Let the numbers do the talking

Lawro – Mark Lawrenson – former player, football pundit and expert took a bit of a battering today, with a website castigating his weekly predictions of upcoming football fixtures:

The table based on Mark Lawrenson’s predictions is actually ridiculous

Reading the article, it goes on to rubbish Mr Lawrenson’s weekley predictions for Premier League fixtures. Joe.co.uk has turned his predictions into a league table and the Give Me Sport article goes on to say

we’ve discovered how the Premier League table would look going into the final game week if all of Lawrenson’s predictions had come true.

To say it’s ridiculous would be an understatement.

Firstly, his former club Liverpool sit top on 89 points having not lost a single game. Not biased at all…

Also, Lawrenson clearly still rates Leicester with them sitting seventh with 71 points! After the Foxes, there’s then a bizarre 22-point gap to Everton in eighth.

Despite Sunderland being rock-bottom with 24 points in real-life, Lawrenson’s predictions has them in the dizzy heights of 14th on 34 points.

Later on the article describes Lawrenson’s predictions as


Harsh words, so do the numbers back up such criticism?

In a word, no.

Using Spearman’s Rank Correlation Coefficient we actually discover that Mark Lawrenson has done rather well.

Spearman’s Rank is a great tool to compare how closely two different variables compare. It is perfect for a case like this where we want to compare a predicted league rank with the actual league rank.

It is based on the difference between the square of the two different rankings.

For example, Lawrenson predictions would have Bournemouth in 13th place, in reality they are in 10th place, so the difference is 3, the square of the difference is 9 (we square the values so we don’t have to worry about negative numbers).  This is the formula used to calculate Spearman’s Rank Correlation Coefficient:

Spearman's rank correlation coefficient

Formula for Spearman’s Rank Correlation Coefficient

where d is the difference and n is the number of pairs of data (in this case 20 as there are 20 teams we’ve compared.)

The value will always be between -1 and 1, where 1 tells us there is perfect agreement in ranking.

So how did he do?

Pretty well – he ended up with a coefficient of 0.890 (to 3 decimal places) which suggests that there is a strong positive correlation between his predictions and the actual league positions. In other words, the boy’s done good.

So ignore the sensationalist headline of the Give Me Sport piece and let the numbers do the talking. Mark Lawrenson has been proven, mathematically, to be a pretty good pundit and predictor.

This entry was posted in Handling Data and tagged , , . Bookmark the permalink. Post a comment or leave a trackback: Trackback URL.

One Comment

  1. Sumit Rahman
    Posted 10/07/2018 at 12:38 pm | Permalink

    I agree this is an excellent example for using Spearman’s rank correlation coefficient – and no doubt easily explored in a classroom. How does the correlation look when we look at points though instead of rank? I suspect it’ll look worse.

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>