Oct 8, 2014
Oct 8, 2014

How high quality chances and game states relate to expected goals

How high quality chances and game states relate to expected goals
This week 7 Premier League review explains how high quality chances and game states relate to expected goal totals in soccer and why understanding this is important for live bettors.

The two pace setters in the Premier League - Chelsea and Manchester City - gained three points during week seven after 2-0 wins. Chelsea defeated Arsenal at home on Sunday afternoon and City travelled to Aston Villa for the late kick-off on Saturday.

Although the score line was the same on both occasions, the manner in which the two wins were achieved was very different, especially from the basis of chances created and allowed.

Chelsea were outshot 2:1 by their opponents, Arsenal, but still managed to record a higher expected goals total from their five efforts by virtue of creating better quality opportunities, most notably Hazard’s penalty kick that opened the scoring after 27 minutes.

By taking the lead, Chelsea were also able to adopt a more typically Mourinho approach, restricting Arsenal to attempts from distance with lots of defensive pressure, while relying on counter-attacks to potentially increase their advantage.

In stark contrast, Manchester City were frustrated for long periods of their game at Villa Park, as the home team adopted a spoiling tactic by absorbing lots of pressure by getting bodies behind the ball to block shots, combined with a very occasional counter attack.

Ten of City’s 13 first half shots were blocked by Villa players and a shot count of 27-6 in City’s favour paints a picture of quantity, but little from quality positions.

Likely match outcomes in Week Seven of the EPL based on 10,000 shooting simulations

Home TeamHome Win %Draw %Away Win %Away TeamActual ScoreMost Likely Score
Aston Villa 8 12 80 Man City 0-2 0-2
Chelsea 57 23 20 Arsenal 2-0 1-0
Hull 53 21 26 Crystal Palace 2-0 1-0
Leicester 44 28 28 Burnley 2-2 1-1
Liverpool 62 22 16 WBA 2-1 1-0
Man Utd 35 28 37 Everton 2-1 1-1
Sunderland 36 33 31 Stoke 3-1 1-1
Swansea 36 25 39 Newcastle 2-2 1-1
Tottenham 30 30 40 Southampton 1-0 0-1
West Ham 70 18 12r QPR 2-0 2-0

The concept of game states in soccer

The different ways in which these 2-0 victories were achieved owe much to the manner in which the game states developed in each match. The concept of game state is still evolving in soccer and is often used interchangeably with the current score line, but this is an over simplification, especially when the game is currently tied.

Game state attempts to define how satisfied each team is with the current state of the match and therefore how their approach to the game might change as it progresses. It encompasses both the current score, the amount of time remaining in the game, but also the realistic pre game expectation of each team, which itself defines the quality of player available to each manager.

A trailing team will inevitably be prepared to take more risks as time ticks on. Often this is countered by their opponents becoming more defensively minded. And the net result is that the trailing team can on occasion record more frequent, but better defended goal attempts, often from distance and increasingly attempted by less attackingly able defenders.

In short, shot numbers may rise, but expected goal levels per attempt often fall.

Similarly, their opponents may create fewer opportunities, but these may be of higher quality on the counter attack, Diego Costa’s second goal for Chelsea against Arsenal being a prime example.

The match at Villa Park remained goalless until the 82nd minute. But the game state had deteriorated gradually for Manchester City, while improving for their hosts.

At the start of the match City were expected to have a 63% chance of winning and a 23% of securing a draw.  Overall their 63% chance of three points and a 23% chance of one point gave City an expected average of around 2.1 league points from this encounter.

As the game entered the 82nd minute still scoreless, City's probability of winning the game had dropped to 25% and a draw was the most likely outcome (65%). Their expected league points from the game had now fallen to 1.4.

In turn Villa’s expected league points increased to 0.9 of a point, having been just 0.66 from their respective win and draw percentages at kick off.

This gradually evolving game state meant different things to each team and had partly dictated how the City/Villa match was played out. Villa were happy to attempt to secure a point against much stronger opponents, even at home and City continued to push for a winner as time and a scoreless game gradually eroded their points expectation.

The risk reward balance of going for a win apparently outweighed the possibility of losing a point and presenting two extra ones to a side in Villa that was unlikely to be challenging City in May. And this partly drove City to amass 27 goal attempts, many of which yielded a low goal expectation.

By contrast, Chelsea in addition to enjoying a fairly comfortable game state in arriving at the same ultimate score line, also created better individual opportunities compared to those of their title rivals, City.

Penalty kicks are converted around 80% of the time, but they are relatively rare events. However, Diego Costa’s goal also came from a chance which had a higher individual goal expectation than many of the 27 chances created by City earlier in the weekend.

Therefore, bettors should understand that although the raw goal expectation for a team is a useful figure, it does hide the granular manner in which the goal expectation is spread out among the individual chances. And a team that consistently creates high quality chances will do better than one that spreads the same goal expectation over more chances.

