close
Oct 18, 2017
Oct 18, 2017

Why bettors should look at expected goals in more detail

The birth of the big chance and expected goals

Why chance quality and chance frequency matters

Is one big chance better than multiple small chances?

Why bettors should look at expected goals in more detail

There is no doubting that advanced analytics improve our understanding of sport. Whether a team analyses data to improve performance or bettors do it to make profits, it is important that data is collected in the right way and analysed properly. Read on to learn more about looking at data in more detail in soccer.

Understanding expected goals

Expected goals is the most visible manifestation in the emergence of analytics, having gone from an abstract concept that tries to quantify chance quality, all the way to a mainstream spot within soccer punditry.

As with all advanced metrics, expected goals is a mere tool that assists in better describing the likelihood that past events occurred and predicting the likely course of a side’s future performance. 

It’s important to understand that this metric is an output from a model and therefore expected goals from one source are unlikely to tally exactly with those derived from a model created from slightly different parameters or timescales.

All expected goals models incorporate three major parameters that most strongly influence the likelihood that a chance will turn into a goal.

The spread of likelihood from the highly likely to the one in 30 or more for long-range pot shots is at the heart of the expected goals revolution.

These are the distance of the attempt from goal (most easily described by the horizontal distance from the centre of the goal and the vertical distance to the goal line, usually referred to as the x,y coordinates of the attempt) and also the type of attempt made, be it a shot or header.

It is to be hoped that models do substantially agree on the general likelihood of a goal being scored while being flexible enough to project a side’s future prospects and help bettors to predict future performance.

The birth of the big chance and expected goals

Perhaps the most scrutinised chance in soccer is the so-called big chance, where a goal is more likely to be scored than not.

For those with long enough memories, the concept of a big chance entered the popular folklore of English soccer in the dying stages of the 1982-83 FA Cup Final between Manchester United and underdogs Brighton.

“….and Smith must score” refers to Gordon Smith’s late one on one chance with Manchester United’s Gary Bailey. Bailey made the save and Brighton lost the subsequent replay 4-0. 

The implied certainty in the voice of the commentator produces a dramatic and historic footnote in the fortunes of Brighton and Gordon Smith, but when viewed through the detached scrutiny of expected goals, the opportunity, good though it undoubtedly was, would usually be converted only 60% of the time.

Equally, a spectacular shot from distance is often guaranteed to secure a place on the highlights reel, particularly if it requires a save or narrowly misses the target.

However, the frequent omission of more mundane efforts from a similar distance that bobble harmlessly wide of the goal or are blocked well short of their intended target may implant an elevated expectation of success for such speculative efforts in the minds of the casual, highlights viewer.

Why chance quality and chance frequency matters

The spread of likelihood from the highly likely to the one in 30 or more for long-range pot shots is at the heart of the expected goals revolution. More so inside the game itself, where an appreciation of the risks and rewards by shunning speculative efforts and attempting to craft better opportunities can drive tactical aims.

It is also important when bettors evaluate teams.

Expected goals has progressed this evaluation by quantifying that not all shots are created equally and led to the idea that a side with a higher expected goals tally than their opponents is more likely to outscore them.

How each individual expected goal opportunity is distributed between better quality chances and lower quality ones then extends the concept to another level.

All expected goals models incorporate three major parameters that most strongly influence the likelihood that a chance will turn into a goal.

I first introduced this idea that different distributions of expected goals chances may have different payoffs for teams, even if the cumulative expected goals totals were identical, in a 2014 blog post

Soccer is a low scoring sport and sides that attempt to create fewer, but higher quality chances are increasing the likelihood of scoring at least once, compared to a side that chooses to spread a similar expected goals total over a larger number of lower quality chances.

The latter trade an improved chance that they score at least once for the rare occasions when they score a larger number of goals. 

Over the recent Premier League season, a team that scores once in a match has taken an average of 1.13 points from those home and away games; an adequate level of points to survive in the top flight year after year.

So far we’ve worked on intuition, but to expand the usefulness of expected goals and to quantify the effects we have described it is useful to have a methodology.

Is one big chance better than multiple small chances?

Simulations, be it in excel or R, are powerful applications that utilise often dry and abstract concepts, such as expected goals to produce a multitude of “what ifs” ranging from individual events in an FA Cup Final, to extracting the range and likelihood of possible finishing positions based on the expected goals created and allowed by all 20 sides in the most recent Premier League campaign. 

I’ll combine the two in repeating the simulation from my 2014 blog.

We’ll give each side 1.2 expected goals. One, we’ll call “Smith United” creates two chances each having a 60% chance of being scored based on historical precedence and their opponents “Pot Shot FC” muster an impressive 20 long-range efforts, each with a 6% chance of scoring. Overall the expected goals created by each team is 1.2 xG

Occasionally, it ends badly for Smith United - there are a couple of 7-1 defeats that appear in 10,000 simulations of this particular game. But overall the strategy of creating fewer, higher quality chances pays off better than the one based on more, but lower quality chances.

Smith United wins 375 games, draws 315 and loses just 310 per 1,000. 

This highlights how raw expected goals totals hide the variance that may appear in outcomes based on the differing distributions of big and smaller chances. 

It is a mathematical validation of the attempts of some teams to utilise the talent at their disposal to create higher quality opportunities for their strikers, while largely shunning attempts from distance.

As with all advanced metrics, expected goals is a mere tool that assists in better describing the likelihood that past events occurred

Digging a bit deeper into a team’s chance quality, particularly now many websites are beginning to list the numbers of big chances a side creates, therefore can give an added nuance to xG numbers. 

Arsenal, both Manchester clubs and Liverpool currently head the table for creating the most number of big chances, but other less fashionable teams, such as Stoke when Delap’s long throws played an integral part in the chance creation, have also placed big chances at the forefront of their process.

It could be argued that some teams have enjoyed success when prioritising a high frequency of low-quality chances - Roberto Martinez's Wigan side being one such example. However, one thing that's for sure is that teams who continually concede high numbers of big chances will deservedly struggle to compete in any soccer league.

When scratching beneath the surface of xG, there’s still a lot to see.

Betting Resources - Empowering your betting

Pinnacle’s Betting Resources is one of the most comprehensive collections of expert betting advice anywhere online. Catering to all experience levels our aim is simply to empower bettors to become more knowledgeable.