On Saturday the San Jose Sharks will host the Los Angeles Kings at Levi's Stadium in what will be the 15th outdoor regular season game in NHL history. With the number of outdoor games on a steady rise, there's an open and important question of whether or not the quality of play differs between indoor and outdoor games. Do lower temperatures and potentially subpar ice conditions result in noticeable differences in the way the games are played? We can't quantify the eye test, but we can see if the changes in environment show up in the stat sheet.
To answer this question, I compared several key statistics from the 14 outdoor games (listed in the table above) to those from over 14,000 regular season games since the 2002/2003 season. Here's the baseline information for those games:
|Date||Game||Location||Temperature (F)||Wind Speed (MPH)|
|Nov 22, 2003||Montreal Canadiens (4) at Edmonton Oilers (3)||Commonwealth Stadium||-3||2|
|Jan 1, 2008||Pittsburgh Penguins (2) at Buffalo Sabres (1)||Ralph Wilson Stadium||34||9|
|Jan 1, 2009||Detroit Red Wings (6) at Chicago Blackhawks (4)||Wrigley Field||28||10|
|Jan 1, 2010||Philadelphia Flyers (1) at Boston Bruins (2)||Fenway Park||33||10|
|Jan 1, 2011||Washington Capitals (3) at Pittsburgh Penguins (1)||Heinz Field||50||10|
|Feb 20, 2011||Montreal Canadiens (0) at Calgary Flames (4)||McMahon Stadium||14||30|
|Jan 2, 2012||New York Rangers (3) at Philadelphia Flyers (2)||Citizens Bank Park||39||10|
|Jan 1, 2014||Toronto Maple Leafs (3) at Detroit Red Wings (2)||Michigan Stadium||12||1|
|Jan 25, 2014||Anaheim Ducks (3) at Los Angeles Kings (0)||Dodger Stadium||55||5|
|Jan 26, 2014||New York Rangers (7) at New Jersey Devils (3)||Yankee Stadium||21||10|
|Jan 29, 2014||New York Rangers (2) at New York Islanders (1)||Yankee Stadium||20||10|
|Mar 1, 2014||Pittsburgh Penguins (1) at Chicago Blackhawks (5)||Soldier Field||16||0|
|Mar 2, 2014||Ottawa Senators (4) at Vancouver Canucks (2)||BC Place||41||20|
|Jan 1, 2015||Chicago Blackhawks (2) at Washington Capitals (3)||Nationals Park||46||10|
In particular I looked for an indoor/outdoor difference in the total number (for both teams) of goals, shots on goal, Fenwick attempts (SOG plus missed shots), Corsi attempts (SOG plus missed shots plus blocked shots), penalties, faceoffs, turnovers, and hits. The table below provides the average number and standard error of each of these statistics (summed across both teams) for outdoor and indoor games.
|Goals||5.143 (0.662)||5.363 (0.019)||No|
|Shots on Goal||60.643 (2.573)||55.469 (0.087)||Yes (p < 0.10)|
|Fenwick Attempts||82.538 (3.473)||77.857 (0.121)||No|
|Corsi Attempts||105.385 (4.437)||103.395 (0.158)||No|
|Penalties||7.857 (0.662)||9.958 (0.041)||Yes (p < 0.05)|
|Penalty Minutes||15.714 (1.324)||21.535 (0.116)||Yes(p < 0.05)|
|Faceoffs||62.857 (3.065)||55.783 (0.085)||Yes (p < 0.05)|
|Turnovers||27.000 (3.695)||28.938 (0.114)||No|
|Hits||44.923 (5.539)||40.288 (0.134)||No|
The table reveals some interesting trends. We might predict that poor ice conditions would lead to an increase in the number of turnovers. Somewhat surprisingly, there were slightly fewer turnovers in outdoor games. On average, there have been 27 turnovers per outdoor game and 28.9 turnovers per indoor game; this difference however is not statistically significantly different from zero. Because the difference is not statistically significant, this is encouraging evidence that weather and ice conditions are not making outdoor games more sloppy.
There is also a statistically significant difference for the number of shots on goal. Outdoor games have, on average, 5 extra shots on goal compared to indoor games. What is interesting is that there is no indoor/outdoor difference in Fenwick and Corsi attempts, which add missed shots and blocked shots to the shots on goal statistic. This means that the shots on goal are not being driven up in outdoor games by more attempts being taken, but rather from missed or blocked shots making it through to the goaltender.
One possible reason for this is that defenders may be out of position more often in outdoor games, perhaps because the cold weather slows them down. More shots on goal may also explain why there are more faceoffs in outdoor games, since the goaltenders will have more opportunities to cover the puck and force a faceoff.
Another interesting finding in the table above is that there are about two fewer penalties called in outdoor games (7.9) than indoor games (10.0) resulting in about 6 fewer penalty minutes per game. There are two possible explanations for the decrease in penalties called during outdoor games. The officials may tend to call outdoor games more loosely, perhaps because they do not want to influence the outcome of these high-profile games. Alternatively, the teams playing in these games may be less penalty-prone than the other 28 teams in the league.
To tease out the explanation, I use a statistical process called matching. I match each outdoor games to the other regular season matchups between the two teams in the same season as the outdoor game, as well as the seasons before and after. This cuts the size of my dataset from about 14,000 games to 134 games. By limiting the analysis to these games, I can account for the team attributes which may influence the values of these statistics. In effect, I can get a better estimate of the "effect" of playing outdoors by controlling for the team matchup.
The figure above shows the results from nine separate linear regressions using the matched dataset. The outcome variables (listed on the left side of the figure) were regressed on whether the game was indoor or outdoor as well as matchup fixed effects. Each dot in the figure is the coefficient estimate for the indoor/outdoor variable, and the thick and thin lines represent the 90 and 95 percent confidence intervals around those estimates.
Just like the first table, this analysis suggests that there are indeed more faceoffs and shots on goal in outdoor games than indoor games. Because this analysis controls for the team-matchups, we know that these statistically significant results are not the consequence of systematic patterns in the types of teams being selected to participate in outdoor games. For the other seven statistics, there is not enough evidence to suggest that indoor and outdoor games are different because the confidence intervals for each of those seven overlaps with zero.
The lack of a significant result for the penalties and penalty minutes variables conflicts with the evidence from the earlier analysis. When I compare outdoor games to all indoor games, I find that there are fewer penalties in outdoor games. When I cut out all indoor games from the analysis except those between the teams featured in the outdoor games, there is not a difference in penalty rates.
What this tells us is that the lower number of penalties being called in outdoor games is not a result of the game being played outside, but rather of the teams being selected to play in outdoor games. This seems like a valid conclusion since the NHL tries to schedule its best teams in outdoor games, and high quality teams are more likely to play with discipline and commit fewer penalties.
So what should we expect from Saturday's game in Santa Clara? History tells us that the stat sheet will not look very different than it would if the Sharks and Kings were meeting at the SAP Center in San Jose. The only possible exception is that there will be more whistles, resulting in more faceoffs and a potentially slower game.
Stephen Pettigrew is a PhD candidate at Harvard University, where he studies political science and American politics. He also has a master's degree in statistics from Harvard. In his spare time, he writes about sports analytics, particularly in hockey and football.
Photo via AP