Re-Introducing Expected Points – (xP/xPoints)

Written By; Jamie Kilday

For anyone who is into sports analytics there was usually an idea or question that dragged them in. When I first starting reading about xG the question ‘How can I use xG to determine the expected outcome of a game?’ was the question that took me from having a passing interest in analytics, into actually writing about it myself. The answer to the question is xP or Expected Points.

In much the same way that xG isn’t a new thing, its just having a more broader appeal every season, expected points (xP) is nothing new. Much like xG there are different models to determining xP but lets first lets acknowledge the elephant in the room, why do we need xP?

There will be many people who think that xG is just a statistic too far, you don’t have to look too hard to find many ‘There is only one stat that matters’ tweets on twitter, and xP is not going to convince that crowd any further so this is your chance to stop reading if xStats are not your thing.

Stilll reading?


xP is useful for, amongst other things, determining how lucky a club are. You may have noticed that with our shot maps that we post every Thursday we post probability charts. Using the xG data we can break that into the odds of a specific outcome of the game by running it through a Monte Carlo Simulation. Sports Data Analyst Danny Page explains in a more eloquent way than I ever could how xG can be used to determine potential outcomes and is well worth a read HERE

A very simple breakdown of how xPoints are determined is, once the win ratio of the game is determined then this can be used to establish how many points a team would have won. If the outcomes are, Team A Win 40%, Draw, 30%, Team B Win 30% then points would be;

Total Points = [3 x (Percentage Outcome of a win)] + [1 x (Percentage Outcome of a Draw)]

Team A = (3 x 0.4) + (1 x 0.3) = 1.5

Team B = (3 x 0.3) + (1 x 0.3) = 1.2

In reality both teams cannot receive more than one point and points cannot be decimalised but xP is not about determining who should have got the 3 points or if a point each is fair and attributing the points that way.

As I mention earlier there are many ways of calculating xP. You could say, a Team A win is statistically the most likely outcome so their xP is 3 but for me you are removing the performance based element of xP.

The benefit of xP is you can determine whether you are right to feel that you team should be doing better or if indeed they are doing much better than their performances would indicate.

I’ll be posting these every week but seeing as we are early in the season here’s a look at how last seasons xP table compared to actual points after 33 games created using @TheGersReport xG data, which I was using prior to the formation of Modern Fitba.

n.b. The split causes some headaches as you can see as some sides have an xP worthy of finishing in the top 6 but just missed out.

AnalyticsModern Fitba