Simulation and Visualization Unlock Sports Analytics' Hidden Advantages

The subtle art of simulation and visualization in sports analytics reveals how understanding complex systems can unlock hidden advantages, even when the immediate path forward seems clear.

This conversation with Micah McCurdy, a mathematician and creator of HockeyViz, delves beyond surface-level statistics to explore the deeper implications of modeling sports performance. McCurdy’s work highlights how visualizing data, rather than just crunching numbers, is crucial for genuine understanding. The core implication is that true insight comes not from predicting the obvious, but from building models that capture the slow-moving, fundamental abilities of individuals and the intricate physics of the game. This podcast is essential for anyone in sports analytics, data science, or even general management who seeks to move beyond conventional wisdom and leverage sophisticated modeling for competitive advantage. It offers a roadmap for understanding how to build predictive power by focusing on the underlying mechanics of performance, rather than just recent outcomes.

The Physics of Player Ability: Why Teams Evolve Slowly

The conversation with Micah McCurdy underscores a fundamental principle: player ability, the bedrock of sports performance, changes remarkably slowly. This contrasts with the common tendency to overreact to recent results, a cognitive trap that can lead to flawed strategic decisions. McCurdy’s approach, rooted in a physicist's mindset, treats player capabilities as stable entities, with changes occurring over long timescales. This perspective is critical for understanding how to build robust predictive models.

When McCurdy simulates hockey games, he begins with an assessment of player abilities and coaching effectiveness, treating these as the primary drivers of outcomes. The immediate results of a few games, or even a series, weigh surprisingly little in his simulations. This is because, from a systems thinking perspective, a team’s underlying strength is a composite of individual talents that don't fluctuate wildly from one game to the next. The implication is that strategies based on short-term momentum or recent wins, while appealing, often fail to account for the deeper, more persistent qualities that determine long-term success.

"The simulations that I'm doing, I'm, I'm starting from this is how good I think the people that the Flyers are going to put on the ice are, as well as this is how good I think their coaches, this is how good I think anybody else who makes a relevant decision for on ice stuff. And so I try not to include anything about laundry, if you like, anything about the teams themselves."

This quote reveals a crucial distinction: McCurdy separates the intrinsic quality of players and coaches from the extrinsic factors of team performance, like recent wins or losses. By focusing on these slowly evolving primitives, his models can offer more durable insights. This is where delayed payoffs create competitive advantage. Teams that embrace this perspective can make more stable, long-term talent assessments and strategic decisions, while competitors chasing fleeting momentum might be making decisions based on noise rather than signal. Conventional wisdom, which often emphasizes recent performance, fails when extended forward because it doesn't account for the inertia of individual skill.

Simulation as a Craft: Bootstrapping Complexity

The discussion around simulation highlights its power not just as a mathematical tool, but as a craft that allows for the iterative building of complex models. McCurdy’s journey from a personal hobby to a sophisticated modeling approach illustrates how simulation can bootstrap understanding, even without starting from absolute first principles.

"And so it's funny, I left physics behind, so I thought at the age of 22, only to recover it, 10, now 28 years later. And part of it was the sort of fun of just saying, well, here's a little gadget, what happens when you, you know, it has a kind of a crafty aspect to it. It's, I, you know, you don't have to convince me about the joy of figuring something out from first principles and then convincing somebody else of it, you know, just on a sheet of paper or just on a whiteboard. Like that kind of thing, that's really satisfying. But it doesn't have that kind of like, oh, here's a little thing, what does it do? And so, so simulation was part of why I chose that project as like a fun, fun project."

This points to the value of simulation in exploring system dynamics. By treating the game as a series of interacting components--players, coaches, and the game's physics--McCurdy can simulate outcomes at a granular level, from shift-by-shift play to the entire series. This "simulation all the way down" approach allows for the coherent modeling of various elements, from player interactions to game flow. The implication is that complex systems, like a sports season, can be understood by building up from fundamental interactions, rather than relying solely on aggregate statistics. This method allows for answering a multitude of questions from a single, coherent model, revealing deeper insights than isolated analyses.

The Moving Elevator Problem: Aging and Teammate Effects

A critical challenge in sports analytics is disentangling the effects of player aging from the influence of teammates and opponents. McCurdy’s work on “moving elevators” illustrates this complex interplay, where a player’s performance can be masked or amplified by the abilities of those around them.

