# Considering Advanced Statistics in the NCAA

Updated: March 8, 2017

Ice hockey is a sport unique in its complexity. As the legendary William Faulkner once wrote, “To the innocent, who had never seen it before, it seemed discorded and inconsequent, bizarre and paradoxical like the frantic darting of the weightless bugs which run on the surface of stagnant pools.” It is a sport that is difficult to get into, but rewarding once you are.

But despite the many components that go in to playing the game, the main methods of statistical analysis are fairly simple. Goals win games, so the team that scores the most goals will usually win the most games. Unfortunately, with the growing intricacies in offensive and defensive systems that are increasingly being deployed both in the college and the professional ranks, this simplistic conclusion may no longer be entirely accurate.

To that end, teams in the National Hockey League have begun hiring analysts to look more closely at the stats that are more predictive of success. From this has sprung the advent of “advanced stats.” This broad term encompasses many different types of analysis, but most of the discussion today relies on a measure called Corsi, which is another term for Total Shot Attempts. Essentially, all shot attempts taken, whether they are on net, miss the net, blocked, or in the goal, are counted, and a percentage is calculated by taking your team’s Corsi and dividing it by the sum of both teams’ Corsi. A Corsi greater than 50% indicates the better team in the game, or even in general.

This data has been shown to correlate very strongly with future success. Since the bulk of the game is played at 5 on 5, most analysts choose to use that measure in their work. The following chart shows a comparison of the last nine NHL Stanley Cup champions and their league-wide rank in 5 on 5 Corsi.

 Stanley Cup Champion (Year) Champion’s Corsi% (Rank) Detroit Red Wings (2008) 58.77 (1st) Pittsburgh Penguins (2009) 48.07 (19th) Chicago Blackhawks (2010) 56.55 (1st) Boston Bruins (2011) 50.70 (14th) Los Angeles Kings (2012) 54.74 (2nd) Chicago Blackhawks (2013) 54.14 (4th) Los Angeles Kings (2014) 56.82 (1st) Chicago Blackhawks (2015) 53.59 (2nd) Pittsburgh Penguins (2016) 52.72 (2nd)

Other than a couple of outliers, every Stanley Cup champions ranked amongst the top four in 5 on 5 Corsi%.

Based on the success this measure has had in predicting champions in the NHL, interest has risen in its place in college hockey. In 2014, an interview of many college hockey head coaches showed responses varying from excitement to skepticism regarding advanced stats. Given how long it has taken advanced stats to catch hold in the NHL (advanced data has been available since the 2007-08 season and it is still a major point of disagreement among hockey purists), it is no surprise that it is not yet prevalent in the NCAA, a league that is largely seen, at least in the hockey world, as a developmental system to teach youngsters the “proper” way to play the game.

This is no different at Cornell, a school with a long, proud hockey tradition that has seen legends such as Ken Dryden and Joe Nieuwendyk pass through its ranks. Men’s Hockey Head Coach Mike Schafer has long been considered to be one of the best defensive coaches in college hockey, yet seems to be skeptical with the growing sentiment in the hockey world that speed and skill are the key ingredients to a championship-winning team.

“We use some statistics in preparation and evaluation, Schafer told us in an email. “I have seen some of the most complicated reporting of some NHL teams with their statistics and I don’t believe that the coaches truly know how to use them and I don’t know if the people preparing them really understand the intangibles when producing the statistics.”

“Our team places emphasis on both offense and defense,” Cornell men’s hockey captain Jake Weidner added via email. “There are times throughout the season where we find weaknesses in our offense or defense and we tend to focus on shoring up those spots, but for the most part we focus on a complete game plan.”

On the surface, there is nothing wrong with these sentiments. A successful team obviously needs a balanced game plan, lest they expose their weaknesses. And it is also not surprising to hear Schafer talk about the intangibles, especially in a sport such as hockey, where physicality and effort, better known as “grit,” have long been prized by fans of grueling, punishing, “old-time hockey.”

Unfortunately, as more and more teams are finding out, both in the NHL and in the NCAA, grit, while important, cannot win a team a championship by itself. The problem with grit lies, not in the correct belief that effort is more effective than skill, but with the incorrect assumption that grit and skill are mutually exclusive. A player can play hockey with incredible passion and effort while still using his talent and skills to good use. A good team needs skill and grit, not one or the other.

As for shooting the puck, Schafer told us the following, “Like anything else statistics can be very useful or very misleading because of the intangibles that go along with them. For instance if you are a puck possession team then shot attempts will be lower than if you are a team that is built to force pucks to the net at every opportunity. Both teams would be successful but would have skewed numbers.”

