Shots on target data (SOT) is now easily accessed, so it can be of interest when analyzing how good a team is.
One way of looking at the strengths and weaknesses of a team is to give a value to the attack and the defense respecitvely. This can be done using SOT analysis.
There are obviously other variables, such as motivation and injury, but SOT provides a very good starting point. In simple terms, it allows us to give a value to the attack and the defense of any team and compare that to the value of their opposition. This allows us to anticipate some features of their match-up.
For example, we might want to find a value bet for a game in which Man Utd are playing at home to Man City. Firstly, we’d probably like to work out the probability of either team winning.
We need to find the value of the defense and attack for each team and work out the expectancy of the number of goals that Man Utd and Man City will score.
Man Utd’s goal to SOT ratio is 0.45, as you can see in the table below. Their average number of SOT per game is 5.8.
If you were to multiply 0.45 by 5.8 then you have a value of 2.61 goals. Sadly, though, it is not that simple.
In this hypothetical example:
- Rooney is injured
- Man United are playing in the European Cup Final the week after
- Manchester United have already won the league
- Kompany is injured (the strength of the opposition’s defense is subject to a number of other variables, too)
After taking all the above into consideration, I came up with the following hypothetical values:
- Expected shots on target for Man Utd: 4.33; Expected shots on target for Man City: 6.14
- Expected goal to SOT ratio for Man Utd: 0.27; Man City: 0.23
If we multiply 0.27 by 4.33 for Man Utd, we see in this hypothetical example that they are expected to score 1.16 goals
Man City in this hypothetical example are expected to score 0.23 times 6.14, so 1.412 goals.
As Man Utd, on paper, should be less motivated for this game, we would look to have a bet on Manchester City, if we can get value.
Looking in more detail at the table below we see that Sunderland have the best Goal to SOT ratio at 0.46. Steven Fletcher has scored 5 goals and had 8 shots on target, so his goal to SOT average is an impressive 0.625.
By comparison, Robin van Persie has had 16 shots on target and has scored 8 goals, so he boasts a 0.5 average, though in a much more creative team.
Carlos Tevez appears in the top 10 in terms of shots, with 34. But if we look in detail, he has 14 shots on target, and just 4 goals. This means that Tevez has a goals to SOT average of just 0.28.
QPR have the wost goal to SOT ratio in the Premiership and Norwich the second worst, with 0.18 and 0.2 respectively.
QPR are certainly shooting too much outside the box. Meanwhile, Grant Holt for Norwich has had 7 shots on target and scored 3 goals, so he has a ratio of 0.42. The problem is that there are no creative players around Holt to help him out, with only Hoolahan contributing more then 100 completed Final Third Passes.
Another interesting insight pertains to Wigan. Before you back them to be relegated on the basis of this data, last season Wigan had the worst goal to SOT data for most of the season, only to escape the bottom three in dramatic style.
The average goal to SOT ratio in the Premiership currently stands at 0.31. You can identify during a game if a team are having an average game by checking the shots on target.
With the millions that Man City have spent on forwards, and a collective ratio of 0.28, below the league average, we might expect more bang for buck than the stats at present indicate.
