Putting stats into context is key to their significance

Everton scored and conceded in 16 of their first 19 games of season 2012-13. They scored and conceded in only eight of the last 19.

The same season there were more than 2.5 goals in 15 of Fulham’s first 19 games. There were more than 2.5 goals in nine of the last 19.

That was also true of Manchester United. In 15 of their first 19 games there were over 2.5 goals. In only nine of their last 19 games were there over 2.5 goals.

Those contrasts between the first and second halves of a season were stark, but they were not unusual. There is hardly any correspondence in goals markets between a team’s make-ups in different parts of a season.

Goals stats have what investor and author Michael Mauboussin calls low persistence. What happens in one period tells you little about what will happen in the next.

That might sound discouraging but it should not. What matters is not whether the past record of a thing reveals a lot or a little about what will happen in the future, but that you know whether it does.

Many football bettors seem to believe there is a strong connection between past and future goals make-ups. Bookmakers know there is not, but they have to allow for the views of their customers – at least sometimes and to some extent. In goals markets you might be able to find opportunities to make money by betting against a repeat of a team’s recent past.

I looked at what happened in the first and last 19 games for every Premier League team in each of the last 21 seasons, all those in which teams played 38 games.

In 50 per cent of all games both teams scored. No matter how often a team scored and conceded in the first half of a season the chance of them scoring and conceding in a game in the second half of the season was rarely far from 50 per cent.

The best estimate of how often a team would score and concede in the second half of a season was a calculation in which less than ten per cent of the weight was given to how often they scored and conceded in the first half of the season and more than 90 per cent of the weight was given to the average over any half of a season for all teams.

And the same for games with more than 2.5 goals. Forty-nine per cent of all games featured more than 2.5 goals. The best estimate of how often a team’s games would feature more than 2.5 goals in the second half of a season was a calculation in which less than ten per cent of the weight was given to their number from the first half of the season and more than 90 per cent of the weight was given to the average for any half of a season for all teams.

Newcastle scored and conceded in only four of their first 19 games of season

2005-06. They scored and conceded in nine of the last 19. In only four of Stoke’s first 19 games of season 2012-13 were there more than 2.5 goals. In nine of the last 19 there were more than 2.5 goals.

No stat is useful in betting unless you know what it means – that is to say, unless you know what it signifies about what might happen next.