How Hockey ChangedThe Vanishing Game

Hainsworth’s 22 Was Mostly Arithmetic

George Hainsworth’s 22 shutouts in 1928-29 have stood for almost a century. We set every NHL season since 1917 beside the shutouts its scoring alone would produce. Shutouts rise and fall with goals. A goalie’s extra shutouts vanish the next season. And since 1980, empty nets aside, teams get blanked less often than their scoring says.

Data story

By Edgehalla ResearchPublished November 10, 202610 min read

Post on X

On March 14, 1929, George Hainsworth shut out the Montreal Maroons. It was his 22nd shutout in 44 games, and six of them had ended 0-0. No goalie since has had more than 15.

He did it “while playing with a broken nose sustained early that season in a pre-game warm-up,” The Hockey News wrote. He also did it in the lowest-scoring season the NHL has ever had.

The hockey historian Eric Zweig told The Hockey News that Hainsworth’s “numbers are way better, even in a crazy low year.” Zweig is right on both counts. The data can say how much of the 22 belongs to each.

Use the toolNHL goalie hubStarters, save percentage and workload for every goalie.

On this page

An ordinary goalie would have had about half of Hainsworth’s 22

So was Hainsworth’s 22 him, or his season? A shutout looks like the purest goalie number there is. One man, one game, nothing allowed. Over a century of NHL games, though, it behaves like a scoring number with a goalie’s name on it. When goals dry up, everybody’s shutouts pile up. And when a goalie gets more shutouts than his goals-against average says he should, the extras do not come back the next season.

Start with 1928-29. NHL teams scored 1.46 goals a game, the fewest of any season. The league’s ten teams were shut out 120 times in 220 games, about once every four games for each team, and 15 games ended 0-0. Hainsworth’s Canadiens went 22-7-15 and scored only 71 goals all season.

Now take the 22 apart. Put an average goalie in Hainsworth’s crease that season, shut out at the league’s rate, and he gets about 12 in 44 starts. Hainsworth allowed only 0.92 goals a game. Some of that was him and some was the Canadiens’ defence in front of him. A goalie who allows that few would normally finish with about 18 shutouts. He had 22. Some of it was him. He beat even that number by about four, the 16th-biggest margin among the 2,064 seasons in which a goalie made 30 or more starts. Still, a little over half of the 22 belonged to the season. And the year before, the same Hainsworth with the same team had only 13 shutouts, about five fewer than his goals-against average would have predicted.

Shutouts peaked in Hainsworth’s season and have followed scoring ever since

Shutouts out of every 100 games a team played, every NHL regular season since 1917-18, with the number each season’s scoring predicts

