GOAT SeriesSkaters
Who Is the NHL GOAT? Gretzky, Lemieux and McDavid, Adjusted for Era
Raw points flatter the 1980s, and adjusted points ignore how good the league was. We measured both on 108 seasons of NHL data, then tested how much of a player’s scoring is just opportunity. Here is what holds up, and the one assumption that decides McDavid’s case.
By Edgehalla ResearchPublished 13 min read
The short answer: who is the greatest hockey player of all time?
It depends on whether you mean the best peak or the best career, and on one assumption about how much hockey has improved. Here is where the evidence lands.
Lemieux has the best era-adjusted peak, Gretzky the best career, and McDavid the best peak of the current era.
Our overall pick is Wayne Gretzky. He has the second-best peak and the best career, and no one else is near the top of both lists. Mario Lemieux was the better player at his best, game for game. Connor McDavid is the best player of the modern era and the only active forward in the peak top 10.
The rest of this article shows how we got there, where the data is solid, and where it runs out.
Why raw points, and even adjusted points, mislead
Gretzky’s 212 points in 1981-82 came in a league where each team scored 4.01 goals a game. In 2025-26 that number was 3.08. Raw totals reward the era as much as the player.
Old box scores also counted assists differently. In 1917-18 the league recorded 0.42 assists per goal. Today it records about 1.7. Early playmakers look worse than they were.
Sites like Hockey-Reference fix those two problems. They restate every season in a standard scoring environment. We did the same, using 2025-26 as the reference season.
But that correction has a blind spot. It says nothing about who a player was scoring against. A goal against a six-team league stacked with talent is not the same as a goal against a league that just doubled in size. Every popular McDavid-vs-Gretzky comparison we found stops here.
How strong was each NHL era?
To measure league strength, we followed players whose careers overlap. If the same group of forwards suddenly scores more one season, after accounting for their age, the league probably got easier. Statisticians call this bridging eras. Berry, Reese and Larkey used it in 1999 on forwards up to 1996. We extended it to 2026 and to every skater since 1917.
Expansions show the effect clearly
When the NHL doubled from 6 to 12 teams in 1967, the forwards who stayed scored about 34% more the very next season. Half the jobs went to players who would not have made the old league.
The 1979 merger with the WHA added four teams too. This time there was no jump (+1%). Those players were already NHL-calibre, so the league did not water down. The model can tell an expansion from a merger, which is a good sign it measures something real.
Talent per job tracks league strength
We built a second, independent measure from Statistics Canada birth records. It counts young men in the main hockey-playing age range, adjusts for the share of NHL players born outside Canada, and divides by the number of NHL jobs.
The two measures rise and fall together (r −0.80 after removing trends). The deepest leagues were the Original Six and the late 1980s to early 1990s, when the baby-boom generation and the first wave of European and Soviet players arrived. The thinnest was 1967–1979.
How much easier it was to score than in 2025-26
Established forwards’ scoring vs today, by decade, at the talent-pool setting
The 1970s were the easiest decade to score in (+18%); the Original Six era was the hardest (−18% to −30%); the 1980s were close to today.
One surprise: by talent per job, Gretzky’s 1980s were not a weak league. The baby boom flooded junior hockey with players, and the jobs had not grown to match. The 1970s were the soft decade.
The catch: the data cannot see slow, steady improvement
Here is the honest limit of any study like this. A player’s age, the season and the year he was born are locked together. That means scoring data can spot every jump and dip in league strength, but not a steady, decade-by-decade improvement that lifts everyone at once.
Better coaching, training, nutrition, video and equipment all work that way. They make every player better, so no comparison inside the league can see them. Statisticians call this the age-period-cohort problem. The 1999 study solved it with an assumption about aging, which is the same choice made quietly.
We make the choice in the open. Our default setting assumes the league improved only as fast as its talent pool grew. Then we show what happens if hockey also improved for other reasons, by some percent per decade.
Best three-season peak at different improvement settings
Points per 82 games in 2025-26 terms
At the talent-pool setting Lemieux and Gretzky lead; with 5–10% extra improvement per decade McDavid closes in and passes Gretzky at about 7%.
- Talent pool only
- +5% per decade
- +10% per decade
McDavid barely moves on this chart because his peak is recent. Howe’s 1950s peak moves the most. Past about +13% per decade, the ranking simply rewards whoever played most recently, which is why we treat that as the edge of a sensible range.
NHL GOAT rankings: the best peaks, adjusted for era
We ranked every forward by his best three full seasons in a row. Short, injury-hit seasons do not count, so a 22-game hot streak cannot top the list.
Best three-season peaks: scoring-only vs era-adjusted
Points per 82 games in 2025-26 terms
Adding league strength lifts Lemieux and the Original Six stars, and leaves McDavid where scoring alone puts him.
