Following on from last year, we have again collected Eredivisie predictions (you can download them all here) but instead of 13 to follow through the season, we now have a total of 48 from online media, fans, statistical models and prominent media figures. Of the 13 who took part last season, nine are back again which means that their consistency of success (or otherwise) can begin to be looked at. The nine returnees include the best objective predictor from 2013/2014 (Soccermetrics) and the best subjective predictor (Nikos Overheul). Can they repeat their success?
The original motivation for collecting predictions a year ago was that so many are made before the season but they are rarely collected or assessed. Any assessment that does take place tends to be based on a measure of absolute deviation from what actually takes place but as one of my favourite analytical bloggers James Grayson explains below, there is a better way:
“Standard deviation was used to evaluate the predictions as it gives an excellent indication as to the uncertainty and spread in a given set of predictions. It can also be applied well to a 20 team league. For a quick worked example, a set of predictions with a standard deviation of 10 points would have predicted the points totals of roughly 13 teams correctly to within 10 points, and the points totals of roughly 19 teams correctly to within 20 points. The smaller the standard deviation, the better the set of predictions was. As such, assessing the results by standard deviation provides a layer of detail that cannot be drawn from just using the mean absolute error, as the spread of the uncertainty in a set of predictions cannot be drawn from the mean absolute error.”
This year’s 48 forecasts have been divided into four groups; online media, fans, statistical models and prominent media. They will be compared to one another and also to a number of benchmarks such as the teams getting the same points as last season, all teams recording 47 points, a bookmaker average, a median of all forecasts which could be termed “wisdom of the crowd” and medians of each subgroup of predictors.
Two plots will be produced with one being based on points, which have been submitted by most predictors, and the other based on the rank order of the 18 clubs in each of the forecasts. Where points were not submitted, the average number of points for the last nine seasons for the position predicted has been used. It is important to use both assessments as the aim of many predictors, particularly in the media, will be to produce an accurate rank order of teams. What follows is a look at how each of the 18 Eredivisie clubs has been predicted to perform this season.
The fans predict the most points for Ajax (median of 77) and the prominent media forecasters the fewest (72). Nine of the 10 statistical models have Ajax winning the league with only simple TSR going against this. Ajax are forecast to finish in the top two in all 48 models.
The prominent media forecasters are the most positive on PSV with a median of 77 points. Seven of the 12 in this group predict PSV to win the title. Only two of the other 36 forecasters predict PSV to win this season’s Eredivisie. Statistical models are the most negative on PSV with a median forecast of 65.5 points.
The fan group is the most positive about FC Twente with a median prediction of 69.5 points. The prominent media forecasters put Twente outside the top three on median points as eight of the dozen predictors place them fourth or lower. The statistical model based on last season’s TSR is the only one of the 48 forecasters to make FC Twente this season’s champions.
The prominent media group forecast Feyenoord to finish third with a median of 69 points. The fan group is the least positive with a median of 62 points. Three statistical models put Feyenoord second in the Eredivisie and Hans Kraay junior of the prominent media group predicts them to win the league.
Three of the groups place Vitesse in fifth place but the prominent media group puts them sixth behind AZ. None of the four groups has a median points total reaching 60 or more. Three statistical models and six of the prominent media group have Vitesse finishing outside the top six but no one from the online media or fan groups does.
The statistical models are the least positive about AZ with a median of 53 points. No one has AZ finishing higher than fourth and only the statistical model using last season’s TSR has AZ outside the top eight, in ninth place. Three of the prominent media sample have AZ in fourth position compared to just two of the other 36 predictions.
sc Heerenveen and FC Groningen
The online media group is the only one to have FC Groningen finishing higher than their northern neighbours sc Heerenveen and all groups classify these two clubs in seventh and eighth overall. One prediction in each group has sc Heerenveen finishing in the top five with Hugo Borst the most bullish as he has placed them in fourth position. One statistical model (Steve Lawrence) and one online media entry (Michiel Jongsma) have Heerenveen in the “rechterrijtje” (bottom half).