per 100

Forward passing in all zonesFewest shutoutsMost shutouts (192)Shootout begins

0102030Shutouts per 100 team-gamesPredicted by scoring and team strength

Show data

Shutouts peaked at 27.3 per 100 team-games in 1928-29 and bottomed out at 1.73 in 1981-82; the number predicted from scoring tracks almost every turn.
Season	Shutouts per 100 team-games	Predicted by scoring and team strength
1917-18	2.9 per 100	1.2 per 100
1918-19	5.6 per 100	4.2 per 100
1919-20	5.2 per 100	3.0 per 100
1920-21	4.2 per 100	2.3 per 100
1921-22	2.1 per 100	2.7 per 100
1922-23	7.3 per 100	6.4 per 100
1923-24	8.3 per 100	9.7 per 100
1924-25	12.8 per 100	12.4 per 100
1925-26	15.1 per 100	12.8 per 100
1926-27	19.3 per 100	15.4 per 100
1927-28	22.3 per 100	18.4 per 100
1928-29	27.3 per 100	26.3 per 100
1929-30	5.9 per 100	7.0 per 100
1930-31	13.4 per 100	12.5 per 100
1931-32	12.2 per 100	10.5 per 100
1932-33	13.7 per 100	12.2 per 100
1933-34	12.3 per 100	10.8 per 100
1934-35	11.8 per 100	9.7 per 100
1935-36	13.8 per 100	12.9 per 100
1936-37	9.4 per 100	9.2 per 100
1937-38	9.4 per 100	9.7 per 100
1938-39	11.9 per 100	10.4 per 100
1939-40	10.1 per 100	11.8 per 100
1940-41	6.9 per 100	9.1 per 100
1941-42	6.0 per 100	5.5 per 100
1942-43	3.7 per 100	4.0 per 100
1943-44	4.3 per 100	4.0 per 100
1944-45	3.7 per 100	4.4 per 100
1945-46	4.3 per 100	4.4 per 100
1946-47	5.6 per 100	5.5 per 100
1947-48	6.4 per 100	6.2 per 100
1948-49	8.3 per 100	8.1 per 100
1949-50	9.8 per 100	7.8 per 100
1950-51	8.8 per 100	8.8 per 100
1951-52	7.9 per 100	9.1 per 100
1952-53	11.2 per 100	10.8 per 100
1953-54	13.1 per 100	11.8 per 100
1954-55	8.6 per 100	10.4 per 100
1955-56	10.2 per 100	10.0 per 100
1956-57	7.1 per 100	8.1 per 100
1957-58	8.8 per 100	6.9 per 100
1958-59	6.7 per 100	6.3 per 100
1959-60	6.0 per 100	6.4 per 100
1960-61	4.5 per 100	6.1 per 100
1961-62	6.9 per 100	6.7 per 100
1962-63	4.5 per 100	5.9 per 100
1963-64	9.5 per 100	7.3 per 100
1964-65	6.4 per 100	6.8 per 100
1965-66	6.2 per 100	6.2 per 100
1966-67	6.9 per 100	6.4 per 100
1967-68	8.5 per 100	7.4 per 100
1968-69	6.7 per 100	6.5 per 100
1969-70	7.6 per 100	6.8 per 100
1970-71	5.7 per 100	6.4 per 100
1971-72	5.8 per 100	6.5 per 100
1972-73	5.9 per 100	5.5 per 100
1973-74	5.1 per 100	5.8 per 100
1974-75	6.0 per 100	5.4 per 100
1975-76	4.5 per 100	5.0 per 100
1976-77	5.5 per 100	5.0 per 100
1977-78	4.7 per 100	5.2 per 100
1978-79	3.8 per 100	3.9 per 100
1979-80	3.5 per 100	3.7 per 100
1980-81	2.3 per 100	2.8 per 100
1981-82	1.7 per 100	2.5 per 100
1982-83	2.7 per 100	3.0 per 100
1983-84	2.7 per 100	2.6 per 100
1984-85	2.4 per 100	2.7 per 100
1985-86	3.0 per 100	2.4 per 100
1986-87	1.8 per 100	2.9 per 100
1987-88	3.5 per 100	2.9 per 100
1988-89	2.6 per 100	2.9 per 100
1989-90	2.4 per 100	3.1 per 100
1990-91	3.0 per 100	3.8 per 100
1991-92	3.6 per 100	3.8 per 100
1992-93	3.4 per 100	3.5 per 100
1993-94	4.5 per 100	4.6 per 100
1994-95	4.8 per 100	6.0 per 100
1995-96	4.6 per 100	5.3 per 100
1996-97	6.0 per 100	6.0 per 100
1997-98	7.5 per 100	7.9 per 100
1998-99	7.2 per 100	7.9 per 100
1999-00	5.6 per 100	7.2 per 100
2000-01	7.6 per 100	7.3 per 100
2001-02	7.5 per 100	8.0 per 100
2002-03	7.1 per 100	7.9 per 100
2003-04	7.8 per 100	8.5 per 100
2004-05	–	–
2005-06	4.8 per 100	5.5 per 100
2006-07	6.1 per 100	6.4 per 100
2007-08	6.6 per 100	7.1 per 100
2008-09	6.4 per 100	6.4 per 100
2009-10	6.3 per 100	6.9 per 100
2010-11	7.3 per 100	7.1 per 100
2011-12	7.2 per 100	7.7 per 100
2012-13	7.4 per 100	7.9 per 100
2013-14	6.6 per 100	7.6 per 100
2014-15	7.0 per 100	7.7 per 100
2015-16	6.2 per 100	7.3 per 100
2016-17	6.5 per 100	7.2 per 100
2017-18	5.7 per 100	5.8 per 100
2018-19	6.1 per 100	5.7 per 100
2019-20	5.1 per 100	5.6 per 100
2020-21	6.9 per 100	6.4 per 100
2021-22	5.2 per 100	5.3 per 100
2022-23	4.5 per 100	5.1 per 100
2023-24	5.5 per 100	5.3 per 100
2024-25	5.9 per 100	5.5 per 100
2025-26	4.3 per 100	5.0 per 100

1929-30: Forward passing in all zones
1981-82: Fewest shutouts
2003-04: Most shutouts (192)
2005-06: Shootout begins

A shutout is a team held scoreless through regulation and overtime. Each game counts twice, once for each team. Prediction: each team’s goals for a game times its opponent’s goals against a game, over the league rate, with a home or road factor, as Poisson goals. Shootout goals are left out, so a 1-0 shootout game counts as a shutout for both teams, as the NHL counts it. The two Montreal Wanderers defaults of January 1918 are dropped. 1994-95 and 2012-13 (48 games), 2019-20 (stopped early) and 2020-21 (56) are rates like any other season but noisier. The 2004-05 season was cancelled.

Find 1928-29 on the solid line: it is the tallest peak on the chart. The dashed line, the shutouts that season’s scoring predicts, peaks right beside it. Hainsworth’s low-scoring world ended the next autumn, when the NHL allowed forward passing in the attacking zone and scoring doubled. That season has its own story. Hainsworth had four shutouts in 1929-30.

The winter before, the same hockey gave Alec Connell six straight

Hainsworth was not the only goalie that hockey made look unbeatable. The shutout record that may be harder to break than his was set the winter before, by Alec Connell of the Ottawa Senators. Between Jan. 31 and Feb. 18, 1928, he had six straight shutouts. The NHL’s game records show what they were made of: two wins in regulation, a 1-0 win in overtime and three 0-0 ties, each with a ten-minute extra period on top. Ottawa scored once in those four extra-time games.