| TEAM | GOALS | TOTAL SHOTS | SHOTS ON TARGET | GOALS/SHOTS RATIO |
|---|---|---|---|---|
| Arsenal | 18 | 179 | 56 | 0.32 |
| Aston Villa | 10 | 127 | 36 | 0.27 |
| Chelsea | 23 | 162 | 60 | 0.38 |
| Fulham | 24 | 167 | 71 | 0.33 |
| Liverpool | 14 | 196 | 47 | 0.29 |
| Man City | 20 | 204 | 70 | 0.28 |
| Man United | 29 | 175 | 64 | 0.45 |
| Newcastle | 12 | 138 | 44 | 0.27 |
| Norwich | 8 | 144 | 39 | 0.2 |
| QPR | 8 | 149 | 44 | 0.18 |
| Reading | 12 | 110 | 35 | 0.34 |
| Southampton | 15 | 144 | 42 | 0.35 |
| Stoke | 9 | 117 | 36 | 0.25 |
| Sunderland | 7 | 87 | 15 | 0.46 |
| Swansea | 16 | 150 | 53 | 0.3 |
| Spurs | 18 | 185 | 62 | 0.29 |
| WBA | 17 | 146 | 49 | 0.34 |
| West Ham | 14 | 141 | 41 | 0.34 |
| Wigan | 12 | 136 | 43 | 0.27 |
| Everton | 21 | 218 | 69 | 0.3 |
- Excellent
- Informative
- Awesome
- Good Read
- ok
About the Author
Categories: Arsenal (NN), Aston Villa, Betting Tips, Chelsea, EPL Index Featured Article, Everton, Fulham, Liverpool, Manchester City, Manchester Utd, Newcastle Utd, Norwich City, QPR, Reading, Southampton, Stoke City, Sunderland, Swansea City, Tottenham Hotspur, West Bromwich Albion, West Ham United, Wigan
Tags: EPL, epl opta stats, EPL Stats, Manchester City, Manchester Utd, Norwich City, Opta Stats, premier league, Premier League Stats, QPR, Wigan
This article has had 2,703 Views
You could also look at shots per game/possession per game to see how well teams utilize possession but bear in mind that a team like Stoke are not interested in ball retention and that possession can be used as a defensive tool to slow a game down. Passing backwards and sideways is not an indicator that a team are looking to score.
Because we now have all the data available we can see trends quickly.
Could you direct me to a dataset (preferably SAS compatible) that I could use for my own studies of this sort of data? Thanks1
Man City 204 Shots and 11 games so average shots per game is 18.5(204/11)Possession average 59.18%
59.18/18.5= 3.19= 1 shot per 3.19% of poss.
That took me one minute to work out thanks to the OPTA data.
liverpool 196 Shots and 11 games 196/11=17.8
Poss average 57.51% 57.51/17.8= 1 shot per 3.23% of poss
QPR 149 shots and 11 games 149/11=13.5
Poss. average 50.13% so 50.13/13.5=3.71% so not as effective with their poss
My favourite team to analyse. Stoke 117 shots /11 games 10.6
Poss average is 41.03%/10.6= 3.87% . Stoke are not interested in ball retention.
Off to write my next blog on ball retention.
A very interesting and informative piece. If only there was a way to analyse clear cut chances rather than shots on target…
Thank you David. You can Find CCC – clear cut chances on Opta.
A quick story on a quiet night. Chap behind me at work is talking about Aston Villa and I hear him discussing the data. I said are you reading a blog and he said Yes. I said did a chap called Jonny Grossmark write that? He said. i replies ” I am Jonny Grossmark”
Key point this is the first time we have ever spoken.
lol
*YES …
26% ok? David was that you pressing the Ok button 20 times?
Jonny, good piece. Any conversations about shots, stats and metrics are useful. Glad you chose shots on target rather than their less useful, uglier sister: total shots.
Could do with talking about regression, though. Man City will rise, united will fall in scoring% unless you think them the best PL team in terms of scoring in quite sometime.
Have a gander through some of the work I did in the offseason.http://www.bitterandblue.com/2011-12-numbers-review
Lots on shots on target, SoT +/-, SoT expectancy for and against, like this http://www.bitterandblue.com/2012/7/20/3170781/2011-12-epl-season-review-possession-part-3c-expected-shots-on-target
Hello shudderothink.
I will certainly have a look at your work as it is a great time for football bloggers to share their knowledge.
I agree that total shots is the uglier sister(very well put)
“reverting to the mean” – Yes, expectation that Man UTD sot will drop….
Interestingly expectation that one of the relegation candidates will go on a run and pull away(just no idea which one at this stage).
Kind regards.
Out of interest if you voted ok I am wondering what you are wanting from a post on SOT.
Certainly an in vogue subject with the predictive model variables added for good measure. I thought this was a very solid article.
