Chelsea's League Form Under Villas Boas and Di Matteo.

This season's highest profile managerial departure rather predictably came at Chelsea following their 1-0 defeat away to WBA on the 3rd of March.The defeat left Chelsea three points adrift of fourth placed Arsenal,seven shy of Tottenham and confirmed their irrelevance in the Championship race where Manchester City were already swinging away in a 20 point lead over The Blues.

After an opening day draw at Stoke,an early season run of four wins in all competitions hadn't been bettered and fatally for Villas Boas two mid season four game sequences of winless games had recently stretched to one of five games.In the currency most usually used to describe a team's well being,his Chelsea team had averaged just 1.7 points per game.Fantastic if you were managing a mid table team on a budget,but a totally unacceptable level of form for a team used to challenging for top spot.

Abramovich must enter every season with a mind to a replacement candidate for his top job should the need arise and Roberto Di Matteo was swiftly promoted from within.He has now taken charge of Chelsea for ten Premiership matches and has interspersed this mundane fayre with a Wembley rout of their near rivals,Spurs and Cup Final victory over Liverpool and has plotted a course past Barcelona to reach the Champions league final against Bayern Munich.

However,critics of Di Matteo are quick to point out that his points per game haul is inferior to that of Villas Boas.In the Premiership Chelsea have average 1.5 points per game since the mid season sacking and although they have around a 36% chance of qualifying for next year's Champions League as holders,they cannot do so by their EPL position where they can finish no higher than 6th.Had they kept accumulating points at the pre sacking rate of 1.7 per game,they would current have around 65 points,making their chances of qualification through the EPL route somewhat on par with the chance currently enjoyed by Newcastle.

Points per game averages can be skewed by the arbitrary awarding of three points for a win.Few would argue that a win is three times better than a draw,so a more objective approach is to look at a team's success rate,where a draw is treated as half a win.If we do this for each portion of Chelsea's season we again see AVB coming out on top of RDM in performance in the Premiership with success rate mirroring the difference in points per game for each manager.

So by both standards,AVB appears to have got more out of Chelsea on a league day than did RDM in subsequent midweeks and weekends.

Chelsea's EPL Raw Points Per Game and Success Rate Record,2011/12.

Chelsea Team Performance.	Villas Boas.	Di Matteo.
Points per Game in Premiership.	1.70	1.50
Success Rate in Premiership.	0.61	0.55
Success Rate.All Games.	0.60	0.74

We could leave the analysis there and conclude that AVB was unlucky to be sacked and his replacement has failed to even reach the level of performance that he extracted from an ageing Chelsea team,ripe for change.
However,RDM's detractors have failed to credit him for the additional cup games he has overseen.If we include FA,Carling Cup and Champions League matches in the records of both managers,we not only increase the sample sizes of the games we are looking at,but we also see a large shift in the two success rates.Di Matteo is now the more successful coach and by a significant margin and although his record now contains relatively easy games against Birmingham and Leicester,these are counterbalanced by two games against Barcelona.

We can continue this theme of strength of opponent by returning to only EPL games and once we correct for opponent strength even the argument that AVB outperformed his replacement in EPL matches begins to look much less secure.The average league position of Chelsea's opponents at the time when the game was played when AVB was selecting the side was 11.5,by contrast RDiM's Chelsea faced sides with an average league status of just over 9th.So latterly Chelsea have gained less points per game,but they have on average faced stronger opposition.

We can attempt to quantify this effect by devising match odds from 30 game rolling averages for each team in all of Chelsea's EPL league games played so far this year and compare the expected results against the actual results for the reign of each manager.If we do this we shouldn't be surprised to see that The Blues have underperformed all season,but this declining EPL track record was slightly more pronounced when AVB was in charge than when RDiM was selecting the side once opponent strength was accounted for.So under a system that takes into account who Chelsea played,RDiM is now also the more successful Premiership coach as well as bettering AVB overall.However,...............

Chelsea's EPL Record Corrected for Opponent Strength.

Chelsea Performance.	Andre Expected.	Andre Actual	% Under Performance	Roberto Expected.	Roberto. Actual.	% Under Performance.
Points Per Game in Premiership.	2.11	1.70	19.4	1.84	1.50	18.5
Success Rate in Premiership.	0.74	0.61	17.6	0.65	0.55	15.4

........this post shouldn't be taken as a declaration that AVB was inferior to RDiM because we simply don't have enough data points to be sure.The 27 EPL games for which the former was in charge is barely enough to gain a decent estimate of a team's worth and the ten game league stint enjoyed so far by his replacement is of insufficient size to make any definite statements.But it should demonstrate that points per game averages are no more than rough estimates,even over dozens of games and are often completely misleading as a means of presenting an opinion about a team or manager.The two men's points per game average alone will tell us very little about their relative quality.Indeed if Chelsea beat Blackburn at the Bridge on Sunday,DiM's average will approach that of his predecessor partly because he has been in charge for so few league games.