Since 1976-77 the record book has said Connell went 461 minutes and 29 seconds without allowing a goal. In 2014, Don Weekes of The Hockey News showed it was too long. In the overtime win, on Feb. 2, 1928 against the Montreal Maroons, Connell “only played 9:20 of the 10-minute extra session before teammate Hec Kilrea scored the 1–0 OT winner.” The NHL’s per-game record gives him 69 minutes and 20 seconds that night, which agrees. Charles Coleman’s history of the Stanley Cup, Weekes wrote, “counted correctly — 460:49 — in his own stat box.”

The streak ended on Feb. 22, 1928 at Chicago, in a 3-2 Ottawa win. Connell’s six and the five Boucher strung together in 2003-04 are the only runs of five or more straight shutouts in the NHL’s game-by-game goalie records. Five goalies had runs of four, all between 1927-28 and 1948-49.

Hainsworth’s 22 a year later included six more 0-0 ties like Connell’s three. That kind of shutout is gone. 186 regular-season games ended 0-0, and the last was on Feb. 24, 2004, Boston Bruins at the New York Islanders. The record book has two stranger kinds. Since the shootout arrived in 2005-06, 56 games have been 0-0 after overtime, and the NHL credits both goalies with a shutout even though one of them lost. And 82 times since 1917, a team has shut out its opponent with two goalies. That counts as a team shutout but goes into neither goalie’s total.

For a century, fewer goals has meant more shutouts

Hainsworth’s and Connell’s winters sit at the far end of a rule that has held for a century. A team that scores three goals a game still gets blanked a few nights a season by chance alone. Score fewer, and those nights come more often. For every season, we worked out how many shutouts chance alone would produce at that season’s scoring and set it beside the real count. Know a season’s scoring and you can guess its shutouts closely. Chance alone gets about three-quarters of the rise and fall from season to season right.

Fewer goals, more shutouts

Every NHL regular season since 1917-18 (108 seasons)

Shutouts per 100 team-games

0.010.020.030.01928-291981-822003-042025-26

Goals per team-game

Show data

Seasons with fewer goals a team-game have more shutouts: from 27.3 per 100 team-games at 1.46 goals in 1928-29 to 1.73 at 4.01 in 1981-82.
	Goals per team-game	Shutouts per 100 team-games
1917-18	5.0	2.9
1918-19	4.1	5.6
1919-20	4.8	5.2
1920-21	4.2	4.2
1921-22	4.0	2.1
1922-23	3.3	7.3
1923-24	2.7	8.3
1924-25	2.5	12.8
1925-26	2.3	15.1
1926-27	2.0	19.3
1927-28	1.9	22.3
1928-29	1.5	27.3
1929-30	3.0	5.9
1930-31	2.4	13.4
1931-32	2.5	12.2
1932-33	2.3	13.7
1933-34	2.4	12.3
1934-35	2.5	11.8
1935-36	2.2	13.8
1936-37	2.5	9.4
1937-38	2.5	9.4
1938-39	2.5	11.9
1939-40	2.5	10.1
1940-41	2.7	6.9
1941-42	3.1	6.0
1942-43	3.6	3.7
1943-44	4.1	4.3
1944-45	3.7	3.7
1945-46	3.3	4.3
1946-47	3.2	5.6
1947-48	2.9	6.4
1948-49	2.7	8.3
1949-50	2.7	9.8
1950-51	2.7	8.8
1951-52	2.6	7.9
1952-53	2.4	11.2
1953-54	2.4	13.1
1954-55	2.5	8.6
1955-56	2.5	10.2
1956-57	2.7	7.1
1957-58	2.8	8.8
1958-59	2.9	6.7
1959-60	2.9	6.0
1960-61	3.0	4.5
1961-62	3.0	6.9
1962-63	3.0	4.5
1963-64	2.8	9.5
1964-65	2.9	6.4
1965-66	3.0	6.2
1966-67	3.0	6.9
1967-68	2.8	8.5
1968-69	3.0	6.7
1969-70	2.9	7.6
1970-71	3.1	5.7
1971-72	3.1	5.8
1972-73	3.3	5.9
1973-74	3.2	5.1
1974-75	3.4	6.0
1975-76	3.4	4.5
1976-77	3.3	5.5
1977-78	3.3	4.7
1978-79	3.5	3.8
1979-80	3.5	3.5
1980-81	3.8	2.3
1981-82	4.0	1.7
1982-83	3.9	2.7
1983-84	3.9	2.7
1984-85	3.9	2.4
1985-86	4.0	3.0
1986-87	3.7	1.8
1987-88	3.7	3.5
1988-89	3.7	2.6
1989-90	3.7	2.4
1990-91	3.5	3.0
1991-92	3.5	3.6
1992-93	3.6	3.4
1993-94	3.2	4.5
1994-95	3.0	4.8
1995-96	3.1	4.6
1996-97	2.9	6.0
1997-98	2.6	7.5
1998-99	2.6	7.2
1999-00	2.7	5.6
2000-01	2.8	7.6
2001-02	2.6	7.5
2002-03	2.7	7.1
2003-04	2.6	7.8
2005-06	3.0	4.8
2006-07	2.9	6.1
2007-08	2.7	6.6
2008-09	2.8	6.4
2009-10	2.8	6.3
2010-11	2.7	7.3
2011-12	2.7	7.2
2012-13	2.7	7.4
2013-14	2.7	6.6
2014-15	2.7	7.0
2015-16	2.7	6.2
2016-17	2.7	6.5
2017-18	2.9	5.7
2018-19	3.0	6.1
2019-20	3.0	5.1
2020-21	2.9	6.9
2021-22	3.1	5.2
2022-23	3.1	4.5
2023-24	3.1	5.5
2024-25	3.0	5.9
2025-26	3.1	4.3

