Does The NHL Preseason Matter? – An Amateur Statistical Analysis

facebooktwitterreddit

Mandatory Credit: Rick Osentoski-USA TODAY Sports

There’s people who overreact to every loss in the preseason. Then there’s people who don’t even bother tuning in because they claim it doesn’t matter. Personally, I’m a little in between both. Preseason is fun, but I don’t take it too seriously. I’m more happy to find hockey than I am seriously looking at the teams and expecting a preview of what’s to come. But like many of my articles, it got me thinking: is there data to support the idea that the preseason doesn’t matter?

I did something similar for my college statistics class, wondering if the NHL’s elite goal scorers score more often in the 1st, 2nd, or 3rd period. (in case you’re wondering, 1st period scoring is significantly outweighed by the 2nd and 3rd period scoring).  You can argue that the preseason has intangibles, this I won’t deny. Chance to see if your new guys can gel, gives coaches a chance to try out new systems, there are plenty of reasons to value the preseason aside from less than ideal hockey. But what I wanted to know was if theres any correlation between how a team performs in the preseason and how they perform in the regular season. Sure you can have an opinion on a way to value or devalue the season, but opinions are all subjective. Statistics on the other hand, are not.

Well sure you can lie with them but I’m not nearly smart enough to be able to pull that off. And if I was I wouldn’t do it for something like a hockey blog.

First, a disclaimer. I only took one statistics class and this question gets real complicated real quick. If you want a pro looking at stats, check out Tyler Dellow , Hockey Numbers or some of the other solid hockey stats guys like Pension Plan Puppets. In fact, add them to your favorites because understanding advanced stats can make you understand the game better, and the more you understand it the more you can enjoy it.

Ok, here’s how I tried to do it.

Methods

Looking at this sounded simple at first. Look at how teams did in preseason, look at how they did in regular season, compare. This isn’t quite as easy as I hoped, however. The data is easy enough to have a look at thanks to ESPN, but knowing what to do with it is another story. There’s no way to account for roster changes or experimental systems. For example, Detroit pounded on Boston last week 8-2 with Malcom Subban in net and without Bergeron or Marchand in the line up. Like wise, Detroit got spanked by Pittsburgh earlier this week while missing Kronwall, Alfredsson, Zetterberg and Datsyuk. These kinds of wins and losses aren’t a very accurate reflection of any potential performance in the regular season and sadly, there isn’t a very good way around this.

Luckily, the next problem had a potential solution. While the regular season has the same number of games for every team, the preseason does not. Most teams play between 6 and 8 games but it’s far from consistent. The Rangers played 3 in 2011. To get around this rather than compare points earned, I would compare Percentage of Possible Points Earned. Therefore, rather than trying to compare the Rangers’ 3 points in 2011 to the Islanders’ 2, I’d compare the Rangers’ 50% to the Islanders’ 20%.

Lastly, it’s also worth mentioning that with such a small sample of games in the preseason, there isn’t much variance in opponents. So a team that plays Pittsburgh and Chicago can’t accurately be compared to a team that plays Florida and Calgary. There isn’t much of a way around this either.

So to make a comparison and look for a correlation, we divide the points earned by the possible points (Games Played * 2) to get Percentage of Possible Points Earned. We do something similar to the regular season records and finally come to see a difference. If theres a ridiculous difference, there isn’t much correlation. If there’s a small difference, then there’s a correlation.

Please don’t get correlation confused with causation, that’s a whole other headache.

Also, to try and see if there’s a trend without creating a large sample size that would take me hours to sift through, I took the last three preseasons into account. So from 2012 to 2010. Oh wait, there was a lockout. Not that I’m still bitter or was supposed to go to the Winter Classic but probably won’t be able to this year or anything… not bitter at all… 2011-2009. Then we’ll average everything!

For the sake of simplicity, Atlanta and Winnipeg will be considered the same team, mostly because they are (you can’t re-write history Winnipeg, your old franchise is in Phoenix, you inherited the Thrashers and their records, not Temmu’s Jets!!!)

If you don’t want to break down each year’s chart, I promise I summarize it.

