This is an exploratory study of NHL scoring in the last three completed seasons, covering 2023/24 through 2025/26. It describes what happened. It does not measure mental toughness or offer a validated prediction of the next game.
We included all 3,936 regular season games and 256 playoff games. We excluded preseason games, international competition and games that were scheduled but never played. The player comparisons cover skaters and their scoring contributions. Goals, primary assists and secondary assists are reported separately.
A goal qualifies when it is scored in the third period with five minutes or less remaining and the scoring team is behind by exactly one or two goals immediately before the goal. The boundary at exactly five minutes remaining is included. We credit the scorer and recorded assistants. A goal that reduces a deficit from two to one counts even if the team eventually loses. A goal scored while tied does not count.
We reconstructed the score before each goal. Regular season and playoff results are kept separate, with an explicitly labelled combined table. Games in which the scoring team has pulled its goalie remain eligible. Leading teams scoring into an empty net fail the deficit condition.
Overtime means an event recorded in an overtime period, including power plays. Shootout goals and attempts are excluded. Regular season overtime and playoff overtime are analyzed separately. All playoff games count toward postseason totals, including blowouts.
A player faces elimination when the opposing team already has three wins in the series before that game. Both teams qualify for Game 7. A team with an opportunity to eliminate its opponent does not automatically qualify. The Stanley Cup Final is also shown separately. These categories overlap and must not be added together.
The source is the NHL event and shift records stored in the BetProps database, retrieved on 18 September 2026. All 25,597 regulation and overtime goals reconcile to the final scores of all 4,192 games after removing the bookkeeping goal awarded to shootout winners. All credited player identities resolve. We normalized the historical Utah team identifier when matching events and shifts to games.
We also checked McDavid’s 42 playoff points in 2024 and Draisaitl’s four playoff overtime goals in 2025 against NHL reporting. Final score reconciliation checks goal coverage. It does not independently verify every assist or timestamp against a fresh NHL feed.
The scoring rate charts report points per 60 minutes of eligible ice time. The calculation is points divided by eligible seconds, multiplied by 3,600. The playoff baseline chart divides two such rates to produce a ratio with no units. League ranks quoted in the article are taken from the same rate file, restricted to skaters with at least 60 minutes of eligible ice time in that category. Hertl leads the late comeback rate at every minimum from 60 through 120 minutes.
The main scoring charts compare the same selected players in the rate and total views. They are not league rankings by rate. Shift based rates use covered games only, while totals include all completed games. Minutes and points in each rate use the same retained games. Select Total points in the chart controls to see the original scoring counts.
For the late comeback rate, we count only the seconds a player is on the ice during the final five minutes while their team trails by one or two goals. The eligible clock stops when the player leaves the ice or the score moves outside those conditions. Overlapping shifts are merged.
Some games marked as synced had no usable shift records. There were 562 such regular season games and 20 playoff games. Those gaps were concentrated in the later seasons, so we cannot assume they are random. Full scoring totals still include every completed game.
Rate calculations exclude inadequate coverage games from both points and minutes. We require both teams to have at least 1,200 recorded player seconds within the final five minutes. For full playoff games the minimum is four player seconds per elapsed second in each regulation period. For overtime it is two in the regular season and four in the playoffs. We also exclude a category within a game if a credited skater has points but no recorded exposure. These are coverage screens rather than a complete reconciliation of manpower.
For late comeback rates, 3,374 regular season games pass the coverage screen. Each player’s rate then uses only games where they have positive eligible exposure. Hertl’s rate sample has 14 points in 128.2 minutes. McDavid’s has ten in 121.1 minutes. Their complete scoring totals are 15 and 12 respectively. The empty net counts in the article’s table are taken from complete scoring totals rather than the covered games behind the rates, so the two columns do not describe an identical set of games.
We resample each selected player’s eligible games with replacement 10,000 times, keeping the points and minutes from each game together. We calculate a scoring rate per 60 minutes for each sample and show the middle 95 percent of those rates. Games with eligible ice time and zero points are included. The random seed is fixed at 20260918.