Goals per team-game leave out shootout goals. The relationship is curved, as chance predicts (the share of games with zero goals falls off quickly as scoring rises), so no straight trend line is drawn.

Each dot is a season. Slide right toward more goals and the dots fall toward fewer shutouts, steeply at first, then more gently.

The low point came at the top of the scoring boom. In 1981-82 teams scored 4.01 goals a game and were shut out 29 times in 840 games, about once every 58 games for each team, the fewest ever. The league leader, Montreal’s Denis Herron, had three, the fewest for a league leader since 1923-24, matched only by 1980-81. Save percentages hit their own low in the same years.

The most shutouts in a season came in the dead-puck era: 192 in 2003-04, when goals were down to 2.57 a team-game. In 2025-26 the league had 113, about one every 23 games for each team, the rarest since 1992-93.

Hainsworth’s league was even stingier than its scoring. Today’s is less.

Scoring does not explain everything, and what is left over points back at Hainsworth’s era. To find it, we gave every team its own scoring and its own defence, added home ice, and replayed each season 200 times with chance deciding every score. That gives the number of shutouts each season should have had. It also shows how far luck alone can push that number. Before the forward pass, the NHL had about 22% more shutouts than its teams’ scoring predicted. Luck alone usually keeps an era within about 6% of its prediction, either way. The extra shrank to about 11% in 1929-43 and 8% in the Original Six years.

Then it flipped. First, the empty-net goals have to come out. Picture a team up 1-0 with a minute left, firing the puck down the ice into an open net. That goal pads its scoring average. It can never end a shutout, because the team that scores it is already ahead. Leave those goals out, and every era since 1980 comes in 10% to 12% short of the shutouts its scoring predicts. In each of those eras, the shortfall is bigger than luck explains. Of every nine shutouts the scoring calls for, roughly one never happens.

About 22% more shutouts than scoring predicts before 1929, 10% to 12% fewer since 1980 with empty-net goals set aside

Shutouts above (+) or below (−) the number predicted from each team’s scoring and defence, by era, regular season

1917-29before the forward pass

22.4%

n/a

1929-43forward pass, red-line-free

11.4%

9.7%

1943-67red line and the Original Six

7.9%

5.9%

1967-80expansion

3.8%

0.1%

1980-93the high-scoring years

-4.1%

-9.9%

1993-2004the dead puck

-4.7%

-11%

2005-17shootout era

-3.6%

-12.2%

2017-26the empty-net era

-11.5%

All goals
Empty-net goals set aside

Prediction checked against 200 simulated seasons per season; luck alone (5th to 95th percentile) moves an era by 3% to 7% either way. “Empty-net goals set aside” removes empty-net goals before fitting team scoring, where per-game empty-net goals are known: 1933-34 on (game records to 2003-04, NHL per-game totals from 2005-06). Each era is named for the first season it starts: “1929-43” is 1929-30 to 1942-43.

Read the chart left to right. The early eras stand above zero. Set empty-net goals aside, and every era since 1980 drops below it.

Even the dead-puck record year fits. In 2003-04 a team was shut out in 7.8 of every 100 games, short of the 8.5 its scoring predicted. The record 192 came from how few goals there were. If anything, that season’s goalies finished slightly fewer games clean than its scoring called for.

There are vivid team cases, and they are mostly luck. The 2010-11 Philadelphia Flyers went the whole season without a shutout when their scoring and their opponents’ predicted about 6; the 2025-26 Flyers also had none, against about 5. A team that gets more shutouts than expected one season is barely more likely than any other team to do it again the next. The year-to-year correlation is .05, on a scale where 0 means no link at all and 1 means a perfect repeat.

The pulled goalie explains only a small part of the missing modern shutouts

Hainsworth never had to survive the modern finish. A team down 1-0 with two minutes left now sends its goalie to the bench for a sixth skater, and coaches pull goalies earlier than ever. That makes a trailing team more dangerous. If that were why modern teams get shut out less, it would show in the last minutes of games.

It barely does. From 2017-18 to 2025-26, 1,267 teams were still scoreless and behind with two minutes left in regulation. About 10% of them scored before the buzzer. At their normal scoring pace, about 9% would have. That is about 13 extra goals in nine seasons. Those teams scored at 1.11 times their usual pace, give or take 0.10, and luck alone could produce a bump that size. Those same seasons came up about 160 shutouts short of what their scoring predicted, leaving empty-net goals aside. So even if every one of those extra late goals were real, the last two minutes would explain less than a tenth of the missing shutouts.

