Those are related, but distinct, questions. A factor can influence performance without offering a betting advantage if the market has already incorporated it into the odds.
This project investigated both possibilities using the 2024–25 NHL season.
02 / INDEPENDENT RESEARCH · SPORTS ANALYTICS
Python · NHL · Statistical modelling
Behind every NHL game is a demanding schedule of travel, recovery, and adaptation. While the action unfolds on the ice, some of the factors that might influence performance are far less visible.
My interest in this question sits at the intersection of human performance, analytical research, and betting markets. I wanted to explore whether the demands of an NHL schedule could be translated into measurable indicators of performance — and whether those indicators might reveal anything beyond what betting markets already anticipate.
Those are related, but distinct, questions. A factor can influence performance without offering a betting advantage if the market has already incorporated it into the odds.
This project investigated both possibilities using the 2024–25 NHL season.
This research required a multi-stage analytical approach, connecting reconstructed travel demands with on-ice performance and betting-market expectations.
I combined historical NHL schedules, game locations, and sequences of games to estimate travel distance, time zone changes, and recovery opportunities. Because actual itineraries and individual sleep data were not consistently available, the circadian measures are modelled estimates, not direct observations of players' biological clocks. The reconstruction included assumptions about travel timing and gradual adaptation with alternative adaptation rates tested for sensitivity.
I examined expected goal differential, shot differential, and goal differential using regression models that accounted for team and opponent differences, venue, rest, and scheduling factors. The aim was to test whether travel-related variables provided a consistent signal beyond more obvious explanations of performance.
I then examined moneyline, regulation tie, and puck line markets. Market calibration simulations and regressions incorporating market implied probabilities helped distinguish potential performance relationships from evidence of market mispricing.
The analysis did not identify a dependable travel related performance signal or betting market advantage.
Across the team performance models, travel distance and estimated circadian disruption did not show a consistent, statistically persuasive relationship with the outcomes examined. The market analyses likewise did not establish reliable mispricing attributable to these factors.
This does not establish that travel never affects performance or that betting markets are perfectly efficient. It means that within one season, the available data, and the assumptions tested, the research did not support a repeatable advantage.
What made this project especially valuable was the challenge of turning a plausible human performance mechanism into variables that could be tested responsibly. That required reconstructing data, making assumptions explicit, evaluating alternative specifications, and separating an interesting hypothesis from the evidence needed to support it.
For me, that is an important part of applied research: bringing behavioural and physiological questions into quantitative analysis, then asking whether the findings can genuinely improve real-world decisions. An inconclusive result is still informative when the process is rigorous and the limitations are clear.
For those interested in exploring the research in greater detail, the complete project is available on GitHub, including the methodology, data preparation, statistical analysis, visualisations, and technical documentation.