At the very least a rudimentary correction for opponent strength needs to be made and even then any conclusions should be tentative rather than definitive.                

Chelsea Shouldn't Chase Clean Sheets.

Ever since Chelsea's Performance Director mentioned in an interview that "there is a stronger correlation between clean sheets and where you finish than goals scored and where you finish.",there's been a spate of blog posts extolling the virtues of clean sheets.To be fair to the guy he didn't explicitly say that Chelsea were actively using strategies designed to increase their number of clean sheets,but there was certainly an implication that more clean sheets were always a good thing to aspire to.


At first glance the evidence seems compelling.If you keep a clean sheet,you're obviously not going to lose.Score and it's three points and on average you'll pick up upwards of 2.5 points from matches where your opponents remain scoreless.However,by looking at how many points a team gets when they successfully keep their goal intact you are omitting those games where a team goes out with this defensive strategy and it fails.The question a Performance Director should be asking is are the points we get when the strategy works sufficient to compensate for the occasions when we concede a goal (or two) and find ourselves with insufficient time or offensive resources to make an adequate response.You've got to account for the games where you adopt a passive approach,but succumb to a late goal and are unable to reply with a goal or two of your own.


In short,is an overtly defensive strategy the best way for every team ?


If your going to go for clean sheets your games are likely to have less goals,so let's see how a team such as Chelsea fare in first an environment where less goals are scored.Chelsea typically score 70% of the goals scored in their games,so in a low scoring,clean sheet orientated environment where say 2 goals a game on average are scored,they will score 1.4 goals per game and allow 0.6 goals.If we now simulate season after season using these parameters we find that Chelsea,if they adopt a clean sheet strategy will gain on average of 1.97 points per game.If we now simulate seasons where Chelsea still score on average 70% of the goals scored in a game,but this time they adopt a less defensively minded approach and their games now on average see 3 goals per game scored.In an environment where goals,both scored and allowed are more likely and clean sheets are less common a much superior team will see their points per game jump to 2.15.


An average of 2 total goals per game is abnormally low,but it has been chosen to emphasis the effect of a changing goal environment.However,the trend for a superior team to do better in a more goal laden situation is consistent.Over more typical ranges an attacking approach will yield a team of Chelsea's quality around 3 points a season more than if they opt to try to prevent their opponents from scoring.


Chelsea have the wealth and prestige to attract the best attacking talent available,therefore they would be better served forgetting about trying to increase their quota of clean sheets and instead take a more attacking approach and challenge their opponents to keep up.It'll get them more points in the long run and they should leave overtly defensive strategies to teams with less attacking resources.


Footnote.Headline from Daily Mail following Chelsea conceding a late equaliser in a 1-1 draw at Wigan."Chelsea paid the price for negative play, admits AVB after dropping points at Wigan."

Read more: http://www.dailymail.co.uk/sport/football/article-2075628/Andre-Villas-Boas-Chelsea-paid-negative-play-Wigan.html#ixzz1gw4KAOVK

Red Cards and the Big Four.

The win probability for each team is tracked on a minute by minute basis throughout each game.At any point in a game a team will have a probability of winning the game outright and an associated probability of drawing the game.By multiplying the probability of winning the game by 3 and the probability of drawing by 1 and adding the results together you will get an expected longterm average for the number of points that team will achieve from that position or Expected Points (EP) for short.

Every football fan can remember games or incidents within games that are especially significant for their club.They usually signify that their team has reached a new high in it's development or plummeted to new and possibly permanent depths of ineptitude.Stoke fans would probably put November 1st 2008 into the former category.The initial 45 minutes of their first EPL season where they shipped a goal every 15 minutes was still a raw and recent memory,but they had partially steadied the ship with wins over Villa,Sunderland and an obliging nine man Tottenham.They occupied 15th place in the EPL,the average finishing spot for promoted teams in their first year in the higher grade,but defeats at the hands of Bolton, Middlesbro', Everton, Chelsea, Man City and Portsmouth meant that the visit of third placed Arsenal to the Britannia was a daunting prospect for the hosts.

Pre game expected points values made the Gunners huge favourites.They would expect to return with an average of 2.34 points from their visit north,while Stoke would be lucky to average half a point a game.

However,today was a day when one of the least favoured scenarios materialised.Fuller's 11th minute strike would rarely have been a decisive first goal and Stoke had to maintain that lead until after the half hour before they became the team favoured to clinch more points on average from that particular game situation.By the 70th minute Stoke still held the advantage and would be expected to gain an average of 2 points per game in such circumstances,but they tightened their grip three minutes later when Olofinjana stumbled his way into the box and bundled a trademark Delap throw over the line.Instant cult status for "Seyi".