In most eras, scoreless teams with two minutes left score about as often as their usual pace predicts

Share of teams that scored before the end of regulation, against the share their own season scoring rate predicts

1917-29

6.4%

1929-43

9.2%

7.7%

1943-67

8.4%

8.7%

1967-80

7.9%

9.4%

1980-93

10.4%

11.2%

1993-2004

11.3%

8.5%

2005-17

10.8%

8.5%

2017-26

10.3%

9.2%

Scored before the buzzer
Expected at own scoring rate

Goal-by-goal records from the NHL’s game files, 1917-18 to 2016-17, and from our play-by-play database, 2017-18 to 2025-26. 2009-10 is left out: its game files match the final score in fewer than 99% of games. Teams: 1917-29 326, 1929-43 492, 1943-67 691, 1967-80 900, 1980-93 647, 1993-2004 1,636, 2005-17 1,756, 2017-26 1,267.

In most eras, the two bars sit within about a point and a half of each other. In 1980-93, 10.4% of those teams scored against 11.2% expected. In the Original Six years, it was 8.4% against 8.7%. The one stretch that runs high is 1993-94 to 2016-17. Then trailing teams scored at 1.26 to 1.32 times their usual pace in the last two minutes, more than luck usually allows. Even there, the extra goals cover less than a quarter of those eras’ missing shutouts.

Give today’s trailing team five minutes, or a whole period, and the answer holds

The pull changed how a shutout dies. It barely changed how often one dies. Of the 130 late goals since 2017-18, 79 (61%) came with the scoring team’s goalie on the bench.

Two minutes is a short window, so we widened it. Over the last five minutes of 2017-26, 22.7% of trailing, scoreless teams scored, against 21.6% expected. Then we took the whole third period. Teams that were behind and had not scored found the net at 98% of their usual rate.

One small clue runs the other way, in Hainsworth’s own era. Before 1929, a team that was behind and scoreless after two periods scored in the third at only 88% of its usual rate, on 443 cases. That is not many, and a gap that size sits near the edge of what luck alone can produce.

So whatever thins out modern shutouts is mostly not happening late in games, and this data can’t say where it is. It belongs to the era. No single goalie owns it. That leaves the other half of the question: the shutouts that were a goalie’s own.

The part that was the goalie doesn’t come back the next season

Hainsworth’s extra four were a lot, but not the most. Tony Esposito was a rookie with Chicago in 1969-70 when he had 15 shutouts, the most by anyone since Hainsworth’s season. He was “Nicknamed ‘Tony O’ for his record number of shutouts,” as The Canadian Encyclopedia puts it. A goalie who allowed what he allowed that season would normally get about 9 shutouts. Esposito’s extra six are the most of any of the 2,064 seasons in which a goalie made 30 or more starts.

The next season he had 6, against about 5 expected. Dominik Hasek had 13 in 1997-98 when about 8 were expected, then 9 the next year against about 8. Martin Brodeur had 12 in 2006-07 against about 8, then 4 against about 7.

The biggest shutout seasons all fell back the next year

Shutouts above what the goalie’s own goals-against average predicts, that season and the next

That seasonNext season

6.4 shutouts5.5 shutouts5.0 shutouts4.7 shutouts4.6 shutouts4.6 shutouts4.4 shutouts4.4 shutouts

0.8 shutoutsTony Esposito, 1969-700.8 shutoutsDominik Hasek, 1997-980.7 shutoutsMarc-Andre Fleury, 2014-15−0.5 shutoutsSergei Bobrovsky, 2018-190.7 shutoutsHenrik Lundqvist, 2010-110.3 shutoutsGreg Millen, 1988-89−0.2 shutoutsPascal Leclaire, 2007-082.8 shutoutsSteve Mason, 2008-09

Show data

The eight biggest overshoots, led by Esposito’s +6.4 in 1969-70, came back to about what their goals-against predicted the next season.
	That season	Next season	Change
Tony Esposito, 1969-70	6.4 shutouts	0.8 shutouts	−5.6 shutouts
Dominik Hasek, 1997-98	5.5 shutouts	0.8 shutouts	−4.7 shutouts
Marc-Andre Fleury, 2014-15	5.0 shutouts	0.7 shutouts	−4.3 shutouts
Sergei Bobrovsky, 2018-19	4.7 shutouts	−0.5 shutouts	−5.2 shutouts
Henrik Lundqvist, 2010-11	4.6 shutouts	0.7 shutouts	−3.9 shutouts
Greg Millen, 1988-89	4.6 shutouts	0.3 shutouts	−4.3 shutouts
Pascal Leclaire, 2007-08	4.4 shutouts	−0.2 shutouts	−4.6 shutouts
Steve Mason, 2008-09	4.4 shutouts	2.8 shutouts	−1.6 shutouts

Expected shutouts = starts × the chance of a zero-goal game at the goalie’s goals against per 60 minutes, scaled so the league’s expected total matches its actual total each season. Goalie-seasons with 30 or more starts; the next season is whatever he played. Sorokin’s 2025-26 has no next season yet and is left out.