- Scoring-only
- Era-adjusted
| Rank | Player | Seasons | Era-adjusted pts/82 | Scoring-only rank |
|---|---|---|---|---|
| 1 | Mario Lemieux | 1991-96 | 207 | 2 |
| 2 | Wayne Gretzky | 1983-86 | 184 | 1 |
| 3 | Gordie Howe | 1951-54 | 159 | 23 |
| 4 | Jaromir Jagr | 1997-00 | 157 | 4 |
| 5 | Frank Boucher | 1927-30 | 154 | 15 |
| 6 | Eric Lindros | 1994-97 | 150 | 9 |
| 7 | Jean Beliveau | 1958-61 | 149 | 65 |
| 8 | Connor McDavid | 2020-23 | 145 | 3 |
| 9 | Howie Morenz | 1927-30 | 145 | 22 |
| 10 | Bernie Geoffrion | 1958-61 | 143 | 91 |
How sure are we? We re-ran the whole model 200 times on random resamples of players. Lemieux ranked first in 99% of runs. Gretzky’s middle rank was 2nd (range 1st–4th), Howe’s 4th (1st–9th), and McDavid’s 9th (4th–19th).
Defencemen: Orr, just ahead of Coffey
Among defencemen, Bobby Orr (1969-72) leads at 116 points per 82 games, just ahead of Paul Coffey (113). Cale Makar is 9th (96). Defencemen got league strength from the forwards’ model, because the defence-only model mistook Orr’s own revolution, defencemen joining the attack, for a weaker league.
Wayne Gretzky’s case
Gretzky’s argument is dominance plus time. From 1981-82 to 1986-87 he stood 4.15 to 4.40 standard deviations above the league’s top-six forwards every single season. No other player has topped 4 on all-situations scoring.
How far the scoring leader stood above the top forwards
Average leader z-score among each season’s top-six forwards per team
The 1980s scoring leader averaged nearly 4 standard deviations above the top group; every other decade sat between 2.2 and 3.1.
Could that gap be a weak league in disguise? Our tests say no. The spread among the top forwards was no wider in the 1980s than in other decades. Gretzky was simply further out. And by talent per job, his league was close to as deep as today’s.
Then there is the career. Gretzky has 2,675 era-adjusted points, 268 more than Howe and 447 more than Jaromir Jagr. Even at his career rate, McDavid would need about ten more full seasons to catch him.
| Rank | Player | Games | Era-adjusted points | Per 82 games |
|---|---|---|---|---|
| 1 | Wayne Gretzky | 1487 | 2,675 | 147 |
| 2 | Gordie Howe | 1767 | 2,407 | 112 |
| 3 | Jaromir Jagr | 1733 | 2,228 | 105 |
| 4 | Mark Messier | 1756 | 1,903 | 89 |
| 5 | Ron Francis | 1731 | 1,890 | 90 |
| 6 | Sidney Crosby | 1420 | 1,886 | 109 |
| 9 | Alex Ovechkin | 1573 | 1,809 | 94 |
| 10 | Mario Lemieux | 915 | 1,771 | 159 |
| 43 | Connor McDavid | 794 | 1,290 | 133 |
What about Jari Kurri? Gretzky’s linemates scored a lot, but he did not need them. In the 345 games where one of his regular linemates was out, he kept 96% of his usual even-strength scoring.
Mario Lemieux’s case
Lemieux has the best peak in our data, and it is not close once league strength counts. His 1991-96 stretch is worth 207 points per 82 games in today’s terms. Over his whole career he produced 159 per 82, the highest of anyone with 500 or more games.
He did it in a deep league. The early 1990s had the baby-boom generation at its peak plus the first big wave of Soviet and Czech players. That is why the era adjustment lifts him while barely moving Gretzky.
The 1999 bridging study reached the same order: Lemieux first, Gretzky second, at their peaks. Our version adds 30 more seasons and still agrees.
What holds Lemieux back is games. Back injuries, cancer treatment and retirement left him at 915. His career total ranks 10th. As a peak player he is the best ever. As a career, he is not in Gretzky’s range.
Connor McDavid and the current era
McDavid’s best three seasons (2020-23) are worth 145 points per 82 games. That ranks 8th all-time on our default setting and 3rd on scoring alone. His career rate, 133 per 82, trails only Lemieux and Gretzky.
His best single-season z-score was 3.44 in 2020-21. That is elite, but well short of Gretzky’s run above 4. Today’s top forwards are bunched no tighter than in past decades, so McDavid’s smaller gap is not just a deeper field hiding him.
