Monday, 21 November 2011

A Century of Competitive Balance and Dominance.

The competitive balance in the EPL and it's predecessor,the English Division One has been steadily becoming more unequal for the last 100 seasons.The share of all the good things that can happen to a team over the course of a season,the goals,the wins and the points are increasingly ending up in the hands of fewer teams.Whether you look at wins or success rate (wins + half the number of draws divided by total games),the standard deviation,which measures the dispersion from the mean or average value has trended to larger figures for most of the last century and continued into the present one.A larger standard deviation indicates that the data points are spread out over a larger range,in other words the top teams are grabbing more wins than their counterparts did previously.

The upward trend can be seen if we plot the seasonal standard deviation for the top division's success rate from 1920 onwards.Seasonal variation appears to have been more volatile in the immediate post war period,but the race to a more polarised sharing of the spoils is less steep in the league's earlier years.In the EPL era and especially from the noughties onwards,season on season variation is much more consistent,but the steepness of the trend curve is more relentless.

The Dispersion of the Success Rate per Season in the English Division One 1920-2010.

We can illustrate how unequal a league has become if we compare the actual standard deviations to the theoretical values that would occur if the league consisted of entirely equal teams.A league consisting of equally talented teams would see each team on average having a seasonal success rate of 0.5.However,random chance would dictate that while some would end individual years with exactly that figure,other teams would have slightly more or less than 0.5.The size of this dispersion around 0.5 would determine the standard deviation for a coin toss season and by comparing this random SD with actual SD's from real seasons,we can gauge how unbalanced the real life situation is.Repeated simulations of a coin toss season,for either a 38 game season (as the EPL has predominately been) or a 42 game season for the old First Division gives standard deviations of around 0.08 for team success rates.

The standard deviation for success rate over the period of topflight English league football post World War One averages out at 0.105 an increase of 34% on the value obtained from a league decide entirely by random chance.However,the trendline shows a distinct upturn after the formation of the EPL in 1992/93 and this is borne out by the average SD increasing to 0.132 if we just consider the EPL years.An increase this time of  just over 60% compared to a perfectly equal initial state.

If we refer back to the graph we can say that in the early post WW1 years some seasons produced tables that were not inconsistent with the league members being extremely closely matched in terms of talent.As we moved through the decades the disparity of talent became more pronounced and in the EPL era,talent and therefore proportion of wins became concentrated amongst a select group of four,possibly five teams.

By contrast with the current EPL,the NFL actively seeks to prevent the concentration of talent in a few teams and tries to promote parity of opportunity.If we repeat the process using NFL wins and ties mimicking the EPL's success rate and further account for a season that has just 16 regular season games we find that since the NFL AFL merger in 1970 the expected win/tie standard deviation was about 0.125.The actual average SD over the same period was 0.19,an increase of about 50% compared to a coin flip version of the NFL.

The Dispersion of Success Rates for various Sports Leagues.

LEAGUE. Success
Rate SD for a Coin Flip League.
Success Rate SD.
% Increase
0.125 0.190 52%
Div 1/EPL
0.078 0.105 34%
 to Present.
0.081 0.132 63%

So the English top flight football league from 1920 was actually more equal than the genetically engineered NFL since the merger,although the situation is reversed if we consider standard deviations for just the EPL since it's inception.The EPL has been won by just four teams over it's lifetime,Man U,Arsenal,Chelsea and Blackburn,while the NFL has had 12 different teams lift the Vince Lombardi Trophy over the same 19 year period.These numbers appear to confirm that the NFL is indeed more egalitarian than the EPL.The standard deviation of NFL team's seasonal success rate is proportionally closer to a coin flip league than is the EPL and many more EPL teams than the thirty or so American football teams who compete in the NFL have tried to win the EPL title and only 4 have actually succeeded compared to three times that number for the NFL.

However,the parity in the NFL is partly an illusion.The knockout nature of the NFL post season,even with the inbuilt advantages afforded to the highest seeds,means that teams who dominate the regular season often fall by the wayside in the Wildcard,Divisional,Championship and Superbowl stages.The wider spread of ultimate winners in the NFL is partly down to the drive for a level playing field through the elaborate salary cap and draft system,but also it's down to the NFL playing a condensed version of the FA Cup as an encore.

Secondly,the draft system means that many teams go through cycles of rebuilding.Poor teams with little chance of competing are always present and at best it is the identity of the poorest teams that changes.The constant churning from good to bad and back again can be as tediously predictable as a title fought for by the same four teams.

Thirdly,the poorest NFL teams have little to play for once they are out of contention in their division other than a contrived contest to finish as badly as possible to claim the highest draft pick.By contrast the average and below EPL teams,while accepting that they are almost certain to have very little title aspirations,do fight out a very real and season long contest to maintain their EPL status.A contest often as compelling as the battle at the top of the table.A Wigan fan can approach a new season with as much anticipation as a Manchester United fan,it's just their expectations that are different.

