Article

Feature:SIMprofiler-ing who will win the league

Sat 29 Sep 2007, 12:03

Running the final SIMprofilerTM projections each year is one of our favourite moments at The Football Review. Not only has everything been updated and uploaded, but all new arrivals to the Premier League have been added and assessed. It also means that we can have some fun with the final output and look at how the model thinks each and every team will fare in the new season ahead, based on what it thinks is in the near future for each player in the league.

Many media outlets like to print their league table predictions before the season starts which is understandable, but very few, if any, have any form of science behind them and they often closely resemble the table from the previous season with only a few minor changes. Adding an element of science behind the figures produces some interesting results, that are often far different from the conventional predictions and give us a great look at the ‘real’ direction in which the different teams might be headed.

We too run a pre-season cut of the numbers, and those of you that have already bought this year’s book will be familiar with the process and the first projections made before the season started. The reason we don’t run with those numbers as the final league projections is because when we go to print there is still a good two months left of player transfers which, of course, have a big impact final team positions. Even though the season has already started, we don’t use any information that has currently been collected - except playing time to get a better indication of where new player additions sit in their squad’s hierarchy.

The player projection process might be complex, but the system of creating the final league table is much simpler. These predicted standings are based on compiling the SIMprofilerTM   projections for each team's squad as at 1st September 2007. This ensures that we are including all new recruits who joined any team before the summer transfer window closed. All players past performance data then has to be added before the forecasting process can begin. Having generated each and every player’s playing time and attacking statistics for the season ahead we also generate each and every player’s defensive figures. We did not include these figures in the book - citing sample size issues for some individual players - but for whole teams this issue is less of a problem. However, it can also be concluded that due to the attacking statistics being based on a greater sample of data, the Goals For data will be more accurate than the Goals Against figures. 

The key statistics in building up each team’s predicted record are each player’s attacking goal value above-average (aGVA) and defensive goal value above-average (dGVA). These two statistics are basically a measure of how many goals each player adds to his team’s attack and how many goals each player prevents his team conceding, above that of an average performer at his playing position given the same playing time. The statistics are setup in such a way that the sum of every player’s aGVA and dGVA in a team estimates how many goals a team will score above the league average total and how many goals each team will concede below the league average total. The value we have used to represent the league average total of goals is the league average total from last season – 46.55 goals.   For example, the sum of aGVA and dGVA values from SIMprofilerTM for Blackburn players this season is 7.17 and 1.42 goals above average, respectively.  Using 46.55 goals as a baseline tells us that the model predicts Blackburn to score 53.72 goals and concede 45.131 goals.      

The next step of the process is to convert these goal figures into points and to do this we use a regression model that estimates the relationship between Goals For, Goals Against and Points based on the last five years worth of data. The very last step in the process is to round the figures to whole numbers for ease of reading and because even Didier Drogba would find it hard to score 0.35 of a goal. 

Table 1.1 The Final SIMprofilerTM 2007/08Premier League Standings

Rank Team Goals For Goal Against Goal Difference Points
1 Man United 85 29 56 89
2 Liverpool 70 22 48 83
3 Arsenal 74 36 38 77
4 Chelsea 59 27 32 73
5 Everton 57 36 21 66
6 Aston Villa 53 38 15 62
7 Tottenham 63 54 9 58
8 Blackburn 54 45 9 58
9 Newcastle 48 41 7 57
10 Man City 40 34 6 56
11 Reading 54 50 4 55
12 Portsmouth 49 49 0 53
13 Middlesbrough 39 49 -10 46
14 Fulham 41 53 -12 44
15 Bolton 46 60 -14 43
16 West Ham 44 59 -15 42
17 Wigan 33 56 -23 37
18 Sunderland 36 60 -24 37
19 Birmingham 31 62 -31 32
20 Derby 23 68 -45 22

(Note: Given above is each team's mean projection in Goals-for and Goals-against in isolation of other team's mean projections in the same statistics.  Therefore, the sum of goals-for and goals-against for the entire league will not necessarily match)   

Manchester United …1st place, 89 points

Defending as current Champions is never easy, but SIMprofilerTM seemed to have very little reason to think that Manchester United couldn’t match their fine 2006/07 season in 2007/08. The model predicted an extra two goals for, a further two goals conceded for the exact same goal difference and points total as last season. When you look at what United added to the squad over the summer (Nani, Anderson, Hargreaves and Tevez) it would be hard to argue with the prediction. What has so far happen this season doesn’t fit with the high octane attack predicted, but has seen United’s defence at a championship level. There is no doubt that it has been a tough start to the season for United’s attack, only registering six goals in seven games, but four straight wins, all without conceding a goal, has Sir Alex’s men in second place even without the attack having moved into second gear. The rest of the title challengers should be very worried about that fact.      

Liverpool … 2nd place, 83 points

Almost every league prediction we have seen before have had Manchester United and Chelsea in the top two spots. Perhaps rightfully so, given the last few seasons, but when you take into account summer transfers and age related improvements or declines Liverpool have improved a lot this season. Even if this prediction becomes hogwash, on paper at least Liverpool have the second best team this year and only major injuries or unforeseen events will see them finishing below Chelsea. The additions of Torres, Voronin and Babel and the further development of a number of returning players, greatly improve the Liverpool attack that has been wasteful in front of goal the last two seasons (15th in SOG-per-goal efficiency in 2005/06 and 13th in 2006/07). SIMprofilerTM thinks that Liverpool’s ‘goals for’ tally will increase by 13 goals from last season – the biggest increase predicted from last season – which isn’t unthinkable, seeing that Liverpool generated the second most goal scoring chances last season and the most in 2005/06. The defence is also predicted to improve by a few goals, with the full time utilisation of Daniel Agger and better depth at full-back with Alvaro Arbeloa being present from the beginning of the season. However, the quality and depth of United’s attack might just be too much to compete with.

Arsenal … 3rd place, 77 points

The first run of the SIMprofilerTM predictions in pre-season had Arsenal finishing just below Chelsea. However, as you’ll read below, the swapping of places has more to do with Chelsea than it does with Arsenal changing a great deal. One thing is for sure we never for a moment had Arsenal finishing below Tottenham and said as much in this year’s book, having seen no real change in the quality of the Spurs defence, which would needed to move into the top-4. We also saw nothing that said Arsenal’s attack would be any worse. We highlighted Cesc Fabregas as the young player of the year award winner and one of the top three candidates to bring home the senior award too (http://www.thefootballreview.co.uk/articles/article.aspx?id=18) and so far he has not made us look foolish. Emmanual Adebayor was also on that list (5th place) and he too has now got going having scored a hat-trick last week against Derby. The only thing left to look like coming true of our major predictions about Arsenal players is that Robin Van Persie will be one of the league leaders in goals scored. He has made a slow start, but we are only seven games into the season and he will no doubt score loads before now and the end of the season.               
   
The Arsenal team has got off to a flyer, but the one problem that still remains to be solved is the defence. Is it championship calibre? Last year the Gunners conceded 35 goals and SIMprofilerTM thinks that it’ll be close to that figure again this season. The major problem remains quality indepth at the back, especially at centre-back, thus making the decision to loan out Johan Djourou all the more confusing. Does the centre-back rotation of Toure-Gallas- Senderos-Gilberto match favourably to United’s (Ferdinand-Vidic-Brown-O’Shea), Liverpool’s (Carragher-Agger-Hyypia-Arbeloa) or Chelsea’s (Terry-Carvalho-Ben Heim-Alex)? Only time will tell.

Chelsea …4th place, 73 points

As we mentioned above, even though Chelsea were predicted to decline from last season they were still three points ahead of Arsenal and very close to Liverpool’s predicted points total before the season began. Arsenal and Liverpool didn’t do too much to change from that point forth, but Chelsea changed greatly, bringing in a below-average defender in right-back Juliano Belletti, and shipping out Arjen Robben. Love-him or hate-him Robben was effective when playing last season and generated many goals for team. SIMprofilerTM predicted that Robben was about to improve again this season and had him generating an aGVA of 6.75 goals above an average winger given the same playing time. Losing him has had a major effect on the Chelsea prediction. After making all the adjustments, SIMprofilerTM has Chelsea scoring five fewer goals, and conceding two more than last season. Although not massive differences they are enough to push Chelsea down the table, given Liverpool’s and Arsenal’s predicted improvements.                                
   