He notes the difficulty of isolating aging curves when players are routinely paired with teammates who are either improving or declining. For instance, pairing a declining veteran with a rising superstar can create misleading performance metrics. A naive analysis might attribute the veteran's struggles solely to age, when in reality, the younger player's ascent might be obscuring the veteran's continued, albeit diminished, contribution. Conversely, a young player paired with a declining star might appear to be improving more than they actually are, simply because the veteran is no longer a significant drag on their performance.

"And so I think, I think at last I have figured out a way that I can start to integrate the aging curve measurements into the model measurements. And so, uh, we'll see if that works. Maybe you can have me back on in a month or two."

This highlights the ongoing nature of research and the difficulty in solving these complex interactions. McCurdy’s pursuit of a regression technique to integrate these factors simultaneously suggests a path toward more accurate player evaluations. The conventional approach often treats these factors separately, leading to potential biases. By tackling the “moving elevators problem,” McCurdy aims to provide a more nuanced understanding of player development and decline, which can inform team building and player management strategies. This requires patience and a willingness to tackle problems that conventional wisdom might deem too difficult or too time-consuming to solve, creating a potential competitive advantage for those who can master it.

The Distribution of Winning: Colorado's Dominance and a Field of Average

McCurdy’s probabilistic assessment of the Stanley Cup playoffs offers a stark illustration of how systems thinking can reveal unexpected distributions of winning probabilities. His analysis highlights a significant deviation from a uniform distribution, where one team, Colorado, stands out with an unusually high chance of winning.

The probability assigned to Colorado (around 37%) is notably higher than one might expect in a typical playoff field, where 16 teams compete. This elevated probability is not simply an artifact of a strong team; it’s a reflection of the underlying quality of the team’s primitives--individual player abilities and coaching. McCurdy emphasizes that this assessment is based on the core components of the team, such as the exceptional play of individuals like Nathan MacKinnon and Cale Makar, and strong goaltending.

"Yes, that's, and that's part of why I don't, I don't feel bad about saying so, even though, you know, for six, seven, eight years in a row, I would have said, oh, that's a mistake. 37%, you know, that's part too high for a single team. But this year, I feel a lot happier about it because I've been watching the, the primitives, if you like, about, you know, Nathan MacKinnon, um, Cale Makar, the, the Wedgewood having a tremendous year, you know, a lot of just really high quality skaters."

This year’s distribution is further skewed by a perceived lack of truly dominant teams in the rest of the field. McCurdy notes that many teams are “middling,” making the top teams appear even stronger by comparison. This creates a situation where a smaller number of teams would need to be selected to cover 50% of the winning probability, compared to previous years. This insight is valuable because it moves beyond simply identifying the favorite; it quantifies the degree of separation between the top contenders and the rest of the league, offering a strategic advantage to those who can accurately assess these distributions. It also challenges the conventional notion that playoff parity is always high, revealing instances where specific team compositions can create significant advantages.

Key Action Items:

• Embrace Slow-Moving Player Abilities: When evaluating talent, prioritize long-term skill assessment over short-term performance fluctuations. This requires developing frameworks to measure intrinsic player value that are resistant to game-to-game noise. (Immediate action, pays off in 6-12 months)
• Develop Granular Simulation Models: Invest in building simulation capabilities that model game dynamics at a detailed level (e.g., shift-by-shift) to capture complex interactions and emergent properties. This is a longer-term investment in analytical infrastructure. (Investment over 12-18 months)
• Isolate Aging and Teammate Effects: Actively develop methodologies to disentangle player aging from the impact of their teammates. This may involve specialized regression techniques or advanced statistical modeling. (Requires dedicated R&D effort, pays off in 18-24 months)
• Quantify Winning Probability Distributions: Move beyond simply identifying a favorite to estimating the probability distribution of outcomes for all competitors. This provides a more nuanced understanding of team strengths and potential playoff matchups. (Immediate application for current season analysis)
• Prioritize Visualization for Understanding: Integrate data visualization as a core part of the analytical process, not just for communication, but for debugging and gaining intuitive understanding of complex data. (Immediate integration into workflow)
• Focus on Underlying Primitives: When assessing team strength, concentrate on the fundamental components of performance (player skills, coaching) rather than relying solely on aggregated statistics or recent results. (Shift in analytical mindset, immediate)
• Resist Overreacting to Short-Term Momentum: Implement checks and balances within decision-making processes to prevent impulsive strategic shifts based on short-term winning or losing streaks. (Requires organizational discipline, ongoing)