader in the Captain Poll, and looks like a good option with Manchester United expected to score at least twice against West Ham.
In terms of clean sheets in Gameweek 1, Chelsea appears to be the standout option, which makes the attack minded Marcos Alonso and Gary Cahill strong candidates for delivering points at both ends of the pitch.
tatistics. The crowd also seems to think that Manchester City and Liverpool are likely to keep clean sheets – despite their away fixtures at Brighton and Watford respectively. This could be good news for those taking a chance on the Reds budget defender Trent Alexander-Arnold, who is expected to fill in at right back in the absence of the injured Nathaniel Clyne.
Among other cheap defensive options, Crystal Palace players, such as Patrick Van Aanholt are worth a look. According to the crowd they are more likely to keep a clean sheet than Arsenal, Everton and Spurs. West Brom v Bournemouth could well be a 0-0 – over 47% of people on Superbru are predicting a Draw in that match. This will be welcome news to those backing Baggies stopper Ben Foster as well as Cherries defenders, such as Charlie Daniels.
If you’re keen to play the Predictor game, join the FFS league here and add your own predictions to the data. Good luck!
