Go figure II – defence vs. attack revisited
Two years ago this month, I penned a piece for this column that investigated which of football’s two most fundamental elements – scoring goals, and ensuring that the opponents do not – bore the closest association to points accumulated over a given season in a given national league. The correlation figures generated as a means to this analysis provide interesting evidence with which to apply hindsight to individual years, but when considered collectively, hint toward a hypothesis with deeper, wide-ranging implications. Attack is usually more significant than defence, but when a certain league becomes the dominant force, this trend is reversed.
Correlation coefficients (Premier League, England)
1997/98: 0.905, -0.822 Attack most influential
1998/99: 0.827, -0.845 Defence
1999/00: 0.852, -0.806 Attack
2000/01: 0.905, -0.901 Attack
2001/02: 0.911, -0.798 Attack
2002/03: 0.931, -0.833 Attack
2003/04: 0.889, -0.899 Defence
2004/05: 0.837, -0.891 Defence
2005/06: 0.914, -0.914 No difference
2006/07: 0.954, -0.859 Attack
2007/08: 0.879, -0.945 Defence
Above is a list of the coefficients representing the positive correlation between goals scored and points achieved, followed by the negative correlation between goals conceded and points achieved, for each of the eleven English Premier League seasons originally surveyed. In very simplistic terms, a correlation coefficient is a statistical measure that gives a figure between -1 and 1, and the closer it is to either extreme, the stronger the relationship between the variables in question. So, for example, a positive correlation of 0.905 in 1997/98 suggests that season’s goals for column had a stronger bearing on league position than goals against, since the negative correlation for the latter was only -0.822.
A trend immediately apparent in this list is that while attack translated to points more effectively than defence in five of the six seasons between 1997 and 2003, this was only the case once in the subsequent five seasons to 2008. The turning point corresponds almost perfectly with English football’s ascendancy to relative dominance over the UEFA Champions League – with Manchester United’s 2008 victory over Chelsea in Moscow the culmination of a five-year period in which the Premier League had produced two European champions, three runners-up, and five beaten semi-finalists. Before 2003/04, there had been just one title for United in 1999 and a combined three other semi-final appearances since English clubs had been invited back into the continental fold post-Heysel in 1990.
Correlation coefficients (Premier League, England)
2008/09: 0.925, -0.925 No difference
2009/10: 0.895, -0.879 Attack
2010/11: 0.858, -0.808 Attack*
(* 2010/11 figures correct up to and including Fulham vs. Chelsea, 14 February 2011)
The reason I felt this topic worth revisiting – he says, 400-odd words in – is that the timing of the previous article in February 2009 now appears to have coincided with a bit of a high water mark, for the time being anyway, in both English and Japanese club football. As it had the previous season, the Premier League provided three of the four UEFA Champions League semi-finalists in spring 2009, but on this occasion, those not eliminated by a domestic rival were accounted for by the magnificent Barcelona side of Pep Guardiola. A year later, only two teams from England made it as far as the last eight – both fell at this hurdle – as the seemingly eternal concept of the ‘Big 4’ suddenly began to disintegrate. Each of its former members (with the possible exception of Arsenal, who nonetheless dropped off the pace for two seasons before bouncing back in style this term) can reminisce fondly about the respective peaks they reached in the latter part of the previous decade but have since slipped from.
Pleasingly – for fans of a mathematical hypothesis if not a supposed elite football team – this retrogression is reflected in the correlation coefficients. When Barcelona finally broke the English hegemony in the final season before Cristiano Ronaldo too headed for Spain, there was no longer any difference (to three significant figures) between the influence of the goals for and against columns in the Premier League table. In 2009/10, attack regained its status as the more decisive factor for only the second time since 2003, and this turnaround is further underlined by the mid-season data for 2010/11 so far.
Correlation coefficients (J. League, Japan)
2005: 0.807, -0.710 Attack
2006: 0.877, -0.740 Attack
2007: 0.820, -0.795 Attack
2008: 0.667, -0.810 Defence
2009: 0.839, -0.648 Attack
2010: 0.817, -0.776 Attack
Though its data range is limited by the less analysis-friendly two-stage season format that persisted until 2004, the J. League has shown near-identical trends in terms of both regional performance and domestic correlations over the past few years. Scoring goals was originally a clearly more effective way of winning points than keeping things tight at the other end, but this tendency became much less significant when Urawa Reds won the AFC Champions League (ACL) in 2007, and was comprehensively reversed the following year as Gamba Osaka romped to Asian glory – beating Urawa in an all-Japanese semi-final. Like England, however, Japan’s clubs have quickly lost their continental superiority, with Nagoya Grampus the only ACL semi-finalist in 2009 and nobody even making it past the round of 16 last year. Over this time, attack has once again replaced defence as the largest contributor to league points.