So McDavid’s GOAT case comes down to one question: has hockey improved faster than its talent pool grew? If coaching, training and equipment added about 7% per decade on top, that is roughly 30% between Gretzky’s peak and McDavid’s, and McDavid passes Gretzky’s peak. At about 13% per decade he passes Lemieux too.
Is that plausible? Goaltending clearly improved far more than talent pools explain (see below). For skaters, we found no direct measure going back to the 1980s. Skating-speed tracking only began in 2021. Reasonable fans can set the dial differently, and that is the honest place to leave it.
One more point for the current era: it has more great players at once. Nikita Kucherov (141) and Nathan MacKinnon sit inside the top 25 peaks. The modern league has a ladder of stars, not one star and everyone else.
How much of scoring is just opportunity?
Put a depth forward on a line with McDavid and Leon Draisaitl, give him power-play time, and his points jump. Vasily Podkolzin is the textbook case.
Vasily Podkolzin: with and away from Edmonton’s stars
Even-strength points per game
Podkolzin scored 0.55 even-strength points a game beside McDavid or Draisaitl in 2024-25, and 0.15 away from them.
- With McDavid or Draisaitl
- Away from them
In 2026-27 he opened on the top line and first power-play unit, with 9 points in his first 6 games. That is a tiny sample, but it fits the pattern.
What 6,054 line vacancies show
We found every game from 2017-18 to 2025-26 where a team’s top forward played without one of his two regular linemates. Using real shift data, we saw who took the empty spot and what happened.
- The fill-in scored 21% more than his usual rate, mostly because he got about 15% more ice time.
- Depth players got the biggest lift: 1.7× their usual even-strength scoring.
- Even so, fill-ins produced only 82% of what the missing regular did.
- Roughly a third of a top-line regular’s scoring comes with the spot. Two thirds is the player.
Ice time explains part of the gap between lines, but not all of it. From 1997-98 to 2025-26, middle-six forwards scored 63% of the top line’s even-strength points per game, but 76% per 60 minutes. The top line also gets about twice the power-play time.
Has that changed over time?
Before 1997 there is no ice-time data. But linemates are on the ice for the same goals, so their game-by-game plus-minus moves together. On modern seasons that signal finds 80% of real linemates and gives the same answers as shift data, so we used it back to 1959.
The promotion bump stayed flat. The gap below the top line grew.
Fill-ins on a star’s line, by era
Moving up a line has always added 2–13%, but since the 1990s the next forward normally scores about 20% less than the regular, where in 1967–92 he matched him.
- Promotion bump
- Fill-in’s usual rate ÷ regular
In the 1970s and 1980s, the next forward up usually scored as much as the star’s regular winger. Points on a star’s line went to whoever played there. Since the 1990s there is a clear ladder: the regular normally out-scores the next player by about 20%.
That matters for the GOAT debate in two ways. First, stars carried their lines in every era: Gretzky (96%), Lemieux (103%), McDavid (98%) and Sidney Crosby (104%) all kept their scoring without a regular linemate. Second, Gretzky’s linemates’ big totals owed a lot to the spot, so they say little about the quality of help he had.
Stars without a regular linemate
Even-strength scoring when a regular linemate was out, as % of normal
Every star kept 96–104% of his scoring without a regular linemate, so linemates do not explain their numbers.
Is the floor rising? For goalies, yes
A popular argument says the modern player is 1 standard deviation above a much better league, so he must be better than an old player 2 standard deviations above a weaker one. Biologist Stephen Jay Gould made the classic version about baseball’s vanished .400 hitters: as everyone gets better, the gap between best and worst shrinks.
In hockey, that is clearly true for goalies. Every goalie does the same job, so we could split each season’s save-percentage gaps into skill and luck. The skill share fell from 66% in the 1970s to 30% in the 2010s. The worst starters caught up to the best.
How much of goalies’ save-percentage gaps is skill
Share of the spread among goalies with 800+ shots that is true talent, not luck
Goalie skill differences shrank from two-thirds of the spread in the 1970s to under a third in the 2010s.
That has a twist for Gretzky. The 1980s had a wide spread of goalie talent, so there were more weak goalies to score on. A standard scoring adjustment does not fully capture that, and it is the next thing we will test.
For skaters, we found no rising floor. The spread among top forwards has not narrowed since the 1940s, and the gap between top-line regulars and their replacements grew. Gould’s idea fits goaltending. It does not explain away Gretzky.
The verdict
- Wayne Gretzky is the GOAT. He has the best career by 268 points and the second-best peak, and his dominance survives every adjustment we tried.
- Mario Lemieux had the best peak, first in 99% of our resamples, and the best scoring rate of any long career. Injuries and illness cost him the career crown.