Scorers;
1-0,Fuller,11'(Delap assist)
2-0,Olofinjana,73'(Delap assist)
Red Card,van Persie,Arsenal,76'
2-1,Clichy,90'

If Arsenal had been unaware of their underdog status from the half hour mark onwards,they certainly knew their likely fate now and three minutes later their predicament worsened even more.Stoke's keeper Thomas Sorensen deliberately dallied over picking up the ball in his box,inviting the loitering van Persie to participate in a game of cat and mouse where the Dutchman was by far the junior partner.Van Persie's half hearted lunge for the ball was never going to arrive before Sorensen claimed it for himself and what the subsequent shoulder charge lacked in violent intent it certainly made up for in stupidity.Sorensen collapsed in a heap,although he was careful to remain in front of his goal line and Stoke's favourite referring pantomime villain,Rob Styles set van Persie packing with a straight red card.The dismissal left Arsenal in a hole too deep for even a team of their attacking flair to emerge from and Clichy's 90th minute strike was merely a consolation.

The view  from the stands confirmed Stoke as a team who could compete on their own turf with the league's very best,dreams of a European Tour were a few seasons down the road,but the Potters were no longer considered to be a team out of their depth.Regarding the dismissal,it was difficult not to conclude that van Persie had reacted to the situation of trailing and most probably losing to a team who should have been dispatched with the minimum of fuss and in a manner Chelsea had achieved on the same ground a month earlier.It was a clear case of red card bourne out of frustration.

It's easy to attach too much importance to individual events,so to test the theory that the best teams see red when they fail to reach the heights they generally attain,I recorded all games involving the Big Four,Liverpool,Arsenal, ManUtd and Chelsea where a red card was shown to their players. I charted the pre game Expected Point Values and the EP for the team in the minute before their player was dismissed.Games between the big four were removed from the sample.

Games against inferior opposition where the big four received a red card.2004-2011.

Teams.	Number of Red Card Games.	Average EP at Kickoff	Average EP just before Red Card.	Average % fall in EP	% of Games where EP fell before Red Card.
Man Utd.	12	2.02	1.54	24	75
Arsenal.	15	2.09	1.73	17	67
Chelsea.	16	2.29	1.98	14	56
Liverpool.	6	1.83	1.01	45	83
Overall.	49	2.11	1.68	20	67

As can be seen the average Expected Points just prior to the red card are below the average EP at kickoff both for the big four as a group and for each individual team.The EP fell on average by 20% and in about 70% of the games it was below it's starting point.Games numbers are relatively small for each team,so conclusions are necessarily tentative.Liverpool appear to be the most bad tempered of the original big four,5 of their 6 red card games came when the Reds were performing worse than expected.Chelsea were the team who appeared least concerned with the current match position before they saw red.

The figures appear to confirm that teams who are playing badly by their expected standards are likely to compound the problem by having players dismissed.The question now becomes is the effect universal or is it more pronounced when those struggling to perform are the best teams in the division,after all they have more to lose than teams with lower pre match expectations and their under performance will be self-evident to their players.The average fall in EP at the time of the dismissal is around 20%,so in the next part of this series I'll take a representative sample of games not involving the big four and see if the red card situations remain the same.

Defensive Performance and Transfer Window Purchases.

At first glance it seems relatively easy to decide which teams have he strongest defences in the EPL.Goals allowed per game is almost universally used to demonstrate defensive ability.A goal conceded is to a large extent down to a team's defenders,so it stands to reason that the less you concede,the better you are.To add even more weight and justification to this choice is that,unlike say the NFL each team plays all the others twice,therefore the schedule is almost identical for everyone.

However,this approach has problems.Defences will face every other attack in the league twice over the season,but they are stuck with their own forward and midfielders for each game.And the quality of a team's other outfield players can impact on the amount of defending a team must do and this in turn can be reflected in the amount of goals they concede.

To take an obvious  example.Manchester United defenders at Old Trafford are unlikely to find themselves under the same amount of pressure experienced by their visiting counter parts.United's forward and attacking midfielders will likely have possession of the ball for a longtime,during which there will be little or no defending for United to do.In short overall team quality makes defending easier and not all teams can afford or attract the kind of quality throughout the team enjoyed by Manchester United and the other top teams.

We therefore need to use numbers that initially at least incorporate a team dimension before we try to decide how effective certain parts of a team perform their various tasks.

Average goal difference (AGD) per game is such a stat.It's the difference between the average number of goals per game a team scores and the average number of goals they concede and it is very strongly correlated to success indicators such as final league position.Over the 2010/11 season,Man Utd's average goal difference per game was +1.08,compared to -0.71 for relegated West Ham United.

AGD incorporates both average goals scored and allowed,so it is a genuine team statistic.A team can accumulate positive AGD by scoring goals and although their goal difference will always be eroded by conceding goals it is still advantageous to allow as few goals as possible to be scored against your side.