The fact that Chelsea management has replaced Jose Mourinho only further enforces the team’s decline and the prediction has an even greater chance of being a reality. Had Mourinho still been in charge, his motivational and coaching quality could have ensured that the loss of Robben and the natural age-related decline of the squad was lessened. This would have probably ensured that Chelsea would have competed with Liverpool for second place (the title certainly looked out of reach even with Mourinho there), but now little known Avram Grant must win over the changing room, and quickly, if Chelsea are to be a real force this season.               

Everton …5th place, 66 points 

Everton were our pre-season pick for 5th and they remain our transfer deadline pick for 5th. This is a good team that only got better in the closed season. A 52 goal for-36 goal against team that won 58 points last season added Middlesbrough’s best player and consistent double-figure goalscorer Aiyegbeni Yakubu, Sheffield United’s most versatile and best player Phil Jagielka and Wigan’s best defender Leighton Baines.. The only major loss was not getting Manuel Fernandez to stay. But cheap loan deals for Thomas Gravesen and Steven Pienaar help to cover for not being able to finalise the purchase. This all adds up to a similar, but far more potent attacking team than last year’s squad and a nice five goal-eight point increase predicted by SIMprofilerTM.

Aston Villa …6th place, 62 points

People always seem to forget that only Chelsea, Manchester United and Arsenal lost fewer games than Villa did last season. Martin O’Neill did an absolutely amazing job in year one. However, it is hard to see Aston Villa finishing this high now that John Carew will be missing for four to six weeks with knee ligament damage. The predicted goals for column was always going to be hard to meet, but without their target men for a major part of the season the team’s options are shortened upfront. There is still enough attacking quality (Agbonlahor, Moore, Maloney, Young) to trouble opponents, but it is hard to see the predicted ten goal improvement from last season becoming a reality. A slight improvement is more likely, which pulls them to up to around a league-average total of 46-47 goals. In addition to that slight improvement, the defence is likely to progress again this season after a 14 goal improvement in Martin O’Neill’s first season in charge. The acquisition of Zat Knight and more minutes from Martin Laursen are the major reasons why SIMprofilerTM sees a further three goal improvement this season.                                                     

There is of course a chance that the defence improves more than predicted, but a 46 goals for-38 goals against season giving the team around 56-57 points looks quite likely. Unfortunately for Villa, a drop in goals for will bring them down the table in the tightly contested upper middle section to a level more likely below Tottenham, Blackburn and Newcastle.   

Tottenham…7th place, 58 points                                                                                        

Wow ! All those people predicting Tottenham would finish in 4th couldn’t have been more wrong. Firstly, they were grossly underestimating Arsenal and the fact that most of the team were about to get exponentially better due to last season experiences and secondly, to move into the top-4 you definitely have to have a league-average quality defence at the very least – only Everton have conceded more than 37 goals and finished inside the top-4 in the last five years. For Tottenham just to improve to league-average would require an eight-goal improvement. Such a large change is usually only capable via a major signing or a change in team tactics. But with the acquisition of Darren Bent to add to the three above-average forwards the team already had, the latter didn’t look like it was about to happen and with only Kaboul and Bale added to the defence, the former wasn’t exactly matched either. The summer arrivals will improve the attack, but having scored 11 goals above-average last season the money was not spent in the obvious area of need.

Blackburn …8th place, 58 points   

Blackburn made big improvements to their squad last season and this season will see the benefits. The main improvements are predicted in the team’s defence where SIMprofilerTM sees that as much as a nine goal improvement could be possible. Last January additions Stephen Warnock and Christopher Samba are two of the major changes from last year as is the better health of Ryan Nelson. As for the attack, Benedict McCarthy is highly unlikely to score as many goals as he did last season, but the continued development of Matt Derbyshire and the acquisition of Roque Santa Cruz will be enough to cover the slack to enable Rovers to fight for a European place.

Newcastle …9th place, 57 points

Without question Newcastle will be one of the teams that improve the most this season and remember we are purely looking at what the players within the squad are going to contribute and not what the new manager’s training regime and tactics will bring. One thing our previous research has shown us is that a change of playing style from a new manager can have a hugely dramatic effect on a player’s production and performance. One only has to look at the massive improvements made by the vast majority of Aston Villa’s defenders last season, where many of them moved from being well below average defenders under David O’Leary, to average and above-average quality last season under Martin O’Neill. Look for the same to occur at Newcastle where a number of players who under performed last season (Nicky Butt, Charles N’Zogbia) have already greatly improved. Sam Allardyce brought in some nice professionals (Smith, Geremi) and some impact performers (Viduka, Rozenhal), who instantly improve this Newcsatle side.
 
