Can you Use Total Shot Ratio to predict the World Cup winner?

Can you Use Total Shot Ratio to predict the World Cup winner?

By Mark Taylor May 16, 2014

Tweet

Total Shot Ratio has emerged as a tool for quantifying the strength of soccer teams; this article outlines what TSR is, discusses its strengths and weaknesses as well as investigating its value for predicting the World Cup winner?

The emergence of Total Shot Ratio

Predicting the outcome of football matches initially relied exclusively on the only readily available data, namely goals. Once applied to a Poisson approach, the ability to forecast individual goal probabilities for each side enabled a wide variety of associated events to be framed, from correct scores to the timing of goals.

Other sports, notably in North America, had access to a much more varied and extensive collection of events. In the NHL, shooting data outnumbered the more valuable, but much less frequent act of scoring of a goal. This led to the additional use of shots, as well as actual goals, as a way of describing the likely difference in ability between teams.

Better sides tend to produce the more offensive output – hence more shots – and in turn face fewer opposition attempts. Total shot ratio – or TSR for short – finally made the crossover from hockey to soccer once shooting data became more readily available in the latter sport.

What is Total Shot Ratio & how to calculate it?

TSR measures the proportion of shots a side contributes (off or on target including headers) in a single match or an entire season. Therefore, if a contest between Team A and Team B contained 30 shots and team A contributed 20 of the attempts, team A’s TSR would be 20/30 or 0.67, with Team B naturally recording a value of 10/30 or 0.33.

Larger samples more quickly tend towards their true values, as the talent begins to shine through. Total match goals in the major soccer leagues rarely stray from an average of around 2.5 per game, while shots invariably reach well into double figures. Hence, it could be expected that the share of shots, rather than goals could better reflect a side’s true talent.

For TSR to be useful as both a predictive tool as well a descriptive one, it needs to correlate well with results. Teams with good TSR values should also perform well in terms of wins and league points. In addition, historical TSR should predict future TSR reliably, in the same way that a weighted sample of goals scored and allowed can be used to predict future scoring rates.

League points do appear to correlate fairly well to a side’s TSR for that season. Below we’ve plotted league points and team TSR for 10 seasons of English Premier League soccer from 2002-03 to 2011-12 and there does appear to be a correlation between the two. A side with high TSR’s over a season tended to gain more points than those with poorer ones.

Similarly, TSR begins to approach the end of season values relatively quickly. Mid-term TSR values show a very strong relationship to the TSR recorded by a team at the end of the season.  Team TSR from one sequence of games is also reasonably well connected to subsequent batches of games.

So, we should be able to use the TSR recorded by two teams over a prior run of matches as a proxy of their likely respective abilities going into a future match.

To try out TSR as a potential talent rating for individual teams, we recorded the overall TSR recorded by EPL teams in their previous 20 matches and used these figures as the rating each side took into a future game. We then compared the actual results of matches using each team’s TSR as the two predictors and a strong relationship appeared to exist.

This relationship was then used to produce win, lose and draw odds for future matches, examples of which from the perspective of the home side are included below. Games were played in the latter stages of the 2012/13 English Premier League season and we’ve included the odds from that day.

Match

Home Team TSR

Away Team TSR

Predicted Home Win

Quoted Implied Probability

Newcastle vs Fulham

0.505

0.447

0.50

0.48

Tottenham vs Everton

0.650

0.586

0.50

0.46

Man Utd vs Man City

0.540

0.640

0.16

0.43

Arsenal vs Norwich

0.607

0.433

0.77

0.80

Stoke vs Aston Villa

0.416

0.387

0.43

0.48

Chelsea vs Sunderland

0.570

0.422

0.72

0.73

Liverpool vs West Ham

0.621

0.440

0.78

0.74

For most of the games, a TSR based approach produces comparable home win odds to those quoted, indicating their possible use as a tool to frame match odds, with one glaring exception. Manchester United was given a much smaller chance of defeating their cross City rivals than was generally quoted.

To calculate Predicted Home win

We used the TSR over the previous 20 matches for the home and away side and whether or not the match was a home win (or away win) for three seasons.

Therefore the two-predictor variables were home TSR and away TSR and the outcome variable was coded 1 for a home win or 0 for any other result.

