UNDERSTANDING PUBLIC MOMENTUM AFTER KEY NHL MOMENTS
Every hockey fan understands the indescribable feeling of following the buzz after incredible, season-defining moments: whether it's an overtime winner, a hat trick, a four-point night, or an incredible 3rd-period comeback. But, behind all of that, I wanted to understand:
What exactly is the pattern of that momentum across the league and how much of it is harnessable?
As I became more fascinated with the business behind the NHL, I was intrigued by the question of: could the data tell us which hockey moments sell the most tickets, move the most merchandise, create the most sponsor value, or even change fan behavior?
But, the useful data to that question is unfortunately not publicly available - it's housed in proprietary ticketing systems, commerce platforms, partnership agreements, and club-specific databases. So instead, I pivoted and found the next best proxy of publicly available data - the public attention that follows those moments.
So I changed the question instead of stretching the evidence. Public data can show when people seek information and when a club enters the news conversation. It can show whether that response arrives immediately, lingers for a few days, or disappears. That is narrower than commercial value—and far more defensible.
NHL schedules, box scores, play-by-play, rosters, and standings are used to establish the game, clock, score, players, and opponent. MoneyPuck is used to add expected-goal and game-state context.
Each raw response is retained with its source URL, retrieval time, and checksum.
An overtime win is defined by the final game state; a hat trick by three goals; and a comeback by a club trailing by at least two goals at the start of the third period and winning the game. The rules are applied across the full archive, so memorable examples are not selected after attention data has been observed.
A new moment type is admitted only after a written rule, version increment, complete rebuild, and test case have been supplied.
Names, relocations, and dates are kept as separate dimensions. Wikidata is used to connect each club and player to the appropriate Wikipedia entity, including redirects and aliases. Arizona and Utah are treated as separate franchises and Vegas, Seattle, Utah enter the model in their respective NHL entry seasons.
All event times are stored in UTC; local time is derived separately.
For every qualifying moment, the prior 14 days are used to establish the baseline of each club’s level of public attention.
A value of \( +0.20 \) is interpreted as a public signal 20% above its recent baseline. It is not interpreted as 20% more people or 20% more sales.
Immediate response is assessed over the first two days. Short persistence is assessed on days two and three. Sustained attention is assessed over days four through seven.
Information seeking is represented by daily Wikipedia page visits. Earned-news presence is represented by GDELT coverage after exact-name matching and a manual precision audit. The measures are analyzed separately because they describe different forms of public behavior.
Neither measure should be interpreted as sentiment, readership, or purchase intent.
Because games occur under different conditions, the league-wide partial-pooling model is adjusted for home or road, opponent, standings context, day of week, month, season, and each club’s usual level.
When another major moment occurs within seven days, the following major event is excluded from the isolated before-and-after comparison, but it is retained in the daily analysis so that both moment windows can be modeled together.
At least 10 comparable events are required for each source, and all evidence thresholds must be cleared. The adjusted result and the club’s raw history must indicate the same direction, while their uncertainty ranges must remain entirely on one side of no change. Where any condition fails, the result is reported as mixed or not yet clear.
Alternative 7-day and 21-day baselines are also checked.
We can get a clearer sense of when a big on-ice moment is most likely to capture and hold the public's attention. This can inform the best strategies for when to put out content, when to ride the wave, and when to pull back to not saturate fans with uninteresting content. Of course, I'm sure there are excellent internal intelligence tools that each club uses, but this was my attempt at recreating some version of that insight just using public data.
Lead with the moment itself, then explain what made it matter.
Give people a reason to come back with context, a player perspective, or a simple breakdown—not just the same highlight again.
Look at the public signals alongside the club’s own measures before deciding whether the format is worth repeating.