Man City …10th place, 56 points 

They have already made a fabulous start and like Newcastle, above, the new manager is having a big impact on returning players. Both the goals for and goals against counts will greatly improve this season. But to see City having an above-average attack would be amazing, seeing how they are coming from relegation quality, just a year ago. In our opinion the team still lacks an out-and-out goal scorer but they will generate far more scoring chances than they did a year ago. Overall, SIMprofilerTM thinks that the team will improve by 11 goals in attack and ten in defence. With new additions Vedran Corluka, Javier Garrido, Martin Petrov and Elano already proving their worth we think the model has this team pretty accurately pegged.

Reading …11th place, 55 points

On recent evidence, perhaps SIMprofilerTM is too optimistic for Reading, but we still think they have enough quality to stay up. Our major pre-season concern for Reading, as we stated in this year’s book, was that they needed to upgrade at centre-back, next to Ivar Ingimarrson. Unfortunately, management weren’t able to make a move and that problem has come back to haunt the team in the early going this season. To make matters worse, Ingimarrson has struggled as well. The other area of concern is in central midfield where Steve Sidwell’s loss is being felt harder in defence than attack. Summer acquisitions Emerse Fae and Kalifa Cisse have not started well and Brynjar Gunnarsson, who is a good occasional starter and substitute for the position, has started all but one game this season for the Royals.                                                                                                   
 
With the model’s goals against prediction coming out below-average at 50 goals conceded the pressure falls on the team’s attack. So far this season they don’t look like getting anywhere near the 54 goals predicted, meaning that a negative goal difference is highly likely. It is almost a given that Reading will decline this season and definitely finish in the lower half of the table. But they should generate more goals in the months to come to be safe from the drop.

Portsmouth …12th place, 53 points

It was interesting to see Portsmouth come out this low especially with a decline predicted in their defensive record. Adding Distin was a masterstroke, but the Hreidarsson/Traore mix at left-back scared SIMprofilerTM rigid, as did the Papa Bouba Diop signing at the end of the transfer window. In attack the team did enough to improve with the addition of Utaka, who looks even better than predicted (seven goals, 14 assists) so far this season, but can Kanu continue in his current place? After scoring eight times before the end of November last season, he only scored twice the rest of the way. If that happens again the Portsmouth attack could struggle.  

Middlesbrough …13th place, 46 points
 
With Middlesbrough this closed season it was all about how they were going to score goals. Firstly, Mark Viduka needed to be replaced and then Aiyegbeni Yakubu moved to Everton before the summer window closed. It is always difficult for a team to replace its best striker, but its two best strikers is another task entirely and especially for a team so heavily dependent on their production. Viduka and Yakubu represented 39% of the team’s total attacking production last season and 59% of the goals the team scored. In came Jeremie Aliadiere, Tuncay, Mido and Gary O’Neil to attempt to cover the loss, but it is unlikely that the production of all four players this season will match that of the two departed frontmen from a year ago.                                    
  
In defence the team was set to improve in pre-season with Woodgate and Pogatetz providing a solid backbone and Taylor, Davies, and Young providing a good balance of youth, quality and experience at fullback. Unfortunatley, Pogatetz’s injury adjustment and the increased presence of David Wheater sees the defence being comparable to last season, with very little improvement likely over the full season - thus putting the pressure back on the attack. If neither Mido nor Aliadiere catch fire and Tuncay continues to struggle as he has in the early going, Middlesbrough could easily fall further down this list.

Fulham …14th place, 44 points 

Lets make this clear from the off, this is a much improved side over last season’s version, but they’ll still be a below-average team this season. Conceding 60 goals as Fulham did last season is relegation quality and only scoring 38 goals is not much better either. This team needed a major overhaul and a major overhaul they got. Some of the summer changes will help and some will hurt, but the overall outcome should be a team that will score a few more and conceded a few less goals than they did last season. This team certainly has enough depth in midfield and attack to score more than they did a year ago and approach a league-average total. We especially liked the additions of Hameur Bouazza (who was above-average in an attacking sense with well below-average Watford last season) Steven Davis (a young player with some room for growth) and Diomansy Kamara (a Premier League quality player playing in the Championship). The defence remains a problem and could potentially pull this team into a relegation dogfight.