Results

2009

TeamPreseason GPPreseason PEPreseason PPE %RankingRegular Season PERegular Season PPE %RankingDifferenceANA8100.6311890.54176BOS8110.696910.55148BUF690.7541000.61106CAR780.5713800.492411CBS440.5018790.48268CGY650.4226900.551511CHI660.50181120.68315COL890.5614950.58122DAL760.4322880.54184DET990.50181020.62711EDM890.5614620.383016FLA9110.6112770.472816LOS890.56141010.6286MIN760.4322840.51220MON790.6410880.54188NJD680.6771030.6352NSH590.9011000.61109NYI850.3129790.48263NYR770.5018870.53213OTT640.3328940.571315PHI760.4322880.54184PHO890.56141070.65410PIT630.25301010.62822SJS770.50181130.69216STL7110.793900.551512TAM7100.715800.492419TOR9120.677740.452922VAN9160.8921030.6353WAS680.6771210.7416WIN/ATL650.4226830.51233Average =9.233333333Standard Deviation=6.128977714

2010

ANA760.4322990.60913BOS770.50191030.63712BUF680.677960.59147CAR660.5020910.55191CBS9100.5618810.49235CGY7141.001940.571716CHI760.4322970.59139COL740.2927680.41292DAL670.5814950.58162DET860.38251040.63619EDM680.677620.383023FLA670.5814720.442814LOS680.677980.60125MIN740.2927860.52216MON780.5717960.59143NJD670.5814810.49239NSH680.677990.6092NYI520.2029730.45272NYR680.677930.571811OTT870.4421740.45265PHI8110.6951060.6532PHO8110.695990.6094PIT6100.8321060.6531SJS760.43221050.64517STL8100.6312870.53208TAM690.7541030.6373TOR9110.6113850.52229VAN860.38251170.71124WAS6100.8321070.6520WIN/ATL610.0830800.49255Average=7.966666667Standard Deviation =6.65651526

2011

ANA760.4322800.49253BOS690.7551020.6272BUF690.755890.541813CAR630.2528820.50235CBS8110.697650.403023CGY660.5018900.55171CHI750.36251010.621015COL640.3326880.54206DAL7120.862890.541816DET890.56121020.6275EDM8100.639740.452920FLA630.2528940.571414LOS780.5711950.58132MIN770.5014810.492410MON840.2527780.48281NJD660.50181020.62711NSH8140.8811040.6354NYI520.2030790.48273NYR330.50211090.66219OTT770.5014920.56151PHI790.6481030.6362PHO760.4322970.591111PIT6100.8331080.6641SJS6100.833960.59129STL8100.6391090.6627TAM660.5018840.51213TOR880.5013800.492512VAN860.38241110.68123WAS770.5014920.56151WIN/ATL770.5014840.51217Average=8.333333333Standard Devisation=6.929862084

Ok, a bit of summary. The difference between where a team ranks in pre-season and regular season over these three years averages is in the last table.

AveragesStd. Deviation20099.246.1320107.976.6620118.336.93Averages8.5133333336.573333333

8.5 positions. Just about every team would fall out of the playoffs if they slipped 8 or 9 spots. Likewise, almost any team could jump into the playoffs by gaining eight or nine spots. That standard deviation is pretty hefty too. No team would be comfortable knowing they’d be tenth in the league plus or minus six spots.

Conclusion

There’s no statistical link between how a team performs in the preseason and how they perform in the regular season based on that standings of each.

Other noteworthy things

*If nothing else convinces you that there isn’t much link, consider the 2010 Calgary Flames. They won 100% of their preseason games, yet failed to make the playoffs. They only failed to make the playoffs by three points, but failure is failure.

*Lowest preseason PPE % I saw was the 2010 Atlanta Thrashers. 6 games played, one point earned. 8% Possible Points Earned. To no one’s shock, they too missed the playoffs.

*On the other side of the coin, the 2009 Pittsburgh Penguins won 3 points in their 6 preseason games, 25%. They finished 4th in the East.

*The 2011 Florida Panthers earned 3 points in 6 preseason games for 25% but finished with 94 points, ending up 3rd in the East for winning their division. If it went just by points this still would have been good for 6th.

*A high preseason PPE% may suggest regular season success, as the top five preseason PPE% in 2010 included Tampa Bay, Pittsburgh, Washington and Philly. As mentioned above, Calgary topped all of them though.

*Vancouver had 38% PPE in 2010, a year they won the President’s Trophy.

*You hear a lot about teams trying to stay above .500 and looking at this, PPE% might be a better measurement, especially in the era of three-point games. Staying above a .500 win % doesn’t do much for you, and it isn’t difficult to miss the playoffs and stay above .500.

*The Wings averaged a PPE% of .48 in these three preseason. For these three regular seasons they averaged .62.

So there you have it. No evidence of a statistical link between preseason performance and regular season performance. If you have a though on this collection of data or know ways to make something like this better (besides increase the sample size) be sure and let us know in the comments below! Especially if you aren’t an amateur like I am!