Tuesday 30 August 2011

Part 3: 22 is the Magic Number


Since my last post, I have been compiling some additional data from previous seasons using the same model.  It has turned out some quite interesting data that I thought I should share.





In each of the last three seasons, the eventual Premier League Champions have won 22 of their 26 ‘most-winnable’ matches.  However, there is no discernable pattern in the results of the one-point-target matches.  In fact, only 13 of the 36 par-1 matches played by the title winners over the last three year were in-fact drawn.

I can only conclude that tracking the results of the most winnable matches is the most valuable indicator of a title assault, as there is a consistent level of performance by the title winning teams in these fixtures.  To win the league, you need to win 22 of these matches.

Looking at the results of the top four teams over the last three seasons in the par-1 matches shows some quite interesting data.

2008-09

1st - Man Utd – 22 points
2nd - Liverpool – 26 points
3rd - Chelsea – 25 points
4th - Arsenal – 21 points

2009-10

1st – Chelsea – 16 points
2nd – Man Utd – 19 points
3rd – Arsenal – 17 points
4th – Spurs – 18 points

2010-11

1st  – Man Utd – 11 points
2nd – Chelsea – 10 points
3rd – Man City – 10 points
4th – Arsenal – 19 points

There is actually quite a bit of variance in the points accrued from these matches from one year to the next, but between the teams within each season, there is a remarkable level of similarity.  There is also no correlation between the points gained in these matches and the final league standings.

Finally, here is the up-to-date Alternative Premier League Table, which shows Man City going two ahead of target and drawing level with Man Utd, while Arsenal and Spurs are both sitting below par, and below the 71-points target line.



Monday 22 August 2011

Part 2: Matches 1 and 2

In my opening article, I wrote about the importance of not dropping points in relation to target in consecutive matches.  A loss at The Emirates would have seen Liverpool doing just that, however, in a controlled performance, two late goals saw Liverpool win there for the first time.  This win pulled back the two points dropped in the opening fixture against Sunderland, which was a bad result regardless of how you look at it.  Liverpool are now back on target for a 90 point season.  Like in golf, bounce-back-ability is a sign of a winning mentality.

Also on target for 90 points after their opening two matches are Manchester City and Chelsea.  Although Chelsea drew their opening match, an away point at Stoke City is a satisfactory result for any team (despite what some people would have you believe), and in this model has a target score of one point.  Manchester City on the other hand have had back-to-back wins, but nothing more than they would expect against Swansea (H) and Bolton (A).

The biggest losers from the opening two matches are Arsenal.  Although their away draw at Newcastle is a target result, any home loss is a very poor result.  As I mention above, avoiding consecutive dropped points (in relation to target) is essential to mounting a title challenge, and they will be delighted to leave Old Trafford next weekend with a point.

The big winners so far are Man Utd with two points gained over target in their opening match away at West Brom.  They were also able to consolidate their gain with an important win at home to Spurs, who themselves have dropped a point in their only match so far.

To illustrate the Alternative Premier League Table throughout the season, I will be keeping the following graph up-to-date.  I have also presented a close up version.  As Man City and Chelsea track the green 90 point target line, their traces are hidden.



Top Four Target

A few people have been discussing various ways this model can be adapted to track a top four target.  However, the way I personally like to look at it is that teams that come top four are able to play with title winning form for parts of the season, but not all, whereas the title winners are able to produce more consistency.  Therefore, on the graph above, you might see some drops followed by plateaus in top four teams.  Their target results should remain the same, however, you just have to apply a handicap like in golf (going back to my original analogy).
To illustrate this, I have presented a descending line to illustrate a 71 point target (19 point handicap), which means that 0.5 points per game can be dropped in relation to target over the course of the season.  As I mention above, teams won’t follow this line, but need to be above it by the end of the season.
Below is an updated version of the 08-09 season data with this 71 point line added. You can see that Liverpool had two title winning standard runs, up to game 20 and from 24 to the end.  Chelsea fell below the 71 point line after game 25, then had an above title winning finish to the season.  Arsenal were below the 71 point line for most of the season, but had title winning form for the last 11 games to sneak above the 71 point line.

 
The importance of winning the three point matches

I have been looking through some past results.  The following tables show the number of three pointers and one pointers won drawn and lost in the 08-09 and 09-10 seasons for the top four teams.  I know the sample is small (I haven’t had time to look at other seasons yet), but in both these seasons, the champions (08-09 – Man Utd, 09-10 – Chelsea) won 22 of their 26 three pointers, which I think adds weight to this model.