Three predictors; Michiel Jongsma, Michel Abbink (both online media) and Terry Tipster (statistical models) have FC Groningen in the top five at the end of the season. Three stat models, two members of the online media group and one fan put FC Groningen in the “rechterrijtje” next May.
FC Utrecht, PEC Zwolle and ADO Den Haag
The online media group has ADO Den Haag as the best of this trio of mid table teams, fans make PEC Zwolle the best and the other two groups have FC Utrecht as the final club in the “linkerrijtje” (top half). Over all 48 forecasters, PEC Zwolle have the largest standard deviation of points at 5.88 with their season’s points total predicted to be between 31 and 54.
Three statistical models, two fans and Kees Jansma of the prominent media group have FC Utrecht finishing 12th or lower this season with one of the statistical models suggesting that they will finish as low as 17th. At the more optimistic end of the scale, three prominent media, two statistical models, two fans and one online media entry have FC Utrecht in the top seven. Wilfried Genee has the club finishing as high as sixth.
All but two of the predictions have PEC Zwolle finishing outside the mid-table zone of seventh to 13th. Valentijn Driessen puts them sixth and Johan Derksen 15th. Three fans and two of the online media group have forecast Zwolle to finish in seventh position this time around.
Forty five of the forecasts place ADO Den Haag between eighth and 13th place this season with the only exceptions being fan Sander Koreman (16th), modeller Niels van de Ven (15th) and prominent media representative Emile Schelvis (14th).
Heracles Almelo, SC Cambuur, Go Ahead Eagles, NAC Breda and Willem II
Cambuur are regarded as the best of this group of five by both the statistical models group and the prominent media group. The online media group have Heracles level with Cambuur as the best and the fans favour Go Ahead Eagles. The online media group is also the only subgroup with Willem II outside of the bottom three.
Heracles Almelo are placed between 11th and 15th position in 41 of the 48 models. Two statistical models (TSR and Niels van de Ven’s goal difference and transfer model) forecast a top half position with finishes of 7th and 9th respectively predicted. Three fans and two statistical models forecast Heracles to finish 16th and thus in the promotion/relegation play-offs at the end of the season.
Go Ahead Eagles are predicted to finish between 11th and 16th position by 39 of the 48 models with one member of each of the online media, fan and prominent media groups placing the club 17th. Three statistical models forecast the Deventer club to finish in the bottom two with Terry Tipster putting them bottom.
With a standard deviation of 2.2 places, NAC Breda are the team with the most variable forecasts for finishing position. However, none of the 48 predictions have placed NAC in the top half of the table with forecasts ranging from 10th to 18th. A quarter of the prominent media group though have them 10th as does one of the online media group. Another online media group member is the only entrant to place NAC last in the table.
Four statistical models, two fans and two of the prominent media group have predicted SC Cambuur to finish in a position outside 12th to 16th. Jan Mullenberg’s model, a Bayesian Network based on the last three seasons, is the only one to place Cambuur in the top half of the table (9th). Steve Lawrence’s squad value model joins Hugo Borst and Emile Schelvis from the prominent media group in placing Cambuur 17th at the end of the season.
Willem II are forecast to finish between 12th and 18th by all predictors. Two fans, Lionel Stute and Wouter Pennings who both blog at SportPreview.NL, have the Tilburg club finishing bottom and returning to the Jupiler League. The only other forecaster to place them last is De Telegraaf’s Valentijn Driessen.
FC Dordrecht and Excelsior
Statistical models rate these two equally, perhaps because of lack of information on these two promoted clubs. The other three groups all have FC Dordrecht as a stronger team than Excelsior. FC Dordrecht are not predicted to fnish higher than 14th by anyone with only Jesse Wieten and the statistical model using last season’s TSR putting them this high. The TSR model is actually based on the previous performance of teams promoted via the play-offs in this case as Dordrecht have not been in the Eredivisie for nearly 20 years.
Excelsior are forecast to finish bottom of this season’s Eredivisie by 27 predictors but two statistical models, from Terry Tipster and Steve Lawrence, have them as high as 13th.
If you’ve got this far, enjoy this ridiculous pre World Cup prediction which removes eventual winners Germany at the first hurdle and ignores all common sense:
Let the games begin……