Was that meant fro me, Jonny? I don’t know I can vote!!
If anyone is looking for a shot on target table for every week then I am not sure it exists. Should do really.
Shuddertothink, i have just read your blog on importance of first goal. Excellent read. My interpretation is coming out tomorrow, Have you got more blogs in the pipeline?
You mentioned sample size and again that is a myth because historically there are very strong trends for years and years re 1-0 ht ie +1/-1 leverage ht re not and 0-1 Leverage.
Seems like people get bullied into saying sample size is not big enough. A football predictive model will not look at a teams ability 10 years ago when looking at a game at the weekend.
Your article is excellent.
“Man Utd took the lead in 31 of 38 EPL games. It’s a great number and was powered by the EPL’s best away number”
Not happening this season thus far,
I am looking at the time of the first goal as well as who scores first and concedes first. Could be out tomorrow. Will be interesting to compare to last season. Blackburn went down as they did not have the “fight back” which i discussed in another blog. They could not get enough point when conceding first. This season Wigan have taken 0 points when conceding first thus far. All my blogs add up to expectation of a goal being dependent on the current score. The time of the first goal can have an effect on the expectation of further goals. Losing -1 goal at ht away and it is nearly impossible to win ft unless you are Man utd or playing away at a weaker team.
Losing -1 goal at Ht at home is still a tall order.
Maybe Wigan could still go down?????
Just comparing your data from last season to this season thus far suggests that Southampton and Reading are going down and Wigan are in trouble. You will see what I mean when the blog comes out re the tables.
My humble advice is not to think the data is not big enough. For me one season is a lifetime.
Bolton 75.00 25.00 0.00
Blackburn 28.50 57.10 14.30
Wolves 0.00 20.00 80.00
from shuddertothing .. the three teams who went down had the above fight back rate when conceding first. Wolves won 0% and blackburn not much better . If you look at this season. Wigan and reading and southampton and qpr fit the bill. I think southampton are down as well as Reading with qpr probably in the third spot.
Interestingly the three teams that conceded the most shots were Bolton, blackburn, wolves. Sometimes people are looking for the holy grail in terms of predictive modelling when the data is there. Other times “the sample size is too small is branded”.
Just to add to the shot on target debate. The most powerful tool when building a predictive football model is shot on target data. I built a predictive model based on that last season and it has been very powerful at predicting the expected shots on target in a game.
This was Arsenal v Fulham
0.39*6.1 Arsenal so expected shots on target =6 and 6 happened. Expected goal to shot on goal 0.39 and it was 0.5 so higher
Fulham 0.37*4.5 Fulham had 5 shots on target and 0.6 goal to shot on target.
Looking at the % of 0k on this post it looks like the article has not gone done well.
Perhaps I did not present it well but in years to come this will be the hottest topic in the football data analyst community and football bloggers will be writing extensively about this subject.
This is Arsenal V Spurs for the weekend game
0.39*5.00 v0.44*3.6=3.5 Goals
Here is my shot on target data for the weekend . Will come back on Monday to review.
Arsenal v Spurs
0.39*5 v 0.44*3.6
1.95 v 1.58
Total Goal Expectation 3.5 Goals
Liverpool v Wigan
0.23*4.83 v 0.3*3.28
1.11 v 0.98
Total Goal Expectation 2.1 Goals
Man City v Villa
0.35*5.25 v 0.23*2.83
1.83 v 0.65
Total Goal Expectation 2.5 Goals
WBA v Chelsea
0.31*4.55 v 0.30*3.88
1.41 v 1.16
Total Goal Expectation 2.5 Goals
Fulham V Sunderland
0.33*5.6 v 0.24*3.48
1.84 v 0.83
Total Goal Expectation 2.6 Goals
West Ham v Stoke
0.44*3.58 v 0.21*4.00
1.57 v 0.84
Total Goal Expectation 2.4 Goals
WBA v Chelsea was spot on. ironically the one most people questioned.
I am looking at arranging shot on target updated tables for the big leagues if anyone is interested.