These approximate bootstrap intervals treat games as exchangeable. They do not correct for missing records, changes in ability, shared linemates, opponents or choosing interesting players after seeing the results. They are not a formal test of who is best. Overlapping intervals alone are not a test of differences.
We counted how often McDavid and Draisaitl received a goal or assist on the same goal. Each goal is counted once within each situation. Categories overlap, so their counts must not be summed. These counts use all scoring events in the relevant category, independently of shift coverage. The summary and shared goal audit identify their credited scoring connection, not causal attribution. We have not fitted a model of performance together and apart. A point without credit to the other player does not establish that the other player was off the ice.
We divide each player’s playoff points per 60 minutes by his regular season points per 60 minutes, pooling the three seasons separately in each competition. A ratio of 1 means unchanged production. A ratio of 1.20 means 20 percent higher playoff production. This measures change relative to the player’s own baseline, rather than absolute ability or a mental response to pressure.
The seven players are the same as in the preceding playoff chart. Both denominators come from all situation player game totals, rather than the partial regular season shifts extracted for late comebacks and overtime. All 1,815 player appearances have positive recorded ice time, no duplicate totals, and points that agree with the goal events. Their game totals can differ slightly from merged shift durations in the earlier charts. We did not substitute full points over partial shift minutes.
Within each competition, we resample a player’s games with replacement 10,000 times, preserving each game’s points and minutes. We divide the resampled playoff rate by the independently resampled regular season rate and show the middle 95 percent of ratios. Each player has a reproducible random seed of 20260918 plus his NHL identifier. Zero point games are retained. All seven approximate intervals cross 1. This is not a test of differences between players.
The calculation pools seasons, so regular season and playoff minutes can have different season weights. As a sensitivity check, we also calculate expected playoff points using each season’s regular season rate and that season’s playoff minutes. Dividing actual points by that expectation leaves the same seven players on the same side of 1. Both ratios, the underlying rates and game counts are recorded in the baseline results.
Players without playoff minutes have no ratio. A zero regular season scoring rate would also make the ratio undefined. The bootstrap treats games as exchangeable and does not adjust for playoff selection, serial dependence, stronger opposition, checking assignments, injuries, power play usage or shared linemates. Being below 1 does not establish poor performance after accounting for those factors. The line at 1 is a descriptive reference, not a prediction of what a player should produce against playoff opposition.
The closing scatterplot extends the full-game comparison to every skater with at least ten playoff appearances and at least 60 recorded minutes in each competition across 2023/24–2025/26. Both axes use all-situation game points divided by game ice time, multiplied by 3,600. We retain zero-point appearances and pool all three seasons in each competition, including regular seasons in which a player missed the playoffs. We check for duplicate player-game totals, missing appearances, missing ice time and disagreement with the original goal-event credits before generating the plot.
The diagonal indicates equal scoring rates, not an opponent-adjusted expectation. A high position on the chart describes a high scoring rate; distance above the diagonal describes improvement relative to that player’s regular-season rate. The chart does not estimate uncertainty or establish a statistically significant gap between nearby players. Playoff selection, season weights, positions, teammates, opponents and usage remain unadjusted.
The comparison is inspired by Data Golf’s Major Performers. Unlike their true strokes-gained measure, these hockey rates are not adjusted for competition. Their age and major-participation restrictions are not applied here. The complete plotted player data and validation summary were verified before generating the plot.
The article checks the last three minutes, deficits of one goal only, separate seasons and goals compared with assists. The featured playoff comparisons include Matthews with 18 appearances and display the sample sizes. They do not claim to be complete league rankings by rate. There is no combined clutch score.
Of the 669 qualifying regular season comeback goals, 425 occurred with the scoring team’s goalie absent. That is taken from the manpower record at the moment of the goal. It includes possible delayed penalties and does not mean every goal was scored with six skaters against five. We compare playoff production with each player’s regular season scoring rate, but have not adjusted for strength, teammates or opponent quality.
The broad approach draws on DataGolf’s study of performance under pressure. A future study could estimate expected scoring using earlier player performance and game context, then test the model on a later season. That work has not been completed here.
Player photographs come from the NHL. Jerseys reflect the photograph season rather than every team a player represented during the study.