So although the figures appear to support the view that the NFL has more parity than the EPL,it does not inevitably follow that it is also a more appealing product*.Competitive balance is but one aspect of a league's attractiveness.Having a dominant and unchanging presence in a league can be good for the health of a league,even if that team skews the competitive balance.Hardcore fans welcome the occasional visit of the very best,even if a favourable result is much less likely and casual fans also care not about seeing a close game,but they too prefer to see the best.Given the choice of a closely matched game or a game featuring Manchester United where the outcome is much more certain,terrestrial TV channels in the UK almost always chose the latter and are rewarded with higher audiences.

*(appealing from the viewpoint of being competitively balanced......both are of course tremendous sports).

Sport and football is greatly affected by randomness even in a league lacking parity and the danger is that an ill considered drive towards equality of opportunity could leave fans as indifferent to a coin toss league as they are supposedly becoming when faced with a league dominated by a small select band of teams.Competitive balance within a league is but one aspect that drives that league's popularity.

Finally,it seems a shame to calculate all these standard deviations and let the go to waste,so as a post script I've calculated standard scores of success rates for each EPL and Division One team since 1920.Standard scores can be used to compare performances from different sets of data,they tell us how many standard deviations a data point is away from that season's average.They cannot tell us that the Champions from 1920 are inherently superior to the champions from 1999,but they can be used to compare how dominant each team was in it's own unique environment.Tactical awareness,fitness levels and the size and distribution of the skill pool obviously changes over time and teams can only perform in their own eras,so we can merely subjectively speculate on how an outstanding team from one era would perform in another.But standard scores can tell us which teams were the most dominant against the opposition they actually faced.

The Most Dominant English League Champions 1920-2011.

The team to dominate a Division One season more than anyone else was Sheffield Wednesday in 1929-30.In an era of two points for a win,The Owls were ten points clear of second placed Derby and 13 clear of third placed Manchester City.To illustrate the scale of Sheffield's achievement,the gap between Manchester City and bottom team Everton was also 13 points.They are followed by Manchester United's Busby Babes from 1955-56 and the list also includes the Danny Blanchflower double winning Spurs,an Everton side featuring their very own goalscoring statistical outlier,Dixie Dean,Don Revie's "dirty" Leeds,inherited and dismantled by Cloughie,a Cantona inspired Man U,an unbeaten Arsenal team and a Mourhinio managed Chelsea.

But by far the most poignant entry is third placed Sunderland.In an era where goal keepers were little protected by the Laws of the Game and sometimes even less so by lax refereeing,their young keeper Jimmy Thorpe died mid way through the season.Officially the cause of death was his diabetic condition,but the reality was that he died through injuries received whilst fielding a backpass in a league game the previous weekend.Increased protection for keepers were quietly ushered into the sport soon afterwards.

The Most Dominant 11 English League Champions 1920-2011.

TEAM. Success Rate. Standard
Sheffield Wed. 0.71 3.13 1929/30.
Manchester Utd. 0.71 3.04 1955/56.
Sunderland. 0.67 2.79 1935/36.
Leeds Utd. 0.74 2.71 1973/74.
Arsenal. 0.70 2.65 1933/34.
Everton. 0.63 2.64 1927/28.
Manchester Utd. 0.71 2.62 1992/93.
Huddersfield. 0.68 2.62 1925/26.
Arsenal. 0.84 2.61 2003/04.
Spurs. 0.79 2.57 1960/61.
Chelsea. 0.87 2.57 2004/05.

Sheffield Wednesday appeared to work on the assumption that fans remember either the very best or the very worst,as they also top the table for the top flight team cast furthest adrift from their competitors with an abject 1954-55 effort.They are chased home by Stoke who won just three games in 1984-85,although two of their victims were Arsenal and Manchester United.

The Worst Individual Seasonal Performers 1920-2011.

TEAM. Success Rate. Standard
Sheffield Wed. 0.31 -2.88 1954/55.
Stoke City. 0.17 -2.76 1984/85.
Sheffield Wed. 0.27 -2.66 1919/20.
Grimsby. 0.26 -2.53 1947/48.
Leeds Utd. 0.21 -2.41 1946/47.
Crystal Palace. 0.23 -2.38 1980/81.
Blackburn. 0.24 -2.36 1965/66.
Portsmouth. 0.25 -2.32 1958/59.
Leyton Orient. 0.25 -2.30 1962/63.
Wolves. 0.27 -2.29 1983/84.
Sunderland. 0.20 -2.24 2002/03.

No comments:

Post a Comment