In today’s part of our odds compiling series, we’ll finally take a look at zeroing in on a closer xG numbers for the upcoming Premier League match in GW 34 between Liverpool and Palace using our 5 season data.
In Part V we are going to look at
refining our xGoals calculation building league tables and evaluating teams over/under performance. With games coming thick and fast this past week, I will wait for a lull in fixtures (after GW 32) to write about improving our xGoal calculations using Attack and Defence ratings.
In this part of our tutorial, we’ll be using our RANK function again to automate everything so it updates after every game week. We’ll also be learning about some metrics commonly used in fanalytics.
So far in this series we have looked at a basic way of estimating xGoals, using a Poisson Distribution, and Bookmaker margin. In Part IV we are going to take a look at using previous results as a way of deriving odds.
Article originally posted on 30th March 2017 here
In Part I of the series we looked getting data and some basic formulas. Part II looked at xGoals in more detail and a basic method of finding the expected number of goals scored between Liverpool and Everton. This method is by no means the end point and there is still some improvements to be made, but we’re beginning from scratch and we’ll come back to improving the numbers later on.
In Part I we looked at gathering and cleaning data from football-data.co.uk, and building our first basic table. In this part we’ll discuss xGoals in a bit more detail and run through our first method for calculating odds, using the upcoming game between Liverpool and Everton.
One question I get asked quite regularly is how I calculate my odds for football. There are many different methods I use, and I never rely on just one model, but I thought I’d share here a very simple method of compiling odds for football.
The difficulty with many of the advanced statistical models is that they use more advanced statistics such as x,y shot locations and other data that is not publicly available, or they use programs that may not be easily accessible to the average fan due to lack of coding knowledge – programmes such as R or Python come to mind. The model that I will walk you through here uses data that is publicly available and needs only a basic knowledge of Excel to build.