Six analyses. Four disciplines. One standard: if it doesn't survive the robustness check, it doesn't get published.
Beyond the Metric applies scientific frameworks from outside traditional sports analytics to questions inside sport. Information theory, control engineering, autonomic physiology, occupational screening science. The tools are borrowed. The questions are original. The methodology is transparent to a fault.
If this is your first visit, here's the map.
The Entropy Series
3 Parts | NHL | 633,318 shots | 8 seasons
Does the predictability of a team's shot locations tell us anything useful about winning hockey games?
We applied Shannon entropy - the same measure Claude Shannon invented to quantify information in telegraph signals - to every 5v5 shot in the NHL across eight seasons. The answer was more complicated than we expected. One finding survived every statistical correction we threw at it. Several others collapsed during robustness testing when we controlled for rink-recording bias and shot volume confounding. The series reports what held, what didn't, and why. It builds on itself.
Part 1: What Shot Entropy Actually Measures — Entropy is a real, stable team trait. It persists year over year (r = +0.57), reflecting genuine organizational choices about offensive structure. Lower entropy predicts higher expected goals per shot. That's the one finding that held. The critical caveat: predicting shot quality is not the same as predicting wins.
Part 2: The Concentration Trap — The teams with the most disciplined shot selection are among the worst in the NHL. Four offensive archetypes emerge. The punchline: volume compensates for shot selection in a way that shot selection cannot compensate for volume. The NHL is a volume game first.
Part 3: The Colorado Problem — The best team in hockey has a perfectly average entropy profile. Rank 11 of 32. Utterly unremarkable. That gap between what entropy predicts and what Colorado actually does is the most instructive result in the entire series.
The Screening Test
NEW | NHL Draft | 4,006 picks | 17 drafts | Occupational Screening Science
The NHL Entry Draft is a screening system. Every June, 32 teams spend millions to rank 200 prospects. Nobody had ever evaluated this decision using the methodology used for occupational screening standards. So we did.
The Screening Test — ROC curves, AUC, sensitivity/specificity, and probability zones applied to 4,006 draft picks from 2000-2016. The draft's AUC for identifying established NHLers is 0.76 - "fair to good," comparable to a cognitive ability test. For stars, it jumps to 0.86. But the draft may be decent at ranking prospects while being much worse at pricing uncertainty. The biggest cliff in draft equity isn't from Round 1 to Round 2. It's from elite to merely first-round.
The Autonomic Frontier
Biathlon | 1,774 shooting bouts | 192 athletes | 8 Olympic events
A different sport. A different science. A finding that runs backwards.
The Autonomic Frontier — At the 2026 Milano Cortina Olympics, the fastest biathlon shooters were also the most accurate. That's the opposite of what motor control theory predicts. The effect holds under five simultaneous statistical controls and within individual athletes across their own bouts. The likely mechanism is autonomic: elite cardiovascular fitness enables faster parasympathetic recovery during the ski-to-range transition. The same physiology that makes them fast makes them stable.
The Restoring Force
NHL | 12,476 episodes | 3 seasons | Control Engineering
What happens in the 180 seconds after a team gets scored on? Not the goals or the highlights. The structural response.
The Restoring Force — We modeled every post-goal episode across three NHL seasons as a damped dynamical system. The finding: good teams and bad teams are indistinguishable for 90 seconds. Then the curves diverge sharply. Nine robustness checks confirm the pattern is real. The practical magnitude is roughly 4 expected goals across an entire season. We report both the significance and the modesty, because that's what honest analysis looks like.
The Metrics Shelf
Reference Guide | 21 metrics | 22,000+ words
Not an analysis. A reference shelf. Something you pull from when you need the full context behind a number.
The Metrics Shelf — Every major metric used across this blog: formal definition, what it measures, what it doesn't, known limitations, and how it connects to the analyses above. Information theory, classical statistics, dynamical systems, physiology, and hockey-specific metrics. Written for anyone who wants to understand what the numbers actually mean before deciding whether to trust them.
New analyses published when they're ready. No schedule, no filler. get them delivered directly.