clock menu more-arrow no yes

Filed under:

Beyond the Numbers: Shots and Success

New, 10 comments

We continue our statistical series by looking at shots and success (points) in Major League Soccer

MLS: Vancouver Whitecaps FC at FC Dallas Matthew Emmons-USA TODAY Sports

It is time for our next installment of Behind the Numbers, where we look at the statistics behind the Vancouver Whitecaps season. In the current series, we are trying to answer the question “What Statistics Predict Success?”. We began by examining possession and continued with passing. There were some great discussions, especially in the passing article. There was also some good feedback for me.

Because of feedback, I am going to try and adjust the presentation of statistics to not overwhelm with too many numbers. That is, I will get to the crux of the argument quicker and try to summarize the results. However, if you are curious about some of the more in-depth numbers, let me know and I will be more than happy to pass those along. Forewarning though, there are a LOT of numbers and tables that have been combined to get to the results you see below. With that covered, let’s get to the question at hand.

Today, we are looking at how shots lead to success in Major League Soccer. As usual, MLSSoccer.com is not the best for statistics, so I will primarily be using WhoScored.com. For the 2017 shot data that I used, here is a link. I also included some 2015 and 2016 numbers, but will not focus too much attention on those. Like with the previous articles, I will use points as a measure of success. Since we are only halfway through the season, I have transferred everything to points per match, and I have broken down comparisons to overall, home, and away. Finally, I have only focused on shots from an offensive perspective. Of course, this is only half of the equation as preventing shots (or certain types of shots) may be very important as well. However, in an effort to keep these articles at a reasonable length, I will save that discussion for another time.

2017 Shots Per Game

Let us start with the simple discussion of shots per game. Real Salt Lake take the most shots, averaging 15.3. This is almost 1 additional shot per game than the next closest teams New York City FC (14.5) and Sporting Kansas City (14.5). As for the Vancouver Whitecaps, they sit middle of the pack (13th out of 22), averaging 12.2 shots per game.

It is no surprise then that when you look at shots on target, the same teams are near the top of the statistics. However, it is not perfectly correlated. While RSL take the most shots per game, they are tied for third with SKC and Houston Dynamo at 5.0. NYCFC lead the league, averaging 5.6, followed by Chicago Fire at 5.1. Once again, Whitecaps sit middle of the pack, although a bit lower this time (17/22), averaging 4.0 shots on target.

Historically, shots per game (SPG) have been moderately correlated with points. In 2015, the more shots a team took on the road, the more likely they were to win (Corr = 0.21). However, if they took a lot of shots at home, they were less likely to win (Corr = -0.12). A similar pattern emerged in 2016. Once again, the more SPG a team took on the road, the more successful they were (Corr = 0.33). At home, the pattern was the same, but slightly weaker (Corr = 0.19). Has this pattern held in 2017?

For 2017, we will go a bit more in-depth. As expected, teams that take more shots are more successful (Corr = 0.40). And, once again, this is more important on the road (Corr = 0.52) than at home (Corr = 0.11). In fact, when we combine the three years of data, we begin to see the pattern that SPG at home is not important, but SPG on the road IS important.

It is one thing to take shots and have them sail into the Southsiders or Curva Collective crowds at BC Place and another to force the keeper to make a save. Here we see a bit more of a pattern in 2017. Overall, shots on target were highly correlated with success overall (0.60), especially at home (0.72). Interestingly, while still important, SOT were less important for success on the road (Corr = 0.47).

Table 1: Correlation between Shots and Success

. Overall Home Away
. Overall Home Away
Shots per Game to Points 0.40 0.11 0.52
Shots on Target to Points 0.60 0.72 0.47

Shot Zone

With the previous articles on possession and passing several people noted that game context was important, as were the locations of the possession and passing. For example, a team could pass the ball around the back many times, resulting in good poss/pass statistics. A team that is constantly attacking may have poorer passing statistics because they are pressing and taking more risks with their passes. While I wasn’t able to address those concerns very well in the other articles (numbers were not available), I can do that a bit here.

I suspect that the first comment about the above data on shots per game and shots on target would be ‘yeah, but where are these shots coming from? A slow dribbler from outside the 18 is less important than one from inside the 6’. True… or so you think. The numbers tell us something a bit different.

Orlando City SC relies on shots outside the 18 the most (52%), followed closely by Sporting Kansas City (50%). There is a big drop-off to third place Montreal Impact (45%), while the Whitecaps are tied for 7th (with Toronto FC) at 42%. On the other end of the spectrum, Minnesota United rely most heavily on shots from inside the 6 (11%), followed by Colorado Rapids and New England Revolution (10%). What you might notice is that Min, Col, and NE are near the bottom of the standings. As for the Vancouver Whitecaps, overall, 58% of shots come from inside the 18 (including inside 6), with a slight up-tick at home (60%) compared to away (57%).

What about the overall story? Do the location of a team’s shots translate to success? Not really. The only place that has a modest correlation is shots at home from inside the 6 (0.28). Otherwise, the correlations are basically ‘0’.

Table 2: Shots Zones and Success

. Overall Home Away
. Overall Home Away
In the 6 -0.078 0.276 0.005
In the 18 (outside 6) -0.028 -0.062 0.073
Outside 18 0.067 -0.081 -0.088

Conclusion

The numbers around shots is interesting. We can see that there is a difference between a shot and a shot on target. Shots per game are not very important at home, however, shots on target are important for success. On the road, while shots on target are still important, simply shooting a lot seems to be correlated with road success. The impact of shots on success does not appear to be effected by the location of the shot. Outside of shots inside the 6 at home, location was not correlated with success.

What are your takeaways from this analysis?

What is Correlation?

Since I know that some people are not that familiar with statistics, I thought I would conclude with a brief definition/explanation of correlation. For those that are familiar with correlation, you can skip this section.

Correlation is a statistical measure that attempts to show a mutual relationship between two (or more) things. The most common measure of correlation is Pearson’s coefficient. This measure ranges from -1 to +1. If there is no relationship between the two variables (e.g., shots per game and points per game) then the correlation will be ‘0’. It is important to note that a negative correlation is just as important as a positive correlation. A positive correlation simply means that when one measure increases, the other measure also increases. In our study, a positive correlation would mean that as a team takes more shots their points per game increases. However, a negative correlation means that as one measure increases, the other decreases. In our study, this would mean that as a team takes more shots, their points per game decreases. As you can see, a negative correlation can be very important.

The final, general, question I will answer here is ‘What is a good/high correlation?’. However, if you have more questions, do not hesitate to ask them below and I will answer them and add them to future articles. Back to the question. A good/high correlation is hard to answer. For example, the correlation between lung cancer and smoking is something like 0.20 to 0.40. However, we are confident that smoking is related to lung cancer. In general, for the articles that I will be posting, I would consider above 0.30 to be good with above 0.60 to be great!