What We Learned Tracking Every NCAA UNC Women's Soccer Shot - Data Analysis

UNC's Anson Dorrance is the most iconic NCAA soccer coach in history.

The NCAA wins leader has 934 career wins, emerged victorious with 21 national championships, won the first-ever Women’s World Cup in 1991, and coached the likes of Mia Hamm, Kristine Lilly, Cindy Parlow, and Tisha Venturini.

He has also managed players who went on to become elite coaches themselves.

Saturday’s USWNT versus England matchup featured former Tar Heel Sarina Wiegman against former WSL and Chelsea manager Emma Hayes.

Hayes didn’t don the Carolina blue, but she did have Dorrance as a course instructor, so his legacy goes well beyond the university.

The allure of playing for Dorrance has consistently brought top talent to his program.

Just last year, three of his players were selected in the first round of the NWSL draft, including the top two picks in Ally Sentnor and Savy King, and another three players were drafted.

Still, more players went undrafted, yet are playing professionally now.

UNC is a model of consistency.

That's what makes them such a fascinating case study.

Given the coach's profile and the talent within the squad, I simulated a postmortem for the team's 2023 season.

Starting with data, I built a sample report that included film from the season to identify strengths within the current group and areas for improvement.

This data analysis is a sampling of the mock report.

While there is no video, we have a sampling of the process used to produce a portion of the report.

For selfish purposes, it was an excuse to build a model for a season review.

This article gives an idea of the direction and workflow.

Unpacking UNC’s Women's Soccer Season

Let's set up the problem.

When reviewing 2023 UNC's Women's Soccer season, we first wanted to look at objective performance data.

That's where we looked not only at goals for and against but also at xG statistics.

Scoring goals is the ultimate objective, but to score those goals, you have to create chances.

In theory, the more opportunities and the higher-quality chances generated, the more successful a team should be in putting the ball in the back of the net.

In the first graph, we immediately notice that UNC was one of the top few teams in xG, making them one of the most prolific attacking sides in the country.

NCAA D1 Women's Soccer Goals & xG Chart - 2023

The issue is that goals per game lingered well behind xG.

While UNC produced 2.72 xG per game, their goals per game sat at 2.04, a 25% decrease.

Throughout a 25-game schedule, that's a lot of goals to leave on the table.

The problem starts to take shape.

We arguably have the country's most dominant side in chance creation, which rates slightly above average in goals per game.

That's our sign to dig even deeper into the shots.

The next step was to look at the shots P90 and the ratio of long shots to shots.

Again, UNC ranks number three in the nation in shots P90 while having a slightly above-average ratio of long shots to total shots.

NCAA D1 Women's Soccer Ratio Of Shots To Long Shots Chart - 2023

Turning to an analysis of shot quality and field tilt, measured here based on touches in the box, UNC is once again elite on the horizontal axis tracking fuel tilt and slightly above average in xG per shot.

NCAA D1 Women's Soccer xG Per Shot & Field Filt - 2023

Putting it into a grading system, we're looking at an A+ team in shot creation but closer to a B- once they get into the box.

Creating wasn't the issue, and even though shot selection and quality weren't on par, they were still decent relative to the data from the top 100 RPI and all Power 4 universities.

We needed to dig in here.

More detail was necessary.

Disparity Between UNC Women’s Best & Worst Opponents

There is a huge quality gap in college soccerbetween the top teams in any given division and those at the bottom.

With hundreds of teams registered at each level, there's a significant gap between the haves and have-nots.

The issue is more prevalent on the women's side, given that the USA is still the primary recruitment focus.

In contrast, the men's game will often have a balance between domestic and international signings.

For the men and women, the top domestic players tend to sign for historically dominant programs or those with the best facilities.

International signings allow men's programs to add top-end talent that can compete with the best domestic recruits, knowing that those best American players only look at a small list of schools.

The perspective player pool is larger on the men's side than on the women's side.

That's why our initial data collection focused on the RPI top 100 and Power 4 universities.

For the most part, that's the level of competition UNC faces any given season.

The only reason to schedule a team outside of those parameters is to get an early game in where the team can have success, score goals, and assimilate the newcomers to the team's game model.

Theoretically, a tougher out-of-conference schedule strengthens the team's RPI for postseason seeding and prepares the group for conference play.

That's where we want to get a sense of how their data compares to the rest of the conference.

Focusing on goal production by UNC and their ACC rivals, a look at the top five teams gives a sense of the number of goals they produced on a game-to-game basis.

UNC and Clemson were the standouts right away, with zero or one goal scored.

Pittsburgh also had some issues, especially getting shut out, but performed better in the number of matches with just one goal.

ACC Top Tier Soccer Goals Per Game Data

Once we get to two-goal games and beyond, we see Florida State's and Notre Dame's consistency.

We find some interesting statistical discrepancies when we segment UNC's shot data into matches against the top 100 RPI teams versus 101+.