That gives you your regression equation, which you can use to input future TSR match ups to give the probability of in this case a home win. That’s the “Predicted home win” column.

To calculate Implied odds

The “Quoted Implied odds” is Pinnacle Sports’ odds for a home win, expressed as a percentage, prior to the game.

Limitations of Total Shot Ratio

A side can over-perform against their fundamental shooting statistics simply through random variation, but United have often performed in such a manner when managed by Sir Alex Ferguson. Therefore, whilst a long lucky streak is possible, if improbable, a more likely explanation is that TSR fails to capture everything about a side’s talent.

A further clue concerning the limitations of TSR can be seen in the under rating of Stoke at home to Villa. Stoke – at the time were managed by Tony Pulis – who prospered in the English Premier League by creating fewer, high quality chances – often from set plays and long throws – were happy to allow their opposition to have a higher numbers of low quality attempts, often from distance.

TSR does not discriminate between speculative longshots and close range chances, or even shots widely off target, but reality and location based shot models do. Therefore, some teams may use repeatable tactical wrinkles that are not picked up by a method based around shot volume and may be consistently over or under rated because of this.

Few teams play in such an extreme defensive shell as Stoke formerly employed and so the majority of sides will see their TSR more accurately reflect their talent base and single abnormal, luck driven seasons will likely be the exception. Therefore, for those looking to a viable alternative to using goals, TSR may fit the bill, as long as they allow for a tactical maverick operating among a more mainstream approach.

Using Total Shot Ratio to predict the World Cup winner

A shot based appraisal may be an attractive addition to evaluating the World Cup in Brazil. Although harder to find, shot data for international games is available and two of the three highest rated TSR sides from the European qualification groups, Spain 7.180* and Germany 6.410* are also quoted as the two most likely European sides in the overall tournament betting.

England 27.000* are broadly mid-table in both the European betting pecking order and TSR.

Russia 61.000* and Switzerland 81.000* are amongst the least likely to lift the World Cup and they also recorded some of the lowest TSR from the qualifying sides, although both teams are still comfortably above 0.50.

However, overall the correlation between TSR recorded by European teams in qualifying and their perceived chance of success in Brazil isn’t particularly strong.

Portugal has an excellent TSR from qualifying, but is only mid-table of the quoted Europeans. Bosnia also has a fairly impressive TSR, but is lowly rated by the odds makers. At the other extreme, Belgium and especially Italy, are fancied to do better than their TSR might indicate.

Shot counts do vary between different reporting sources and TSR’s from qualifying for Italy range from just above 0.5 from one source to 0.55 from another, highlighting a current potential issue with this nascent statistic.

Additionally, a deeper look at the shooting tendencies of the Italian team may indicate a tactical adeptness, rather than a side riding a slightly lucky tide with an unimpressive TSR. So dismiss the Azzurri at your peril.

Of the South American qualifiers, Argentina 5.540* and Uruguay 22.000* are closely matched as the best TSR qualifiers, but the former is much more fancied in the outright betting. Possibly reflecting the difficulty associated with qualifying from Group D. Columbia and Chile’s TSR keeps step with their current respective ranking in the outright odds. Brazil is omitted, as they’ve existed on a diet of friendlies.

In the limited data we’ve found for Concacaf qualifying sides, Mexico hold top spot in both TSR and expectation of World Cup success, but the USA, the spiritual home of the metric, records a poor TSR score compared to both the lesser fancied, Honduras and Costa Rica. Although it should again be stressed that data becomes scarce in the other qualifying confederations, with numerous matches devoid of shot data.

These discrepancies between bookmaker expectation and underlying TSR may provide exploitable edges along the lines of what may be possible by forming odds by using individual team TSR over a previous timeframe as demonstrated in the Premiership examples, but caution is advised. In game effects, such as red cards and time spent leading, drawing or losing can skew figures, as can strength of schedule issues.

Currently, TSR provides food for thought and is a possible opportunity in waiting.

Mark Taylor is a freelance soccer and NFL writer who, along with producing expert content for Pinnacle Sports, also runs his own soccer analytics blog, the Power of Goals.

*Odds subject to change

If you have feedback, comments or questions regarding this article, please email the author or send us a tweet on Twitter.