This article is partly inspired by a conversation I had with Jamon Moore regarding our Where Goals Come From series on American Soccer Analysis – As we move into season two of the project, we wanted to more closely link xG to our progressive passing model: Something which anyone with even a passing interest in analytics will be aware of. Our discussion involved a look at an article written by ModernFitba (RIP) three years ago now regarding Miles Storey – Then of Partick Thistle, and now plying his trade for Inverness. To sum it up (you should read it too, of course) Jason talks about Expected Conversion Rate, a metric of xG divided by  shots (unblocked ones) – Essentially what % of shots should have found the back of the net. Unblocked shots can be additionally measured with “Fenwick-adjusted” Expected Conversion rate, a hockey concept, but that’s neither the time nor place for that.

This conversation further evolved when I got into a debate on Twitter (yeah, yeah, I know – My fault) regarding a player I worked with at the University Of Virginia, Daryl Dike. The other person online was insinuating that because UVA played defensively (not true), Daryl was unable to generate high shooting volume (not *explicitly* true)  thanks to lack of opportunities to shoot. Other than the fact I also spit out my coffee, this was patently false: The UVA teams I worked for were some of the most ball dominant teams in college soccer, regularly racking up high possession numbers in the final third, pressing up the pitch etc. This naturally had an effect on how teams reacted defensively – They dropped deeper, put 11 players behind the ball and looked to bunker in. Space was at a premium. Daryl, despite his talent level in comparison to lots of opponents had very little opportunities to gain yards of space and shoot 5-6 times a match.

These discussions all together served as the inspiration for the article – How does one contextualize an attacking player’s shooting trends? How much of this is a product of an individual flaw/strength of a player, and how much of it is related to team strength (or weakness), team style and opponent’s reactions to these?

Setting The Stage – The Dike Dilemma 

To provide a concrete example with the player who inspired this piece, let’s take a look at Daryl Dike’s final year in college – The 2019 NCAA season in which we went all the way to the national championship game. 

In 23 games that season, Daryl took 58 shots – 2.5 per game. Scoring 10 goals. Compare that to some other top strikers in the game and their shots per 90:

• Lewandowski: 5.03
• Kane: 4.07
• Benzema: 3.87
• Mbappe: 3.86

It’s quite a difference! What’s the issue? The truth is very, very, nuanced. Some players it’s fairly obvious what problems are, or trends within their game in terms of shooting. A classic is Ruben Neves, who is seemingly incapable of shooting from anywhere inside the penalty area. Daryl’s case provides us a good example of the contextualization that is important when looking at shot maps and general shooting trends.

Team Style + Daryl’s Role

UVA during the 2019 season dominated opponents. Featuring the likes of Daryl, Henry Kessler, Joe Bell, etc. we utilized our offensive organization within the 4-3-3 to pin opponents deep, build using our #6, invert our wingers inside to the halfspaces, etc. – Pretty much every buzzword possible for a high possession team. Below will detail some of the metrics which illustrated this (all per 90):

Expected Goals	2.34
xG Per Shot	.13
Passes Into The Penalty Area	17
Key Passes	10
Progressive Passes In The Final Third	22.3
Final Third Passes	132

Looking at these, you would expect Daryl – A current USMNT player, would be the beneficiary of most of these. However, looking at the shot numbers I detailed prior, that’s not necessarily the case. As ever, we need to contextualize them. 

Daryl’s role within the team was as a traditional #9 – Operating as the spearhead of the attack and providing the link up option between the middle third and the final line. It would be too simplistic to call him a target man however. When building up play or looking to progress the ball into the final third, Daryl was stellar at holding off defenders (either by occupying the final line or dropping off) and finding supporting midfielders underneath. This allowed our wide players to hit/create depth in behind or to simply move lines forward. 

Once the ball was entered into the box, with one CF in our team, a lot of the focus was on him – And naturally, in the penalty area, there would be bodies around him. Regardless of how dominant you are in soccer/football, breaking down deep blocks is HARD. This led to most of our attacking movement & patterns ending with Daryl and being facilitated by wingers, midfielders, etc. – That is not to say he was just a finisher: His key pass numbers were still great due to his individual ability, but it didn’t mean he was this creative force. Most of the metrics and ending attacking moves were performed by the midfielders and wingers that Daryl’s sheer gravity (in terms of skill and also….you know…size) facilitated: While we obvious did our most to create scenarios for Daryl to show his prowess in the final third, the style of play sometimes make it difficult to do so. Transitional scenarios and against teams where possession and the flow of the game was more even (such as Wake Forest, Clemson, etc.) were the games where Daryl was more able to show the type of game he’s played at Orlando and with the USMNT.

The Opposition

Looking at the opposition,  the opponent’s average defensive actions were located around 30 yards from their own goal – That’s all tackles, interceptions, fouls, etc. Compare that to us at UVA, which was about 45  yards from our own goal – A significant increase. Alongside the metrics highlighted, it gives a good account of how we played with the ball and how opponents defended. The previously highlighted metrics are very favorable in our direction in regards to getting into the penalty area, but it also means that Daryl was facing some VERY crowded penalty areas. No space = few opportunities to pull the trigger.

