Introduction

I currently work as a supervisor in live data production for primarily South/Latin American professional football leagues including the Copa Libertadores, Brazil Série A-B, Chile Primera División, Colombia Primera A, Ecuador Liga Pro, Peru Liga 1, Liga MX, and MLS. We produce live match details for media clients including lineups, goals, assists, cards, substitutions, and play-by-play details with the league detail level dependent on the data package requested by our clients. As I seek to turn a cool job into a viable career in the sport that I love, I have decided to begin taking statistics and data analytics courses in my free time while practicing practical applications using the extensive and easily exportable data produced in Football Manager. This story will document my journey from data monkey to data expert (hopefully, probably not). I will begin as an unemployed, unlicensed, former Sunday league footballer (as is the case in real life) applying for jobs in the lowest available division of the USA league system (USL League Two). My database will be limited to the countries I currently cover at work and I will use a data-driven approach in my recruitment and team selection. Let the game begin!

The First Step

After applying for every job in the USL2 Southeast Division, where I currently reside, I have agreed to become FC Miami City‘s manager. Nearly half of the clubs in the division turned me away for demanding they fund my first coaching course. I feared I would strike out in every interview in the division before needing to accept a role without the promise of a coaching license on the horizon. Luckily, FC Miami City agreed on the proposal after I promised to guide them to a top-half finish, exceeding their initial proposed expectations. 

The Club

FC Miami City was founded in January, 2014 as a member of the United Soccer League. The club has strong ties to Paris Saint-Germain, playing a role in the development of Paris Saint-Germain Academy USA players by offering them pathways to the USL. Our home ground, Central Broward Park, is a 20,000-capacity dedicated cricket stadium in Lauderhill, Florida, one of just two dedicated cricket stadiums in the country.

Expectations

With the fourth strongest squad in the league (according to the bookmakers), we should be able to easily achieve the agreed upon expectations of finishing in the top half of the table. I have yet to perform an in-depth analysis of our squad and it's strengths/weaknesses, but I will do so and provide a breakdown in my next update. Hala Miami City!

This sounds pretty cool. Looking forward to seeing how you bring the data element in at that level 

 The Squad

Right from the jump, it became clear I would need to be creative in my tactical setup to accommodate our strengths and cover our abundant weaknesses, the most obvious of which being the near complete lack of options at center back. At the moment, our only natural central defender is Jesus Tinajero and the transfer window is closed for another month. We have strength in depth at wing-back and far too many strikers. Two solutions became clear, convert the wing-backs into wide center backs and utilize a two-striker partnership. With relatively high aggression, work-rates, and pace spread around the squad, a gegenpress style of play was also the clear tactical choice. Defending deep was not an option considering the lack of height and strength available in the wide center back positions. 

The System

I settled on a 3-1-3-1-2 formation, which I have honestly never used in any previous save. I normally prefer a somewhat standard 4-3-3 formation with the roles dependent upon the players available. However, I'm relishing the opportunity to build a new system and style to get the most out of my eccentric Floridian squad. The idea is to play a wide game with plenty of crossing, either from deep with our very competent crossers at wide center back or from the byline with our touchline wingers. Garcia will float around the edge of the box to pick up second balls and thread passes into the box. Smith & Jackson both possess strong tackling ability and work-rates, hopefully preventing counter attacks through the middle. As I mentioned before, this system is far outside my comfort zone so it is very much a work-in-progress. 

The Table

The season is already in full swing, with the club currently underperforming in 7th. Hopefully we can begin the climb the table as we work out the kinks in the new system and I will likely make further changes following a dive into our performance data to this point. I will likely post another update looking over our performance metrics and areas for improvement in the following update, stay tuned!

Love a good Moneyball save. Def following.

Team Performance Data

It has been a slow process but I have finally extracted and organized the team performance data in several ways to better understand my team's playing style, strengths, and weaknesses prior to my arrival. As you can see in the screenshot above, my performance analyst was of very little help in this process. He did point out that we have been playing in a 4-2-3-1 setup for the most part, so at least he's aware of formations I suppose.

The first step was extracting the data provided in the Competitions>Stats>Team Detailed page within the game, which is not easily exportable like the player data in the squad overview or the recruitment hub. Being the data monkey I am, I manually entered each value into a Google Sheet to lay the foundation for my analysis. The goal was to identify areas in which we were the among the weaker sides in the league and those where we were strongest, first seeking correlations and possible causations and then digging into each area of the data which could be improved upon or further exploited.

Using a Python script, I derived correlation values between each of the team performance metrics to identify any surprising or interesting correlations that may not be expected. Unfortunately, there were no real surprises and pretty much all of the data points aligned as you would expect. Feel free to peruse the correlation matrix above for yourself but you likely won't find anything out of the ordinary. For reference, dark red indicates strong correlations, dark blue indicates strong negative correlations, and the lighter areas indicate weaker correlations.

FC Miami City Performance Data

The next step was to dig into our performance data and create visualizations that would help us better understand our standing within the league. I used Google Sheets to generate percentile values for each column to ease the creation of radar charts that would help display each area of our performance in relation to the other teams in the league. In the chart above, you can see how our shooting frequency and efficiency compares to the other teams based on percentile values of each metric. While we are taking very few shots per game, we are relatively efficient in our finishing ability. We are slightly underperforming our xG by -0.05 per game, however our shot on target rate (SoT %) is the strongest in the league at 43.33%. Our conversion rate is also very strong, while the rest of the shooting metrics leave us somewhere around the middle of the pack. Essentially, it seems we are creating few chances per game yet we are creating high quality chances in each game, and our finishing ability is strong enough in comparison to some of the other teams. To better understand our chance creation, I then created a radar chart to demonstrate our creativity metrics in relation to the league.

It quickly became clear that we are not a possession-based team and have been relying on our strong crossing ability to generate chances for our lone striker in our 4-2-3-1 formation. As I suspected, we are among the lowest in the league in terms of chances created per game, passes completed per game, final third passes per game, and fouls won per game. However we rank very high in the league in terms of cross completion rate, volume of crosses completed per game, and successful dribbles per game. Surprisingly, despite this direct play-style, our pass completion rate is also quite high in relation to the other teams. Hopefully with the new 3-1-3-1-2 system we can extract more value from these strengths by placing even more emphasis on wide play while also benefitting from an extra striker in the frontline. The final step of my analysis was to better understand our defensive tendencies.

In this visualization, I reversed many of the percentiles to better represent our strengths & weaknesses, meaning the lower percentile in areas such as “Goals Allowed Per Game” (GA/Game) indicates a higher number of goals allowed per game. As you can see above, while we rank very highly in expected goals allowed per game (meaning we are allowing very few expected goals per game), we are allowing far more goals per game than most of the other teams in the league. Considering that we are allowing very few shots per game and very few shots on goal per game, it's clear that our goalkeeping is an area of weakness. In addition, we are very weak aerially, which was already made abundantly clear in my initial squad overview when I realized we only have one natural center back. We rank fairly highly in the league in terms of the number of times we're winning possession of the ball in each game, which correlates strongly with points per game. However, we are allowing a relatively high pass completion rate for our opponents and a relatively high number of passes per defensive action, indicating an inefficient pressing style. Hopefully with our emphasis on a high pressing style of play, we can begin to disrupt our opponents' play more effectively and further limit the number of chances our goalkeeper has to face before we are able to find a replacement for the keeper to shore up that weakness. Over the next week, I will dig into individual performance metrics and potentially create another post here about those findings before finally beginning to play some actual football. Bear with me, please, haha.