Overview of Why Soccer Analytics Works Like Volatility Arbitrage Trading
This Bloomberg Odd Lots episode explores how soccer analytics has evolved from a supposedly “too fluid to model” sport into a data-rich field that increasingly resembles finance and trading. Hosts Joe Weisenthal and Tracy Alloway speak with soccer analytics consultant Joris Beckers and volatility arbitrage trader / soccer analytics practitioner Mike Tracy about how teams use data to scout players, improve tactics, manage risk, and make better decisions in a game defined by uncertainty, low scoring, and lots of noise.
Key Takeaways
- Soccer analytics has matured significantly since the early Moneyball era, moving from basic event stats to tracking data, expected goals, expected possession value, and now body-pose / skeleton-level analysis.
- The game’s randomness is a feature, not a bug: analysts often have to model around red cards, VAR, referee decisions, and skewed match states.
- Finance and soccer are similar in structure: both involve high-variance outcomes, imperfect information, and decisions under uncertainty.
- The real bottleneck is interpretation, not raw compute: models are only useful if analysts can translate outputs into actionable coaching and recruitment insights.
- MLS and roster rules create a “portfolio management” problem, where player decisions are constrained by cap charges and roster slots, not just talent.
- Analytics does not replace scouting; it complements it. The best clubs use both human judgment and data, then reconcile differences.
How Soccer Analytics Actually Works
From event data to tracking data
Early soccer models relied mostly on on-ball event data:
- passes
- shots
- tackles
- goals
- coordinates and timestamps
That evolved into tracking data, which records player and ball positions multiple times per second. The next frontier is full body-pose data, which can capture movement patterns and body orientation in much greater detail.
Common models and metrics
The guests discuss several common analytics tools:
- Expected Goals (xG): estimates the likelihood a shot becomes a goal.
- Expected Possession Value (EPV): estimates how much a possession increases scoring probability.
- Expected Threat: measures how dangerous an action is over the next few seconds or next possession.
- Momentum charts: visualize shifts in control over a match.
These metrics are especially useful because soccer has few goals and relatively few discrete events, making the game harder to summarize with basic box-score-style stats.
Why the Trading Comparison Fits
Mike Tracy draws a strong analogy between soccer and volatility arbitrage trading:
- Both involve distribution-based thinking rather than simple outcomes.
- Both require making levered bets under uncertainty.
- Both depend on imperfect signals and require judgment about what is “priced in.”
- In soccer, as in markets, you are often asking:
- What is the expected outcome?
- What is the realized outcome?
- Where is the mismatch between the two?
This is especially relevant in:
- player recruitment
- match preparation
- in-game adjustments
- roster construction in salary-capped leagues
The Limits of the Data
Not every stat is meaningful
The guests are skeptical of some old broadcast-era stats such as:
- possession percentage
- corners
- yellow cards
These can be interesting, but they are not always predictive on their own.
VAR and red cards complicate models
Referee interventions and unusual game states can distort data. For example:
- a red card can radically change the rest of a match
- VAR can nullify a goal after the fact
- a “great” performance in a weird game state may not generalize
A key modeling practice is to censor or adjust for abnormal match conditions so that one odd game does not overly influence player evaluation.
Data must be translated for humans
A major theme is that machine learning models often produce outputs that are not directly interpretable by coaches. The solution is:
- model the game
- identify meaningful patterns
- convert them into video clips or simple tactical recommendations
- let analysts communicate the takeaway to coaches and managers
In other words, the analyst is the translator between machine output and football decision-making.
MLS, Salary Caps, and “Portfolio Management”
Mike Tracy argues that MLS adds a third layer beyond recruitment and team tactics: portfolio management.
Because of:
- salary caps
- designated player slots
- U-22 roster rules
- cap charges rather than raw salaries
…clubs must think like portfolio allocators. A player’s value depends not just on talent, but on:
- where they fit in the roster structure
- how much cap space they consume
- what alternatives exist in that slot
This makes MLS analytically distinct from European leagues, where club wealth and financial power can dominate more directly.
Big Clubs vs. Small Clubs
The episode also addresses a common concern: if analytics becomes more powerful, do richer clubs gain even more of an edge?
The answer is uncertain:
- Yes, larger clubs can afford more data and compute
- But open-source tools and cheaper models can democratize access
- Smaller clubs may still use analytics to punch above their weight
A recurring tension is that the tools are becoming more available, but the hard part remains turning data into insight.
Human Judgment Still Matters
The discussion pushes back against the idea that soccer can be reduced entirely to computation.
Important points:
- Players and coaches often know things the data can’t capture directly, like what they could or could not see on the field.
- Scouts still matter.
- Good clubs do not rely on models alone; they compare model outputs with human observations and investigate mismatches.
The strongest teams, according to the guests, are collaborative environments where:
- scouts evaluate talent
- analysts quantify it
- coaches decide how to apply it
The Bigger Philosophical Point
The episode ends on a broader reflection: if soccer can be modeled increasingly well, what does that say about other complex human systems?
The hosts and guests flirt with a provocative idea:
- maybe many things we treat as “art” are actually very complex but still computable
- the challenge is less “can it be modeled?” and more “what exactly are we optimizing for?”
That question matters because in sports, “success” can mean different things:
- winning matches
- avoiding relegation
- maximizing profit
- entertaining fans
- producing betting edge
Bottom Line
The episode argues that soccer analytics is no longer just a novelty. It is becoming a serious decision-support system, much like quantitative methods in finance. But the core lesson is not that data replaces intuition—it’s that the best results come from combining data, context, and human expertise.
The sport may look fluid, artistic, and chaotic on the surface, but underneath it is increasingly being broken into measurable micro-events. The real challenge is not collecting the data—it’s knowing what to do with it.