3PW And Goals

A little more on the suggestion from Jamie last week that three points for a win (3PW) might have made a difference to the percentage of matches resulting in draws, or in the number of goals scored.

A little more analysis shows that while there has been a slight increase in goals since 1982, there is no evidence to suggest that this can be attributed to the 3PW change.

Football has changed considerably in the past 30 years, and if we compare the 10 years prior to the points change with any period after, we are also including other changes in the game. It’s interesting to see that despite these changes, many numbers have stayed relatively constant.

The starting point for this post was the conversation about why the draw was (almost) never the favourite outcome, and the 2PW to 3PW change is a red herring.

Before the change, the previous 10 seasons saw the combined goal expectancy per game in the top English division at 2.57. Over the past 10 seasons it is 2.65.

One noticeable change though is that while the total expectancy remains much the same, the ratio of home goals versus away goals has changed. 1971 to 1981, the ratio was 1.58 to 0.99 whereas from 2004-13 it was 1.53 to 1.11.

Home goals were regularly scored at a rate greater than 3:2 (i.e. 60%) in Division One, (11 times in the last 21 seasons), but the Premier League has seen this happen just once, in 2009-10.

Not surprising then that home wins have declined from 49.5% to 45.6%, and away wins climbed from 21.8% to 27.1%. That 27% has held solid over the past 20 seasons.

Back to the draw and how it correlates with the goal totals. Looking at the past 42 seasons, in 1973-74, the goal expectancy was at a low of just 2.4, and the draw occurred in a high of 32.25% of matches. In 2011-12, goal expectancy was at a modern record high of 2.81, and the percentage of drawn matches was just 24.47%, but interestingly the correlation is highest during the eleven seasons of Division One following 3PW – for the statistically minded at 0.83.

During the entire Premier League era, there is no significant correlation although in the early days, there was. The years from 1992 to 2005 saw the correlation at 0.71, but then game the outlying season of 2005-06 when just 20.3% of matches were drawn, despite the average goals per game dropping below 2.5 for the first time since 1974. In other words, the complete opposite of what would be expected.

2005-06 was unusual though, in that while the easy wins (by 3 or more goals) were slightly below their long term average, wins by 2 or more were at an all-time (Premier League) high – 146 games.

Thus the number of ‘close’ games was less, but what was remarkable is that while the ratio of one goal matches to drawn matches averages about 1.42, in 2005-06 it was at 2.04 - a true outlier.

I mentioned in my email to XX Draw subscribers the unusual situation this season where the goals per game average has jumped up in Serie A, but dropped in the EPL, yet there are more drawn games in the latter. The reason is that there have been a number of games in Serie A which have seen a high aggregate of goals, and thus less chance of a draw (although there have been three 3-3 draws already) while in England, just one team has scored more than four goals, (Manchester City 7-0 v Norwich City at the weekend) and just that one game has seen more than five goals.

The conclusion is that while the goal expectancy clearly affects the probability of a draw, the correlation over a season may well be masked by either a high number of blow-out wins, or as in 2005-06, by an outlier.