Bolton …15th place, 43 points

This is one where a lot of people might claim hindsight or blame Sammy Lee for being a worse manager than the departed Big Sam. The fact of the matter is neither is included in this prediction (except for the latter’s skills in player transfers of course). Quite simply, based on age related decline and the balance of player transfers over the summer, Bolton are set to be a lot worse this season. The main area predicted to decline is the defence, where the team lost the well above-average and first choice left-back Tal Ben Heim and one of their staring centre-backs from a year ago when Abdoulaye Faye departed for Newcastle. Jlloyd Samuel was a nice bargain pick-up and should provide steady play at left-back but Gerald Cid and Andy O’Brien are not quality replacements to play alongside Abdoulaye Meite.                                                                                               
 
While the defence has been weakened the midfield and attack were strengthened in depth as opposed to quality. Heidar Helguson should add something as a back-up for Davies, Anelka and Diouf and Alonso and McCann add to the central midfield depth that was already strong, but aging. It all adds up to a team that should score a similar amount, but will concede a fair few more goals this season. Exactly how many is something that Sammy Lee can still affect.

West Ham …16th place, 42 points

Curbishley had an effect last year and will need to again this season. Some of the better players on this team from a year ago have moved on (Carlos Tevez, Nigel Reo-Coker, Yossi Benayoun) and to a man have been replaced with lesser talent (Craig Bellamy, Scott Parker, Freddie Ljungberg). The attack will be better, but the key will be the defence that SIMprofilerTM doesn’t think will improve much this season from last. If the injury bug can be fought off and management can get a long run of continuity out of a Neil-Upson-Ferdinand-McCartney back-four the goals against figure will be improved upon, but SIMprofilerTM just couldn’t forecast that with any level of confidence.

Wigan …17th place, 37 points
 
After surviving the drop on the last day of the season Wigan began the summer well by trying to get younger players and releasing a number of aging players. Unfortunately, they continued the summer by not only picking up below-average players, but some of the worst players available. The Michael Brown, Antoine Sibierski, Marcus Bent and Titus Bramble fearsome foursome combined for a horrible -7.36 tGVA last season. Adding that sort of mediocrity is never good. However, the acquisitions of Jason Koumas, Mario Melchiot and Andreas Granqvist help to absorb some of that below-average fodder, but this team is still a favourite for the drop.

Sunderland…18th place, 37 points

Spending a lot of money in the off season does not necessarily result in success (just look at Chelsea’s £72 million outlay for a worse team last year) but Sunderland are in the best position to stay up, of the three newly promoted teams. Just looking at their squad and their start, that 60 goals conceded looks a little high, but they wont score many more than the 36 predicted. If Keane can get his defence to play well week-in and week-out he’ll have them in the 40-45 points bracket, if not they will be close to the drop.

Birmingham…19th place, 32 points

There is just not enough quality there and they have reminded us of Sheffield United so far this season. They play well as a team and give some opponents problems, but simplly can’t turn that into wins. One only has to look at some of their results from the season so far to understand what we mean. They played well versus Chelsea but lost, they played a crude but clever game plan versus Liverpool and drew and they played well versus Sunderland but drew. There is little to suggest that that won’t continue and could spell disaster. Their defence does look better than one that is predicted to let in 62 goals, however, with good quality Premier League experience in centre-backs Liam Ridgewell and Johan Djourou and therefore could be much better than the model predicts.   The attack is lacking talent, however, and below Forrsell, Jerome, Kapo and Larsson there aren’t many players with a chance to rate above-average at that end of the pitch.

Derby … 20th place, 22 points

They were bottom in our Championship Prediction model used in the book and they are bottom here. Regardless of how many goals this outfit can score, their defence is horrible. Their early form in away games perhaps spells it out best – four games, four losses, no goals for, 16 goals against. They’ll be planning for next year in December at this rate.

About the Author

Oliver Anderson is the director of Sports Statistical Reviews Ltd which is an innovator in the collection, analysis and interpretation of progressive sport statistics.

If you enjoyed this article and are interested in football statistics then we recommend you buy The Football Review 2007 which investigates each Premier League club and player in detail as well as looking at other hot topics that defined the season.