- Connor McDavid is the best player of his era and third-best by career rate. Whether his peak beats Gretzky’s depends on how much hockey improved beyond its talent pool. If you think it was more than about 7% per decade, McDavid has the case.
This article only measures scoring. It does not count defence, faceoffs or playoff results, and points before 1955 carry more uncertainty. Every table behind it is in the “Show the work” section below.
Show the work
- Tests written down
- Not in advance (exploratory)
- Results computed
Every NHL regular season from 1917-18 to 2025-26 (47,644 player-seasons). Points restated for goals per game, assists per goal and skaters dressed; league strength from a player + age + season model on forwards with 20+ games; the long-run trend anchored to Canadian births per NHL job (1955–2025). Holdout: five folds of player-seasons. Uncertainty: 200 resamples of players. Shift-based lines from 2017-18; box-score linemates before that.
| Test | Result | Verdict |
|---|---|---|
| Season effects predict unseen player-seasons | Squared error cut 1.8–2.6% in all 5 folds (n 22,851) | Real, small |
| Expansion waters down the league | 1967: +34%; elasticity 0.36 (90%: 0.25–0.47); WHA merger +1% | Real |
| Talent per job tracks league strength | r −0.80 after removing trends (1955–2025) | Supports the anchor |
| Top forwards are bunching tighter (Gould) | +0.005 log-SD per decade since 1942; EV +0.003 | Not supported |
| Goalie talent spread is shrinking | True SD .0125 (1970s) → .0050 (2010s) | Supported |
| Lemieux has the best three-season peak | Rank 1 in 99% of 200 resamples | Robust |
| Box-score linemates match shift data | 80% of true linemates; retention 0.89 vs 0.92 | Validated (2017–2026) |
| Gap below the top line has grown | −4.8% per decade (90%: −6.2% to −3.2%); bump +1.0% | Real |
| Stars depend on their linemates | Gretzky 96%, Lemieux 103%, McDavid 98% of normal | No |
| McDavid’s peak beats Gretzky’s | Only above +7% extra improvement per decade | Depends on an untestable assumption |
Download the result tables
- Peak rankings (CSV): Top 60 three-season peaks, era-adjusted and scoring-only.
- Career totals (CSV): Top 25 era-adjusted careers for forwards.
- League strength by season (CSV): Season effects, expansion jumps, talent per job and Canadian share, 1917–2025.
- The improvement dial (CSV): Peaks and ranks for key players at −20% to +30% extra improvement per decade.
- Fill-ins by era (CSV): Retention, promotion bump and gap for line vacancies, 1959–2026.
- Goalie talent spread (CSV): True-talent spread of save percentage by season, 1955–2025.
Aggregated results only, free to reuse with credit (CC BY 4.0).
Frequently asked questions
Who is the greatest hockey player of all time?
On era-adjusted scoring, Wayne Gretzky. He has the most era-adjusted career points (2,675) and the second-best three-season peak. Mario Lemieux had the best peak (207 points per 82 games in today’s terms), but played only 915 games.
Is Connor McDavid better than Wayne Gretzky?
Not on the evidence we can measure. McDavid’s best three seasons are worth 145 points per 82 games after adjusting for era, against Gretzky’s 184. McDavid passes Gretzky only if hockey improved about 7% per decade beyond the growth of its talent pool, which no data can confirm yet.
Was Mario Lemieux better than Wayne Gretzky?
At his peak, yes. Lemieux’s 1991-96 seasons rank first in 99% of our resamples, and his career scoring rate (159 per 82 games) beats Gretzky’s (147). Gretzky’s career is far longer and more valuable overall.
Was the 1980s NHL a weak league?
Not by talent per job. The baby-boom generation filled junior hockey, so the 1980s were close to as deep as today. The weak decade was the 1970s, after the NHL doubled to 12 teams and then competed with the WHA.
What does “era-adjusted” mean in hockey?
It restates a player’s points in another season’s terms. Most versions only adjust for goals per game and how assists were counted. Ours also adjusts for league strength, measured by following players whose careers overlap and by the size of the talent pool per NHL job.
Sources and method
- NHL stats (NHL.com), every season since 1917-18
- NHL API reference (community documentation)
- Berry, Reese & Larkey (1999), “Bridging Different Eras in Sports,” Journal of the American Statistical Association 94(447)
- Hockey-Reference: how adjusted stats are calculated
- Stephen Jay Gould, Full House (1996): why .400 hitting vanished
- Statistics Canada, Historical Statistics of Canada, live births 1921–1974 (B1-14)
- Statistics Canada table 13-10-0415-01, live births by month
- Age-period-cohort analysis (why the long-run trend is not identified)
- Bootstrap resampling
Research and analysis, not betting advice. Past results do not guarantee future returns; bet only what you can afford to lose.