I've taken each EPL team's seasonal AGD  and plotted the numbers against the average number of goals per game allowed by those teams since the 1999/2000 season.As you can see from the graphic below AGD is strongly correlated to goals conceded.We would of course expect it to be.

From the graph we can derive an equation that relates the amount of goals allowed by a team for a certain level of average goal difference.This relationship gives us an opportunity to see how much of a team's AGD is derived from it's defensive abilities compared to the typical expectation for the EPL.Teams that are overachieving in relation to the norm can be considered to be performing extremely well,defensively regardless of their overall level of talent.

The Myth of the Clutch Goalscorer.

EP would appear to be the ideal tool to investigate whether a player has the ability to score high value goals almost on demand.However,before we imbue players with a special talent we first have to determine whether or not clutch ability exists in the first place.

It's very easy to take incomplete or sparse data and make seemingly logical leaps of faith to produce incorrect conclusions.Teams are often seen to be draw specialists using reasoning based on little more than one seasons worth of results during which a team draws a lot of games.However, if you look at enough teams you're almost certain to to see one set of results that are far enough from the usual average to justify the use of the specialist tag,when all you are really seeing is perfectly natural random variation.For the tag to be legitimate we need to be seeing the phenomena persist for many seasons.

Draw specialists are almost never that.If you take the teams who drew the most games in one season and then average those teams draws in the subsequent season you'll find that the figures always return to the league average.

Home or away specialists also go the same way,with maybe one or two very special exceptions.

So to make answering the clutch question more palatable I propose to firstly see if it's possible to identify any teams that exhibit the trait.Are they more likely to score goals that are important to the match outcome and does this ability persist.

The average goal difference per game for a team is a very good indicator for the number of points a team will achieve over he course of a season.In general terms for every 0.1 increase in goal difference a team will increase it's average points per game by 0.066 of a  point,put formally a team's goal difference per game multiplied by 0.66 plus 1.37 gives the average number of points per game a team will gain in the EPL.

Average Points Per Game = 0.66 * (average goal difference) + 1.37

If we therefore calculated the number of points that a team would have expected to gain over a season given it's goal difference and compare that to their actual points haul we can see which teams are out performing their goal difference.We can see which teams are scoring lots of important goals as opposed to lots of scoreline boosting ones.

As with so called draw specialists mentioned previously it's important the recognise that one season out performing the trend line doesn't make a team a clutch scoring side,natural variation is to be expected.We are looking for teams to continually out perform over a consistent series of seasons and then,just as crucially we need to try to identify any factors that support the designation we've given them.

The results would seem to suggest that teams can not consistently beat the goal difference by scoring a larger than expected number of important goals.Manchester United are fairly typical of the sample,over the last twelve seasons their points total over performed against goal difference on 7 occasions and underperformed on 5.Newcastle split 5 against 6 and Sunderland 5 against 4.Wolves have overperformed on each of their 3 premiership seasons,but even this has about a 1 in 7 chance of occurring by luck alone.

So clutch teams would appear to be rare if they exist at all.It's hard to imagine a reason why for example Wolves might be one.They are a fairly unremarkable EPL side,although one clue might lie in their astute manager who clearly prioritises winnable games.So perhaps apparent overperformance could be a trait of the manager.

If we look at Chelsea's record it at first glance appears fairly average.Their split is 7 and 5 in favour of overperformance.However,there is a run of three seasons in the mid 2000's coinciding with  Mourhinio's reign when they overperformed by around 5 league points in each of those 3 years.Pre and post Mourhinio they had been a 50:50 split side.

If we delve deeper into Chelsea's record we find that they were indeed more likely to score vital goals at vital times.The chances that a team scores the next goal is strongly tied to their chances of winning the game and so it is with Chelsea.If a next goal was scored in the Mourhinio years there was a 77% chance that Chelsea would score it.However,when Chelsea trailed or the game was tied (other than 0-0's) then the chance that the next goal would come from Chelsea was in excess of 87%.It could be argued that Mourhinio made tactical changes when trailing or drawing late in the game that made scoring more likely,his team carried this out and then together they cemented the gains made by these high EP goals and that was reflected in the final table......or he and they could have been lucky,3 consecutive "up" seasons are still just a 1 in 7 random event.

Except he then did the same in his 2 years at Inter,bringing the total to 5 consecutive seasons,one chance in 32 by luck.

If we look at the EP for all of Chelsea/Morhinio games,we'll probably find a Chelsea player who appears clutch because he scores a higher proportion of important goals.But the truth is possibly more complex.He more likely scores more vital goals because the team embraced the tactical changes,upped their game pace to higher but ultimately unsustainable levels (a very common and obvious occurrence near to the interval in rugby for example) and gave him greater opportunity to score.

If it exists,it's a team and manager trait.
.