Every line slopes down. The winter after the big year, most of these goalies finished within one shutout of what their goals-against called for.

It happens to everyone. We looked at 932 pairs of back-to-back seasons by goalies with 40 or more starts in both. A goalie’s extra shutouts one year had no relationship with his extra shutouts the next. The 25 biggest overshoots averaged about 4 extra shutouts; the following season they averaged 0.1. A goalie’s goals-against average this season predicts how many shutouts he will get next season better than his shutout count does.

Brian Boucher said as much about his own streak. Relegated to third string for over a month, as he told it, he got the net back with Phoenix and shut out five straight opponents between Dec. 31, 2003 and Jan. 9, 2004. His saves, night by night: 21, 35, 26, 27, 21. “You have to catch some breaks and I caught quite a few, like pucks that hit posts or maybe shots that you didn’t see that end up going wide,” he told NHL.com’s Dan Rosen. Two nights later, on Jan. 11, Atlanta’s Randy Robitaille scored off a deflection. “It went off of his chest and beat me high glove,” Boucher said. “It was one of those.”

Brodeur’s 125 came from lasting. Hainsworth’s pace came mostly from his era.

The career list runs on the same arithmetic. On Dec. 21, 2009, Martin Brodeur made 35 saves in a 4-0 win at Pittsburgh for his 104th shutout, one more than Terry Sawchuk’s career total. “This record was held for so long and when you break records and see how long they have lasted it’s pretty cool,” he said. His coach, Jacques Lemaire, went further: “The record will never be broken.”

Brodeur finished with 125. That is a record for lasting as much as anything. He played 1,266 regular-season games, and his 9.9 shutouts per 100 games are in the same range as Sawchuk (10.6), Jacques Plante (9.8) and Glenn Hall (9.3).

Brodeur has the most shutouts; Hainsworth got his twice as often

Regular-season shutouts, with shutouts per 100 games played

Martin Brodeur1,266 games, 9.9 per 100 (1991-92 to 2014-15)

125

Terry Sawchuk971 games, 10.6 per 100 (1949-50 to 1969-70)

103

George Hainsworth465 games, 20.2 per 100 (1926-27 to 1936-37)

Glenn Hall906 games, 9.3 per 100 (1952-53 to 1970-71)

Jacques Plante837 games, 9.8 per 100 (1952-53 to 1972-73)

Alec Connell418 games, 19.4 per 100 (1924-25 to 1936-37)

Tiny Thompson553 games, 14.6 per 100 (1928-29 to 1939-40)

Dominik Hasek735 games, 11.0 per 100 (1990-91 to 2007-08)

Roberto Luongo1,044 games, 7.4 per 100 (1999-00 to 2018-19)

Tony Esposito886 games, 8.6 per 100 (1968-69 to 1983-84)

Ed Belfour963 games, 7.9 per 100 (1988-89 to 2006-07)

Marc-Andre Fleury1,051 games, 7.2 per 100 (2003-04 to 2024-25)

NHL regular-season goalie totals, 1917-18 to 2025-26. Shutouts shared by two goalies are credited to neither, as the NHL counts them.

Hainsworth, the man whose single season still stands, had 94 in 465 games, 20.2 per 100, about twice Brodeur’s pace. He spent his best years in the lowest-scoring seasons the league has known.

A shutout count mostly measures the season

A shutout count mostly measures how many goals were scored around a goalie. To compare goalies across eras, compare goals-against or save percentage, then the shutouts.
A goalie who leads the league in shutouts is very likely to get fewer the next season. Expect about what his goals-against says, not a repeat.
Shutouts are rarer now than scoring alone predicts, and the pulled goalie explains little of it. The shortfall barely shows up in the final minutes, and the data cannot yet say where it does.

Caution

What this cannot tell us

Our prediction is a simplification. It gives each team one scoring rate and one defensive rate for the whole season and ignores injuries, trades and backup goalies. Allowing for those would raise the number of shutouts we predict, which would make the modern shortfall bigger, not smaller. The early excess, though, could partly reflect leagues of four to ten very unequal teams. We can show where the modern shortfall mostly is not (the last minutes), not what causes it. The 2009-10 goal-by-goal records do not match the final scores, so that season is left out of the late-game test. Early records (before about 1950) are the NHL’s reconstruction from old game sheets, and early overtime rules changed from season to season. Single seasons are noisy, especially the small early ones and the shortened seasons.

Sorokin’s seven

Ilya Sorokin led the NHL with 7 shutouts in 2025-26. In a season of three goals a team-game, a goalie who allowed what he allowed would normally get about 2.6. Esposito got 15 where about 9 were normal. Measured against what each man’s goals-against called for, Sorokin’s seven were the longer shot. If the past century is any guide, next season they will look ordinary.

By the same yardstick, an average goalie in 1928-29 would have had about 12 of Hainsworth’s 22. His goals-against average, his own work and his team’s, lifts that to about 18. The other four were his, or his luck, and across the century that is the part that has not carried over to the next season.

Share this article

Post on X

Goalie tools

Live Edgehalla tools on the same subject.