Correlation coefficients (K-League, South Korea)
2007: 0.785, -0.806 Defence
2008: 0.769, -0.749 Attack
2009: 0.834, -0.662 Attack
2010: 0.924, -0.873 Attack
So, what of Japan’s biggest rivals, on the opposite side of the sea on whose name they still cannot agree? Unfortunately, the K-League offers the smallest (and thus least reliable) data set of all, with the current single-stage format only in place for the past four seasons and just 14 (to 2008) or 15 teams (to 2010) competing. Over this period, defence was the more significant factor in 2007 alone, before the goals for column took precedence thereafter. South Korean fans could make the case that their league had not yet been knocked off its perch for most of 2007, with Jeonbuk Hyundai Motors entering the year as Asian champions (having overcome compatriots Ulsan Hyundai Horang-i en route, in the last four) and Urawa needing a tense penalty shootout to finally see off Seongnam Ilhwa Chunma in their ACL semi-final that autumn. Two successive continental titles for K-League clubs in 2009 and 2010, however, do not appear to have had much impact on the correlation coefficients above.
Perhaps, though, this latter point should bring us back to the word ‘dominant’ within our original hypothesis. When Pohang Steelers won the ACL in 2009, they had been the only remaining Korean side in the last four; ditto Seongnam last year, even though all four K-League representatives had reached the quarter-finals. The East Asian section and Seongnam’s performances in particular certainly made club football in South Korea look superior to Japan, but overall results suggest it would be premature to proclaim a dominant force. This provides a handy explanation for the ‘goals for’ correlation remaining stronger, and is further supported by evidence from Europe.
Correlation coefficients (other European leagues, 2009/10)
Germany (Bundesliga 1): 0.861, -0.824 Attack
Italy (Serie A): 0.856, -0.741 Attack
Spain (Primera Liga): 0.941, -0.811 Attack
The hegemony of the Premier League may be over, but in an era of few really good European club teams, no one country has yet stepped up to present a clear frontrunner as far as domestic competitions are concerned. As such, attack remains a more significant contributor to points gained than defence in each of the other three leagues to have produced Champions League finalists over the past two seasons (and, indeed, since 2005); just as was the case in the original article two years ago. Of course, the figures may be influenced by the respective styles of football common to each nation, but our hypothesis still comfortably passes the test if we consider the genuinely dominant European leagues of the past few decades.
Correlation coefficients (dominant European leagues)
Italy (Serie A) – 2002/03: 0.883, -0.916 Defence
Spain (Primera Liga) – 1999/2000: 0.640, -0.827 Defence
Italy (Serie A) – 1993/94: 0.658, -0.889 Defence
England (First Division) – 1980/81: 0.706, -0.856 Defence
Internazionale are the reigning Italian, European, and world champions, but the last time you could argue hands-down that Serie A was the best around was 2002/03, when AC Milan won a last four derby to set up an all-Italian Champions League final with Juventus. Three years earlier, Real Madrid beat Valencia in the first ever final to feature two clubs from the same country, with Barcelona falling to the latter in the semis, and Deportivo La Coruña pipping the lot to clinch their first ever La Liga crown. 1993/94 was perhaps the peak of the Football Italia era, with Fabio Capello’s Milan winning a ridiculously strong Serie A for the third year in a row before demolishing the Barcelona ‘Dream Team’ 4-0 in Athens. Finally, way back in 1980/81, Liverpool secured the fifth in an unprecedented run of six consecutive European Cups for English clubs, with Bobby Robson’s Ipswich Town taking the UEFA Cup for good measure. In all four cases, the goals against column had the greater influence on points accumulated, and often considerably so (even if Milan’s miserly record of just 36 goals scored and 15 conceded over 34 league games in 1993/94 suggests the trend isn’t necessarily conducive to greater entertainment for the neutral).
So, what now?
Without wishing to recycle the Benjamin Disraeli and Homer Simpson quotes with which I concluded the previous article on this topic, the caveat remains that the sample discussed above is relatively small and thus not necessarily all-telling. But the hypothesis does at least appear to provide quantitative support for a trend, qualitatively evident in events on the pitch over the past few seasons, which has left the positions of ‘dominant’ league in both Europe and Asia very much up for grabs. With the former, periods of domination have been cyclical in the past and it seems reasonable to expect a continuation, but the Asian situation is rather different as countries like Japan and South Korea look to further their impressive development in global terms. As a consequence of rapid improvement at both club and international level, a succession of East Asian players are now transferring to and thriving in higher-profile European climes. It may, therefore, be desirable for the J. and K-Leagues to weaken slightly in the short term in order to reach yet higher levels in future.
And if anybody wants to commission a more exhaustive survey of these correlation trends, they know where to find me.
TrackBack URL for this entry:
Listed below are links to weblogs that reference Go figure II – defence vs. attack revisited: