If you’re tapped into the conversation surrounding the Premier League right now, then you’ve probably seen a lot of pundits speculating on how many points clubs like Liverpool will end the season with. They arrive at these figures using two special formulas, one for expected goals (xG), which is then used to calculate expected points (xP). Here’s how, in a little more detail.
Why These Predictive Metrics Exist
Before we go into the technical details, it’ll help to explain why these metrics exist. xG is about measuring the probability that a goal-scoring opportunity will succeed, on an individual or team basis. xP uses the xG of players/teams in a match-up, plays them against one another, and predicts the final score.
Of course, predictions are just that. They can and often do miss the mark, and that’s part of the fun as pundits give competing analyses. Knowing how the metrics work won’t give you the answer. You could know how a random number generator (RNG) works, but it’d still spit out an unpredictable number. A lot of modern entertainment relies on that unpredictability, like online casino slots with RNG software. Those randomness algorithms were inspired by real-life action, where anything can happen on the reels or on the pitch, and the final scores aren’t known until the game is over.
So, take all predictions with a grain of salt. There’s no shortage of dramatic upsets or stunning comebacks in football that leave number crunchers scratching their heads. Besides the betting market, these match stats also help pundits understand how certain metrics influence fixture outcomes. Put simply, it helps the talking heads pay attention to things that actually change the match, so their off-the-cuff commentary and speculation can get sharper.
Explaining xG
As we said, xG is the probability of a successful goal. A more natural term for this would be quality – you don’t need to be a data scientist to see a good shot versus a desperate, long shot. xG puts that into numbers, with 0 being poor and 1 being a certain goal. A close open goal would rank somewhere like a 0.8 or 0.9, an easy shot. A shot from the halfway line would be 0.1 to 0.4, depending on other factors.
That’s because a lot goes into these calculations, fuelled by machine learning AI. Distance, angle, play pattern, kick vs header, all of it factors into that final xG score. That number represents how many times the goal would succeed out of ten, so a 0.8 should score eight times out of ten. Of course, a prediction is only as useful as the data used to make it. Not every source has access to this rigorous match data.
Explaining xP
xP, or xPts in some places, is what you get when you total the xG of two teams and then play them against one another. This is more complicated than predicting goal scores and involves simulation software that can ‘play’ thousands of matches against one another using the xG numbers. Think of it like virtual football, only run privately for research purposes.
The winner of each match and their scores are all tallied, with tens or even hundreds of thousands of match results. Let’s say a simulation of Liverpool vs Arsenal runs 100,000 times – Liverpool wins 64,000 times, Arsenal 28,000, with 8,000 draws. Since we’re focusing on Liverpool, the xP formula will look like this:
xP = (Win Probability x 3) + (Draw Probability x 1) + (Loss Probability x 0)
Probabilities are expressed in decimals, so 64,000 becomes 0.64, 28,000 becomes 0.28 and draws become 0.08. If we work it out…
Liverpool xP = (0.64 x 3) + (0.28 x 1) + (0.08 x 0)
Liverpool xP = 1.92 + 0.28 + 0
Liverpool xP = 2.2
Remember that this is for one fixture. That xP figure is then used in match predictions, and for whole-season predictions, the number crunchers will tally each one to arrive at league table predictions. If you couldn’t tell, a lot of time and resources go into calculating these predictions. It’s fascinating to see where they come from, but if you’re a normal fan, it’s probably best to leave it to the experts and enjoy the match as it’s played.
