Analysis: Total Shots vs Shots on Target

I'll start with a question: Which one of Total Shots or Shots On Target has the stronger correlation with a team winning a football game?

The reason for trying to answer this question is simple; a good healthy dose of curiosity as to which mainstream stat may be best used in analyzing teams underlying performances.

Now, most people's stat of preference is total shots. A bigger sample size is one reason for that and, as JamesWGrayson's excellent work point out, total shots ratio has the slight edge in terms of which of the 'advanced' metrics has the most predictive power.

As for shots on target, this stat may lack the sample size of total shots but some believe that it may be a more relevant metric in terms of analyzing a game.

I can understand both sides of the debate, and it is really about personal preference, but surely one of these statistics has a stronger correlation with scoring more goals and winning a football game?

This study is all about if a team wins or loses the fixture.

Total Shots vs Shots on Target

I want to start by using an example fixture from this season. It is an extreme example, but it was necessary to exaggerate the difference in order to emphasise the point.

The screenshot (below) from ESPN soccernet shows Chelsea's recent home defeat at the hands of QPR. When examining the basic stats from QPR's 1-0 win, we see that Chelsea score a heavy Total Shots and TSR win, 26-7.

The shots on target count is a little more telling, Chelsea lost that battle 2-3. An argument can be made that Chelsea were the more dominant of the sides as indicated by the Total Shots win, but do shots really matter unless those shots are of a high enough quality as to be on target? The answer has to be no. For a shot that ends in row Z, or going out for a throw in, or hitting the side netting of the goal never have any chance of being a goal. A shot which ends up high or wide is, ultimately, a giveaway to the opposition.

Do note that this Chelsea game is an extreme example of the imbalance between Total Shots and Shots On Target, and that usually a team which wins the total shots battle will win the shots on target battle. But not always. One of these stats has a greater correlation with a team winning that fixture.

Chelsea QPR

26(2)	Shots (on goal)	7(3)
13	Fouls	8
14	Corner kicks	2
1	Offsides	4
64%	Time of Possession	36%
1	Yellow Cards	1
0	Red Cards	0
2	Saves	2

Method

I have used every result this current season and the 3 seasons previous to that and looked at games that have solely been declare a win, thus ignoring drawn games (23%ish). I have only used fixtures that have resulted in a win as we are solely looking to see if the team that wins the Total Shots battle (north of 50%) wins or loses the fixture. I have done likewise for Shots on Target.

Sample Size

	Games
2009/10	380
2010/11	379
2011/12	380
2012/13	208

That's 1347 games of data to play with. A healthy enough sample size indeed, although I will return to this topic once I have the time to research seasons past.

For this specific exercise we need to eliminate draws. Why? We are solely interested in a teams ability to win the total shots battle or the shots on target battle and how that correlates with the wining, or losing, the fixture.

Below are the four seasons of data I have collected, with draws subtracted:

draws	remaining
96	284
111	268
93	287
60	148

987 games is our reduced sample size. let's begin.

Total Shots Record

We have the aforementioned 987 games sample now, so this is what we are looking at: Did the team that won the total shots battle win or lose that individual fixture?

Games in which the total shots battle was tied are naturally exempted from the win or loss count and are listed in the far right column.

	shots won	shots lost	shots tied
2009/10	184	85	15
2010/11	165	97	6
2011/12	175	102	10
2012/13	97	47	4

Clearly, when examining our 987 fixtures that have resulted in a win or a loss for a team, we see that the team which records a clear majority of total shots wins more games than it loses.

In Graph Form

The total shots record in gaph form gives us a greater visual understanding of the gap between wins and losses.

There is a far amount of variation from season to season, as would be expected due to differences in teams' quality within the league for each individual season.

Using our original sample size of 987 games:

The team that wins the total shots battle went on to win the fixture 62.91% of the time.

The team that wins the total shots battle but went on to lose the fixture occurred 33.53% of the time.

Shots On Target Record

	sot won	sot lost	sot tied
2009/10	210	49	25
2010/11	192	55	21
2011/12	202	53	32
2012/13	104	34	10

Immediately, and even in table form, we can clearly see that the team which wins the shots on target battle and goes on to win the fixture is far greater than the Total Shots numbers listed above.

In Graph Form

Again, there exists variance from season to season but this is to be expected.