In the 12 matches against top 100 RPI opposition, UNC managed to match their xG total with 24 goals off of 24.34 xG.

The 0.11 XG per shot is slightly below average, but some regression can be expected against better opposition.

If anything, there was simply more of a willingness to take a shot from distance rather than continuing to break the opposition down, which shooting locations and film verify.

Looking at the film, UNC did tend to possess the ball in the final third for most of these games.

One of their difficulties was that the opponent typically had numbers behind the ball and could easily move from mid- to low-blocks.

For UNC, the seven matches against teams outside of the top 100 RPI were feast or famine.

In four games, UNC scored just a single goal, while the other three saw an average of five.

They were averaging approximately 27 shots per game against this group and a per-game xG of 4.24, an outlandish number.

But here we do see some of the issues.

Nineteen non-penalty goals from 29.68 XG is problematic.

0.16 xG per shot is a fantastic number, which only produces more questions than answers.

To get to the heart of the issue, the next steps of data analysis were clear.

It was time to track each shot.

Tracking UNC’s Shot Locations And xG Data

This is where we really got into the specifics of the shots.

Plotting each shot, we identified their distance from goal and which zone they fell into.

To create customized zones, we segmented the horizontal channels using distances in 6 yd increments and used natural markers on the field to create our vertical boundaries.

The width of the goal was labelled Zone 1, the space between the posts and the 6-yard box was Zone 2, Zone 3 was the distance from the 6 to the 18, and the largely consequential distance in the wings was Zone 4.

Here's what that shot breakdown looked like.

UNC took 87 shots from Zone 1, scoring 15 goals, a 17% shot success rate.

49% of their shots were at least on the frame, either scoring or forcing a save from the goalkeeper.

Zone 2, which, like zones three and four, falls on both sides of the goal, saw the team take 99 shots and score 10 goals.

The team's success rate was 10%, with 46% of shots landing on the frame.

Zone 3 is where we see a tremendous dip in production.

From 104 shots, UNC managed just four goals, with approximately a 4% success conversion rate, with 37% of the shots finding the frame.

Adding in the distances from goal, we get a full breakdown of where UNC was the most efficient.

To no one's surprise, within the width of the goal and 0 to 6 yards out led to a 57% conversion rate, and 71.4% of the shots found the target.

Three of the four most efficient shooting zones and four of the top six are within the width of the goal.

The third most successful shot location was between the goal posts and the 6-yard box (from 0 to 6 yards out), meaning the wide spaces of the 6-yard box.

Ultimately, if UNC were to score a goal, the shot would come within the width of the 6-yard box and typically from inside the box.

The other shooting zones ranged anywhere from inefficient to not happening.

Another standout data point is the inefficiency of shots taken between 6 and 18 yards (Zone 3 above).

A 4% conversion rate indicates how difficult it is to score from outside the 6-yard box.

Shooting from that width increases the distance to goal, narrows the angle to the target, and leaves the goalkeeper with less of an angle to cover.

It’s a trifecta favouring the goalkeeper.

After finishing up with a relative view of UNC shooting data, we wanted to compare them to Florida State, the top team in the conference and the 2023 NCAA D1 women's soccer national champion.

It's interesting to note that Florida St.’s average xG when they scored was 0.31, whereas UNC’s was 0.24.

When Florida St. scored, it tended to come from higher-quality chances of getting a goal.

That's our first column.

The remaining columns look at the total number of shots each team took and break them into ranges.

We start with 0.001-0.05 and extend to the top range of 0.51 and better.

The shots are sorted by the percentage of a team's overall shots taken within that xG range.

UNC had a significantly higher percentage of shots fall within that lowest value tier.

Moving through each of the remaining columns, it's clear that Florida State had just a slight percentage improvement over UNC, but those small advantages add up to something significant in the end.

Conclusion

By breaking down the data behind UNC’s shots, we now have clarity on the directions the team could take in training.

Now we know which location on the pitch the goals came from and which parts of the field are less productive.

We can go to the film and identify specific scenarios leading to more favourable shots versus less favourable shots.

Since UNC is a team that dominates possession, you'd want to examine issues arising from sustained possession in the attacking half and find alternative means of engaging the opponent.

IDP training can also factor in.

Whether it's training alternative actions that lead to shots from wide areas or improving efficiency within the width of the six, we can look specifically at recurring patterns and help the players engage those situations with a new framework.

We can train their eyes and awareness to see new patterns and make better decisions in the flow of play.

This type of season postmortem can help teams identify what's working and what's not.

It's a means of correcting mistakes ahead of the next season and part of the process of continually refining the game model.

In our opinion, a season postmortem is an exercise of immense value.

The insights on player habits and the team's execution of tactics lay the foundation for next season's approach.

Data-driven clarity guides our reflection and paves the way for progress, setting the stage for next season's evolution.