I highlighted it in the previous section but pinning teams back into their own box makes things difficult to find your #9 – This is true for teams at the very top of the world’s game as it is for amateurs in college soccer.  His shot locations are still great (thanks in some small part to my stressing of the importance of xG), but plenty of shots did come when we were able to lure out opponents on the counter, or press high and win it in the final third. His key passes also ranked second in the team, largely due to the focus on him and space it created others. 

Individual Player Considerations 

While most players and their shooting patterns require the contextualization of a team environment to understand the “why”,  there are a few examples of players where a players shots can be isolated more succinctly into how they operate as an individual – Such as their role within the previously highlighted team environment and flaws/issues  which are more subjective – such as technique. 

Trends/Roles

By looking at general shot locations and clusters a player has on their behaviors when going for goal, sometimes you can detect flaws or quirks. Take the player below, who was a player in the NCAA in 2019. There is an easily identifiable pattern and “hot spot” – When cross referenced with the types of events that precede these alongside the general role of his team, there is a pretty clear story to be told. The player in question was a left winger. His role dovetailed with the positioning of the left back in support – Allowing him to occupy the halfspaces inside (similar to what we saw with Spinazzola and Insigne this summer in the EUROs). Naturally, this set him up to receive around the left side of the “D” on the top of the penalty area and shoot, or make runs behind the back line and shoot wide of the penalty area. These types of playing roles and responsibilities (as previously mentioned with Daryl) 

Flaws

Trends can also suss out flaws in a player’s game, but often requires more than just traditional event data to find them. The player aboves shooting trends, and similar contextualization, can allow coaches, analysts, scouts, etc. to help detect decision making faults or technical faults in a players game. 10/15 of the shots taken from this athlete were first time attempts. While this could be spun as a positive (not allowing players to get near and close him down), it was almost exclusively a negative – Making the margin in error MUCH greater. Similar issues that can be found are improper technique, shot selection in terms of avoiding a clear pass, defender’s positioning etc. Advanced shot maps like ones on StatsBomb which include goalkeeper and defender positioning DO include these, but don’t provide the total picture of events leading up to it (Wait till 360 data comes up with the pictures it delivers, however!) Using this information, you can coach players to improve their shooting, use to bolster player recruitment, etc. – The possibilities are limitless.

Linking It To xG

Mr. No Name above and the examination into his shooting habits can be used in conjunction with xG analysis to explain WHY a player might be underperforming his expected goals, rather than just “he’ll revert to the mean eventually!” 

Here we have another player who emphasizes that concept. Those shot locations are pretty good, right? However, as we can see, only one goal scored from 14 attempts – a total xG of 2.35. The sample size isn’t *perfect* obviously, but it details a typical story often told in the context of soccer/football analysis. However, as I’ve harked on constantly, we have to contextualize it. Advanced xG models which account for technique, shot height, goalkeeper positioning, etc. do detail this well, but information such as whether it was the correct technique for a given scenario can further diagnose issues. Perhaps I am splitting hairs, but let me give you three examples:

These three attempts on goal from former Pitt Panther (and current NE Revolution player) are of various styles: A half volley, counter attack/1v1, and a cross. Prior info from xG models provides us what body part Kizza used (left foot, right foot, head, other, etc.) where the goalkeeper was located, and as well where defenders were located. This is important info but it doesn’t provide us with so more subjective information to explain an individual’s shooting repertoire: 

• Was the technique or PART of the foot the right one to use given the scenario and was the collection proper? – Looking at the 1v1 with the goalkeeper, we can notice Kizza slightly flubs the technique, bobbling the ball to the goalkeeper and taking the pace and direction out of it. 

• Was it the correct technique given the pass/scenario? The “bicycle kick” attempt on the cross was likely an example of him making his job more difficult. Rather than pulling away so he can get his head on it, or taking a touch to bring it down, Kizza contorts himself and tries an audacious effort. 

This area of shooting analysis is slightly more gray, and coaches might have different opinions, but used when in conjunction with a reliable xG model (the objective side of things) you can better diagnose individual problems – And crucially link these visualizations and data to an actionable insight on the pitch: Deciding drills and sessions to work on. 

The Complete Picture

Hopefully this brief insight into shooting trends has given you a new way to look at how a player shoots on goal. Looking at the pure data and locations of said attempts only gets you so far – If you don’t root these insights into an actionable and contextualized picture of the game: Both in relation to the collective tactics of a team (both the player’s own and the opponents) as well as more subjective individual aspects you can pick up some false narratives. While this piece is used in relation to shots, which is arguably the most analyzed aspect of modern day sports analytics (especially in soccer), blending the objective and subjective together is a MUST for all analytical approaches to the game – Nothing exists in a  vacuum. It’s impossible to look at pure data and get the full picture, nor is it possible to use the “eye test” by itself.