All NHL tools

Show the work

Tests written down: Not in advance (exploratory)
Results computed: October 26, 2026

Sample: every finished regular-season game since 1917-18 (65,240 games), with the shootout decider removed from the winner’s score and the two Montreal Wanderers defaults of January 1918 dropped. A shutout is a team-game with no goal in regulation or overtime; this matches the NHL’s team shutout totals in all 108 seasons. Prediction: each team-game’s scoring rate = the team’s goals for a game × the opponent’s goals against a game ÷ the league rate × the season’s home or road factor, as Poisson goals. Each season is simulated 200 times and re-fitted, which gives the method’s own bias (the ratio is divided by the simulated mean) and how far luck moves a season or an era (5th to 95th percentile). The empty-net-free version removes empty-net goals before fitting, wherever they are known per game. Goalie test: expected shutouts = starts × e^(−goals against per 60), scaled so each season’s league total matches. That expectation is partly circular, because a shutout lowers the goals-against average that predicts it, which makes single-season excesses look smaller than they are. The year-to-year test avoids the problem by comparing one season’s excess with the next season’s. Records, streaks and named games: the NHL’s season and per-game goalie totals.

Every test we graded (About ten comparisons, chosen in the memo before drafting. The era pattern rests on whole eras of 2,000 to 28,000 team-games each, with luck bands from simulation; single seasons and single teams are treated as noise. The late-game groups are a few hundred to a few thousand teams each and are read as “no sign of an effect”, except the last two minutes of 1993-94 to 2016-17, which run above pace but explain a small share.)

Test	Result	Verdict
Our shutout count matches the NHL’s	Team shutouts identical in all 108 seasons; 0-0 shootout games carry two goalie credits each (112 in 56)	Holds

More tests below

Rows 2–4 of 12

Test	Result	Verdict
Scoring explains the shutout rate	Log shutout rate vs log chance rate at league scoring: r = 0.851 (108 seasons), about 73% of the variance	Mostly
Early eras had more shutouts than scoring predicts	1917-29 +22% (luck band −6% to +6%); odd seasons +29%, even +19%. 1929-43 +11%; 1943-67 +8% (halves +15% / +0%)	Holds before 1943; mixed halves in the Original Six
Modern eras have fewer shutouts than scoring predicts (empty-net goals set aside)	1980-93 −10% (band −6 to +6; halves −3 / −16); 1993-2004 −11% (band −3 to +3; halves −13 / −9); 2005-17 −12% (band −3 to +3; halves −15 / −9); 2017-26 −12% (band −4 to +3; halves −14 / −8)	Holds; every era outside its luck band

Continues below

Rows 5–7 of 12

Test	Result	Verdict
Same, counting all goals	1980-93 −4.1%; 1993-2004 −4.7%; 2005-17 −3.6%; 2017-26 +3.0%	Smaller: empty-net goals inflate modern scoring
The early-modern gap is chance (placebo)	Shuffling the 108 season ratios 2,000 times: none produced a gap between pre-1943 and post-1980 seasons as large as the real 0.353. Seasons next to each other are not independent, so this overstates the certainty	Rejected
Pulled goalies break modern shutouts	2017-26, last 2:00: 10.3% scored vs 9.2% expected (ratio 1.11, SE 0.10, n 1,267); 1980-93 ratio 0.93 (SE 0.11); 1993-2004 1.32 (SE 0.10); 2005-17 1.26 (SE 0.09), under 25% of those eras’ missing shutouts; last 5:00 1.05; third period 0.98 (SE 0.02)	Not detected since 2017-18; 1993-2017 above pace but a small share

Continues below

Rows 8–11 of 12

Test	Result	Verdict
Extra-shutout seasons repeat	r = −.03 (90% range −.09 to .02), 932 pairs; odd start years .03, even −.09; 2005-06 on .07	No
Shutouts predict next season’s shutouts better than goals-against does	Shutout rate → next shutout rate r = .42; goals-against expectation → next shutout rate r = .51	Reversed: goals-against predicts better
Sorokin’s 2025-26 was a long shot	7 shutouts on about 2.6 expected; binomial tail p = 0.0145 (Esposito 1969-70: 0.0215). Both tails are conservative because of the circularity above	Long shot; no next season yet
Team shutouts above expectation repeat	r = .05, n=1698 team-season pairs, 90% CI 0.013 to 0.093	Barely; treated as noise

Continues below

Row 12 of 12

Test	Result	Verdict
Late game: how the test works	Late-game test: every team that had not scored and trailed with 2:00 (and 5:00, and a full period) left in regulation, against its own season goals for a game over the same minutes.	Method

Frequently asked questions

What is the NHL record for shutouts in a season?

George Hainsworth’s 22 for the Montreal Canadiens in 1928-29, in 44 games. The most since is Tony Esposito’s 15 for Chicago in 1969-70. Since 2005-06 the most in a season is 12, by Martin Brodeur in 2006-07.

Who has the most career shutouts in NHL history?

Martin Brodeur, with 125 in the regular season (1991-92 to 2014-15). Terry Sawchuk had 103 and George Hainsworth 94. Brodeur passed Sawchuk on Dec. 21, 2009 at Pittsburgh.