In short, creating high quality chances is an important secondary factor to consider when evaluating a team’s overall goal expectation.

Overall goal expectation figures alone are currently barely able to split Chelsea and Manchester City in terms of schedule adjusted performance so far this season. But in terms of creating high quality chances, the Blues are superior.

1740 chances have been created to date across the EPL and when arranged in descending order of expected goals, Chelsea have created 13 of the EPL’s 100 best goal scoring opportunities, compared to none as yet from City.

So while advanced football statistics are in their infancy, their usefulness is still unclear. The drive to be accessible and present complex interactions in a single figure may wash out from the wider picture more granular detail, such as game state effects and goal expectation distributions.

It is well to appreciate that numbers, such as those presented below can omit tactical wrinkles that are often essential to an individual team’s quality, along with the type of chances a side consistently creates or faces and can also fail to account for more obvious squad strengthening or injuries.

After all bettors should remember they are betting tools to be used to evaluate team quality in conjunction with other ingredients, rather than complete solutions.

Team achievement based on chances created & faced (data adjusted for opponents strength)

TeamAttacking rating (higher = better)Defensive rating (lower = better)Overall rating
Chelsea 1.56 0.80 2.01
Man City 1.44 0.69 1.97
Liverpool 1.42 0.74 1.90
Arsenal 1.26 0.62 1.87
Southampton 1.24 0.76 1.77
West Ham 1.17 0.77 1.71
Tottenham 1.08 0.68 1.69
Newcastle 1.21 1.01 1.62
Everton 1.05 1.07 1.46
Stoke 0.87 0.87 1.43
WBA 0.85 0.87 1.40
Aston Villa 0.89 0.94 1.39
Sunderland 0.84 1.04 1.32
QPR 1.00 1.33 1.32
Leicester 0.85 1.20 1.30
Man Utd 0.79 1.14 1.28
Crystal Palace 0.80 1.21 1.24
Swansea 0.86 1.27 1.24
Burnley 0.72 1.08 1.20
Hull 0.73 1.28 1.17

The EPL story so far

With these caveats in place, the final table above gives a snapshot of the core defensive and attacking attributes of each team adjusted for opponent strength after week seven.

As usual, expected goals are used to reward those sides that may be failing to get the rub of the green and to highlight high flyers, which have perhaps profited from some good fortune – something all bettors should be aware of.

As an example, Arsenal have produced chances of quality and quantity that would see them create an average of 1.60 expected goals per game, and concede opportunities consistent with a side allowing 0.78 expected goals per game.

They have done this against opponents which have typically allowed the chance equivalent of 1.27 expected goals and created 1.26 expected goals per game. So Arsenal have played against slightly below average defences and slightly above average attacks.

If we combine these four figures, Arsenal’s attack is creating expected goals totals that are 126% of average typically allowed by their opponents to date and allowing expected goals totals that are 62% of average typically created by their opponents.

Therefore Arsenal are well above average, both in attack and defence. These two chance based figures are good enough to place Arsenal as this season’s fourth best overall performing team and this is denoted by the overall rating figure in the table above.

To make the final rating easier to appreciate, it is expressed as the number of league points each side would gain, on average, against a mythical average team at a neutral venue, if they repeatedly performed to the level of their chance based achievements to date.

Figures such as these may expose the amount of random variation that can contribute to the actual points record of a team and at least lead to questions being asked about how talented, as opposed to lucky or unlucky, a side really has been recently.

Swansea, for example, have a tradition of using possession as a defensive tactic to protect a lead or a point, so shot based analysis alone may be unduly harsh in appraising their qualities. But their expected goal differential should also give some cause to doubt their current league position as we enter another international break.

Since they were noted as potential overachievers with less impressive underlying stats following Gary Monk’s manager of the month award in August, the Swans have taken just two points from a possible 12.

Similarly, Newcastle continue to possess expected goal figures of a mid-table team, following a point from Swansea on Saturday in an eagerly awaited contest between the two teams currently with the biggest disconnect between actual points and their underlying chance creation stats.

Manchester United’s lowly standing owes much to an abundance of usually rare events, such as red cards and penalties and a starting eleven that is now unrecognisable from early August. The honour of playing the league’s easiest schedule so far also still lies with Manchester United, while Villa have played teams with the most impressive combination of attacking and defensive expected goals tallies.

West Ham, a side managed by one of the early statistical innovators, Sam Allardyce, appear well suited to a spot in the top half, even if just three points separate 16th from 4th in the actual fledgling league table. Their shooting statistics place them well above the majority of similarly rated teams from Everton downwards, any of which might be vulnerable to a run of poor, bad luck driven results.

At the top of both tables, Chelsea’s rating even against the Premier League's fourth easiest schedule has them as the league’s best side, a view endorsed by some of their opponents.


Click here for the latest Premier League odds. 

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.