2008-09




Tuesday 9 August 2011

Part 1: Introduction

My motivation for starting this blog is based on my feeling that league tables don’t tell the full picture during a season as they don’t take into account the teams played so far neither do they illustrate whether the fixtures were home or away.  Over the course of this season, I will be updating this blog to apply a simple model that I use to evaluate league standings based on the matches played.  I first started looking at league tables in this way about four years ago, by adapting the old theory that in order to win the league, a team needs to win all of their home games and draw all of their away games.

There are a few assumptions I have to make to keep this model as simple as possible.   Firstly, I assume that 90 points should be a target points total for a title challenge.  You can't control league position entirely yourself, so focussing on points makes more sense to me.  If a team earns 90 points this season and doesn’t win the league, it would just mean another team has done exceptionally well.

In order to achieve 90 points, I make the assumption that a team needs to win every home game (57 points) and their 7 'easiest'* away games (21 points), drawing the remaining 12 away games (12 points).  This means that a draw for Liverpool away at Old Trafford is a par** result for a title winning team, with a win there being 2-over-par, but a draw at home to Norwich is 2-under-par.  During the season (as with during a round of golf to use that analogy further) it is unlikely for you to play to par in every match (hole), but any dropped points need to be made up elsewhere.

* I assume the 7 'easiest' away games are against the teams that finished between 17th and 14th last season and the three promoted teams as I need to determine my par scores before the season starts (I appreciate this isn't ideal, but I have to start somewhere).
** If a team is on par after 38 games, they will hit 90 points, if they are over par after 38 games, they will have more than 90 points (unlike golf, in this model, over par is good, under par is bad).

During the season, I will be plotting the results to par of all of the top six teams from last season, to see how the fixtures played affects the league position.

To get things going, this is a graph to illustrate how all of the top six teams’ fixtures differ during the coming season.  It shows Chelsea and Tottenham should both accrue 47 points after 19 games (not taking into account any rearranged games that affect the order of fixtures), whereas Liverpool should only accrue 41 points if they are to be ‘on target’ for a league title.  In fact, it shows that after 20 games, Liverpool’s target number of points is eight fewer than Chelsea and Tottenham’s (42 compared to 50).  Therefore, in theory, if Liverpool are less than eight points behind either of these teams at this point in the season, they would in fact be ahead of them in relation to their target.  As a matter of interest, Man Utd’s, Arsenal’s and Man City’s 20 game targets are all 46 points.



It is quite interesting to look at how teams have performed in relation to their theoretical target in previous seasons.

This is a graph illustrating the top four teams in the Premier League during the 2008-9 season, the closest Benitez’s Liverpool came to the title:



I think the key thing it indicates is that Liverpool had three significant crashes in their points total, where they dropped points in relation to target in consecutive games.

The first is matches 14 and 15 (home games to Fulham and West Ham), which they bounced back from well with a win away at Blackburn.  The second is matches 21-23, where they drew to Stoke (A), Everton (H) and Wigan (A), which was followed with a home win to Chelsea and a 3-2 win away at Portsmouth.  The final one was matches 26 and 27, home to Man City and a 2-0 loss away at Middlesbrough.

Man Utd only had one of these crashes, which only lasted 2 games in matches 28 and 29.  This was when a loss away at Fulham followed Liverpool’s 4-1 win at Old Trafford.  Significantly, this occurred when they had a 12 point buffer over Liverpool.

The key lesson in this is that every team has bad results during the season, but limiting these to isolated games is the key to a successful title challenge.  This doesn't mean dropping points in consecutive games is bad, if the target result is achieved.  In matches 3 and 4, Man Utd lost to Liverpool at Anfield, then drew away at Chelsea, and although this was 5 potential points missed out on, in the model, they only dropped 1 point, which they more than made up in match 6 with a win away at Blackburn.

A summary of how each of the teams performed in their 3-point-target games and their 1-point-target games also seems to indicate a trait of a title winning team (although I would need to look at data from more than one season to test this further).



I’ve often heard people talk about the importance of winning against the big teams if you want to win the league.  However, Man Utd got zero points off Liverpool in 2008/9, 1 off Arsenal and 4 off Chelsea.  In the same season, Liverpool took 6 off Chelsea and Man Utd and 2 off Arsenal.

What the data above does seem to indicate, is the need to win more of the 'most winnable' games.  In 2008/9, Man Utd won 22 out of 26 of these, Liverpool only won 17.  The model isn't saying that drawing is desirable, just that it isn't that bad to draw away from home against the better teams.
Looking again at the fixtures for this season shows that Liverpool’s first eight away matches are all 1-point-target matches.  Although this means they have less opportunity to ‘make up’ dropped points in the second half of the season, it also means that falling off the pace slightly (in terms of absolute standings) shouldn’t be seen as disastrous to any title challenge.  In fact, it might just work in Liverpool’s favour, as they might have less pressure, and an (on paper) easier second half of the season.