Does a shootout win count as a shutout in the NHL?

Yes, if the game was 0-0 after overtime. Shootout goals do not count against a goalie, so in a 0-0 game decided by a shootout both goalies are credited with a shutout, and both teams with a team shutout. That has happened in 56 games since 2005-06.

What is the longest shutout streak in NHL history?

Alec Connell of Ottawa had six straight shutouts from Jan. 31 to Feb. 18, 1928. The official length is 461:29; The Hockey News found it should be 460:49. The longest since is Brian Boucher’s five straight for Phoenix in 2003-04, a streak of 332:01.

Why are there fewer shutouts in the NHL now?

Mostly because teams score more than in the low-scoring years: 3.08 goals a team-game in 2025-26 against 2.57 in 2003-04. On top of that, since 1980 teams have been shut out roughly one time in nine less often than their scoring predicts. Pulled goalies explain little of it. Since 2017-18, trailing teams have scored in the final two minutes only slightly more often than their usual scoring rate predicts, a gap inside the range of luck. From 1993-94 to 2016-17 the gap was bigger than luck, but it still covers less than a quarter of those years’ missing shutouts.

Is a high shutout total a sign of a great goalie?

Only partly. Shutouts follow goals-against average closely, and the extra shutouts a goalie gets beyond what his goals-against predicts do not repeat the next season (across 932 pairs of seasons). A league lead in shutouts is mostly scoring environment, team defence and luck.

More in The Vanishing Game

All How Hockey Changed research

Sources and method

Research and analysis, not betting advice. Past results do not guarantee future returns; bet only what you can afford to lose.

Hainsworth’s 22 Was Mostly Arithmetic

An ordinary goalie would have had about half of Hainsworth’s 22

Shutouts peaked in Hainsworth’s season and have followed scoring ever since

The winter before, the same hockey gave Alec Connell six straight

For a century, fewer goals has meant more shutouts

Fewer goals, more shutouts

Hainsworth’s league was even stingier than its scoring. Today’s is less.

About 22% more shutouts than scoring predicts before 1929, 10% to 12% fewer since 1980 with empty-net goals set aside

The pulled goalie explains only a small part of the missing modern shutouts

In most eras, scoreless teams with two minutes left score about as often as their usual pace predicts

Give today’s trailing team five minutes, or a whole period, and the answer holds

The part that was the goalie doesn’t come back the next season

The biggest shutout seasons all fell back the next year

Brodeur’s 125 came from lasting. Hainsworth’s pace came mostly from his era.

Brodeur has the most shutouts; Hainsworth got his twice as often

A shutout count mostly measures the season

Sorokin’s seven

Goalie tools

Show the work

Frequently asked questions

What is the NHL record for shutouts in a season?

Who has the most career shutouts in NHL history?

Does a shootout win count as a shutout in the NHL?

What is the longest shutout streak in NHL history?

Why are there fewer shutouts in the NHL now?

Is a high shutout total a sign of a great goalie?

More in The Vanishing Game

The Slapshot Went From One NHL Goal in Four to One in Twelve

For 19 Seasons, an NHL Hit Depended on the Rink

Did Garbage Goals Take Over the NHL?

The Last Goalie Who Played Every Night

Is the Rush Dying in the NHL?

NHL Fights Fell to 0.18 a Game in 2018-19, Then Climbed Past the 0.20 of Gordie Howe’s Era

Sources and method

An ordinary goalie would have had about half of Hainsworth’s 22

Shutouts peaked in Hainsworth’s season and have followed scoring ever since

The winter before, the same hockey gave Alec Connell six straight

For a century, fewer goals has meant more shutouts

Fewer goals, more shutouts

Hainsworth’s league was even stingier than its scoring. Today’s is less.

About 22% more shutouts than scoring predicts before 1929, 10% to 12% fewer since 1980 with empty-net goals set aside

The pulled goalie explains only a small part of the missing modern shutouts

In most eras, scoreless teams with two minutes left score about as often as their usual pace predicts

Give today’s trailing team five minutes, or a whole period, and the answer holds

The part that was the goalie doesn’t come back the next season

The biggest shutout seasons all fell back the next year

Brodeur’s 125 came from lasting. Hainsworth’s pace came mostly from his era.

Brodeur has the most shutouts; Hainsworth got his twice as often

A shutout count mostly measures the season

Sorokin’s seven

Goalie tools

Show the work

Frequently asked questions

What is the NHL record for shutouts in a season?

Who has the most career shutouts in NHL history?

Does a shootout win count as a shutout in the NHL?

What is the longest shutout streak in NHL history?

Why are there fewer shutouts in the NHL now?

Is a high shutout total a sign of a great goalie?

More in The Vanishing Game

The Slapshot Went From One NHL Goal in Four to One in Twelve

For 19 Seasons, an NHL Hit Depended on the Rink

Did Garbage Goals Take Over the NHL?

The Last Goalie Who Played Every Night

Is the Rush Dying in the NHL?

NHL Fights Fell to 0.18 a Game in 2018-19, Then Climbed Past the 0.20 of Gordie Howe’s Era

Keep reading

Sources and method