When looking at the team which wins the SoT battle in a fixture, we can see that not only does that team go on to win the fixture at a higher rate than TS told us, but that team that records that higher share of SoT is losing far, far fewer games.

In short, a team that wins the SoTbattle is far more likely to go on and win that fixture than the team that wins the total shots battle. Shots on target should, logic determines, be able to tell us more about that teams performance than Total Shots does.

Using our original sample size of 987 games:

The team that wins the Shots On Target battle went on to win the fixture 71.73% of the time.

The team that wins the Shots On Target battle but went on to lose the fixture occurred 19.35% of the time.

Year By Year Total Shots v Shots On Target

So I have decided to take things slowly and break the data we have already looked at into smaller pieces. The above graph shows us how many wins the team that recorded the highest proportion of Total Shots and Shots On Target. Those two numbers are placed side by side for each of the 4 seasons.

Blue is Total Shots, Red is Shots On Target.

When we directly compare the win record of Total Shots and Shots On Target it becomes clear that there is a disparity between the two popular stats. Winning the Shots On Target battle is a better fit, and predictor, at who will win that fixture than Total Shots is.

That is what the 4 years worth of numbers tell us.

Total Shots vs Shots On Target Losses

Below we will see the Number of games where a team won the Total Shots or Shots on Target battle but lost the fixture, year by year.

Blue is Total Shots, Red is Shots On Target.

Remember, the loss numbers listed above for Total Shots and Shots and Target are for the team who won those individual battles but lost the fixture.

Again we see fluctuations year to year for the three completed seasons. The 2012/13 season exhibits signs of the gap between Total Shots and Shots On Target being much closer, but that gap will widen as the season progresses.

Losses for Total Shots are 331 in 987 games

Losses for Shots on Target are 191 in 987 games.

Again, the information in front of us points us to the conclusion that a team which wins the Total Shots battle being a poorer predictor for who will win the fixture that Shots On Target is.

Total Shots vs Shots on Target: Win%

In the following series of graphs we take the raw numbers that are listed above and convert into %'s

The methodology is simple, take the games won by the victor of the Total Shots or Shots On Target battle and divide them by the games which didn't finish as a draw.

We can see that the win% for each season varies and that 2009/10 looks like an outlier of sorts. Shots on Target win% is consistently superior to Total Shots win%, and despite year to year variance the gap in win% between the two numbers remains within the 9-10% corridor for the three full seasons of data we have.

The mean line indicates the average difference between the two shot statistics.

Total Shots vs Shots on Target: Loss%

Same method as above, only this time we are looking at the % of time a team lost a game despit winning the Total Shopts or Shots on Target battle.

Loss% displays a touch more variance season to season than win% did, but we arrive at the same conclusion as we did for win% as we do now for Loss%: Total Shots has a poorer correlation to the team that wins the fixture than Shots on Target exhibits.

That is clearly evident in the above data showing us that the team which wins the Total Shots battle loses a far higher % of games than the team which wins the Shots on Target battle.

4 Year win, draw and loss %'s for Total Shots and Shots on Target

Total Shots is the red line, Shots on target is the yellow line

This is purely a alternate take on the information we have seen above.

Shots on target has a higher win rate than total shots. Shots on target has a lower loss rate than Total Shots. And the number of non drawn games that were ruled out due to the Shots on target count/Total Shots count being tied at 50-50 in any particlular fixture is listed at the bottom of that graph as tie%.

Conclusion

Although we know Total Shots and TSR to be the most repeatable skill from season to season, the above information clearly shows that the team which wins the Shots on Target battle has a higher chance of winning the non-drawn fixture than winning the Total Shots battle does.

Winning the SoT battle in non-drawn games, results in a team winning that fixture 71.73% of the time and losing the fixture 19.35% of the time.

Winning the TS battle in non-drawn games results in that team winning the fixture 62.91% of the time and losing the fixture 33.53% of the time.

This study is not designed to win any arguments or state that using total shots is incorrect. The original question was: Which of Total Shots or Shots On Target has the stronger correlation with a team winning a football game?

Now, although both metrics strongly correlate with winning that fixture, it is Shots on target which has the stronger link by virtue of having a win% 8.82 percentage points higher than Total Shots does using a sample of just short of 1000 games.

Cheers for reading and congratulations if you made it to the end!

---

Follow shuddertothink on Twitter