A quick look at Total Shots Ratio
After some decidedly mixed yet encouraging results while using my data for predictive purposes, I’ve been doing some reading about how other analytical bloggers have approached this. I was interested to read about Total Shots Ratio (TSR), particularly in this post by James Grayson, and its ability to predict future performance.
What I like about it is that it’s such a simple measure: the ratio of shots that a team has compared to the opposition, with higher values suggesting a greater amount of time in control of the ball. It’s not a perfect predictor by any means: by definition it ignores shooting efficiency (i.e. the vertical axis on my effectiveness graphs), which is obviously a simplification. The quality of chances created remains important, and some teams do very well with a counter-attacking style while others create plenty of chances but lack a clinical finisher, but analysis shows it to be surprisingly reliable and it’s a much quicker way of comparing teams.
What I’ve done below is to append 2 columns to the current league table for each Football League division (emulating a similar approach by the excellent 11tegen11 blog for the Dutch Eredivisie): firstly the TSR for each team and secondly how this ranks against the other 23 teams (also colour coded for ease of reading). If TSR truly is a reliable predictor of long-term performance, then we can expect to see a team’s league position change over time to correlate with its TSR rank. By implication, those teams whose league position and TSR are very different are likely to be over or under-performing.
- While the top is definitely ‘greener’ than the bottom, it’s far from convincing with 4 of the 12 teams in the top half have ‘bottom half’ TSRs (and vice versa obviously).
- Starting at the top, Crystal Palace and Middlesbrough are definitely outperforming their TSR – they’ve not been dominating matches to the extent associated with their position at the summit, so it’ll be interesting to see if they can maintain this.
- While most of the teams currently in the play-off picture have commensurate TSRs, Huddersfield and Blackburn are significantly outperforming and could well start to slip away.
- Further down, Blackpool, Sheffield Wednesday and Bolton are the most significant underachievers – neither has been able to turn their relative dominance of matches into points, with their TSRs suggesting that we should expect them to rise over the course of the season.
- I was surprised to see that Leeds were the division’s lowest-ranking team – they could be in for frustrating seasons if this doesn’t change, and the number of underachievers beneath them could see them dragged into a relegation battle.
- Again we have 4 teams from the top half with a bottom half TSR rank, and the top 2 teams (Tranmere and Sheff Utd) seemingly overachieving.
- What’s interesting here is that with the exception of Hartlepool the division’s lowest TSR teams are significantly outperforming this metric, sitting in mid-table or above. Does this mean that we’ll see a lot of teams falling away as the season goes on, or is this a feature of pragmatic or counter-attacking football being particularly prevalent at this level?
- Stevenage and Notts County are the most noticeable higher-placed teams outperforming their TSRs – is their recent stuttering form the beginning of a ‘correction’ to a lower league placing?
- Most of the sides in the bottom 7 look capable of a higher placing, with Portsmouth and Oldham in particular underachieving given that they consistently out-create their opponents.
- While the very bottom of the table looks suitably red, overall half of the teams are in the ‘wrong’ half of the table compared to their TSR, making this the most mixed division of the three.
- We again have ‘overachievers’ at the summit, with 5 of the division’s top 8 teams significantly outperforming their TSR. Given the amount of goals they’ve scored, Port Vale‘s low TSR is surprising, although they’re clearly very clinical with the shots they do take.
- Lower down, I was surprised to see Dag & Red with such a high TSR – along with Oxford and Bristol Rovers (who experienced fantastic and terrible starts to their campaigns respectively), the numbers suggest higher league placings in the long run.
- League 2 houses the Football League’s highest and lowest TSRs: Bradford have the highest, which bodes well for their promotion push, while Accrington have the lowest: they’ll do well to hold onto their mid-table position.
- As I mentioned previously, the bottom part of the table is strongly correlated, with the lowest 4 places occupied by teams with low TSRs. Perhaps the relegation candidates have assumed their positions early?
EDIT – a more graphical approach
I realised that I could display this information visually by adapting my existing attack and defence graphic. The dashed line is where a team would have exactly half the shots in a match, so the further away (perpendicularly) from this line a team is, the greater the degree of domination (bottom right) or being dominated (top left) in terms of chances created.
- You can see how busy Burnley‘s matches are compared to York‘s – both give as good as they get, but there are 6-7 more shots on average in games involving the former.
While the tables are inconclusive at present, I’ll revisit this analysis a bit later in the season to see whether positions and TSR ranks start to converge. We’ve seen in the ‘first 10 games’ report that there are a lot of other factors at play which can influence a team’s league performance, so I expect a fair few teams to go on defying this metric and it will be interesting to see just how useful a predictor it is. At the very least it will give us a broad measure for which teams are under or over-performing respectively.
Very interesting and certainly a good bench mark to make a long term prediction.On the data I wondered where you gain your stats from as my records indicate Leicester average 10.31 shots per game but against I get the figure at 6.88 which is way short of the 60% supremacy that you show here.
Your stats are correct: 10.31 / (10.31 + 6.88) = 10.31 / 17.19 = 60%
My apologies.Maths has never been my strongest subject.
Great work and I should think that doing the same tables based on home and away performance will highlight some interesting trends.
Pingback: The quest to manage the ‘state of emergency’ « The Width of a Post