Weidner adds, “We focus on making sure we are hitting the net in order to generate scoring chances on the first shot, but also the possibility of a rebound that may lead to another chance.” However, he adds, “As a team tactic, shooting the puck can increase the number of scoring chances, but there are times when a misguided shot can just be giving the other team the puck.”

These beliefs, that shooting the puck is more-or-less equivalent to giving up possession of the puck, seems to be the newest misconception about advanced stats. First of all, regarding Schafer’s comments, he is considering puck possession to be actually holding on to the puck, which is a very misleading belief. Of course, having the puck a lot is preferable to not having it at all, but having the puck and not taking shots will ultimately not get a team anywhere. Additionally, shot attempts have shown to be reflective of actual puck possession, since a team that takes more shots will invariably have the puck more, or at least will have the puck most effectively.

Second, Weidner’s point about a misguided shot being equivalent to a turnover has become a popular criticism for advanced stats skeptics. In fact, NHL coach Barry Trotz recently said that he does not use Corsi to a great extent because he believes that players can “game” their Corsi by taking low-quality shots (incidentally, Trotz’s team, the Washington Capitals, who currently lead the league, are also one of the best teams in Corsi). However, this claim is not supported by stats, since if a player or team tries to inflate their shot attempt total by taking a poor shot, it will probably lead to chances for the other team, thus leading to an overall detrimental effect on Corsi. Weidner is not wrong when he says that teams should avoid taking misguided shots, but his suggestion that only shots that are on goal can be quality ones is flawed.

While there is no clear answer so far as to whether advanced statistics can make the jump from the NHL to the NCAA, the early returns seem promising. Corsi data is only available for the past 2+ seasons, and the following table shows where Frozen Four teams ranked amongst D1 hockey teams in 5 on 5 Corsi.

 Year Champion (Corsi Rank) Runner-Up (Corsi Rank) Semifinalist (Corsi Rank) Semifinalist (Corsi Rank) 2014-15 Quinnipiac (2nd) Boston (6th) North Dakota (23rd) Omaha (45th) 2015-16 North Dakota (4th) Quinnipiac (7th) Denver (12th) Boston College (18th)

Both years, the NCAA champion ranked top-four in the country in 5 on 5 Corsi%. Cornell, however, struggled, finishing 38th and 42nd in the respective year. Both times, Cornell exited the ECAC tournament early, falling in the first round in 2015 and losing in the second-round to Quinnipiac in 2016.

This year, although Cornell has a strong record, their Corsi% ranking is low, at only 30th in D1. In addition, while many times the numbers will not necessarily match up with the qualitative analysis, or the “eye test,” in this case it does. Cornell, throughout the season, has been very wary of shooting the puck, even when they have established themselves in the offensive zone. They currently have the 8th-fewest shots on goal and the 4th-fewest shot attempts in the country. As they move into tournament season and defenses get better, it will be more and more important to take more shots and take advantage of the fewer opportunities they will have.

This paints a somewhat bleak picture of Cornell’s situation. Yet, something must be working for Cornell to be a top-10 team in the country. In fact, there are two specific things working for them. First, and most obviously, is the defense. As noted above, Schafer is considered to be a defensive guru, and Cornell has one of the best defenses in the country.

“When focusing on defense, it usually involves how we can quickly get the puck back so that we can go play on offense,” said Weidner. “Focusing on defense and how to stifle the other teams attack will launch your offense if you can force turnovers.”

This is a very sound defensive philosophy, and one that actually speaks to the value of using advanced stats in the coaching method. Given that Corsi is based on shot attempts, limiting opponent shots is as important as shooting the puck yourself. And it seems to have been working. They have the fewest goals against and the 7th-fewest shot attempts allowed in the country, both marks of elite defensive squads.

That being said, if the answer to a subpar offense was to have an elite defense, then we would not see the patterns we have seen in the NHL and NCAA in terms of strong Corsi predicting championship teams. So, there must be another factor that can explain Cornell’s success this year, not to mention the success of subpar Corsi teams such as Omaha two years ago. They cannot just be getting lucky, can they?

The answer is yes, they are getting lucky, but not in the way some may think. There is no way to easily quantify the “luck of the bounce” in a game as complex as hockey, and there is especially no way to use randomness to predict future success. Instead, advanced stats defines “luck” using a number known as PDO, which is essentially the sum of a team’s shooting and save percentages.

PDO works as a sort of scale with the average set at 100. This is because any shot on goal that is taken will either be saved (increasing save percentage and lowering shooting percentage) or scored (vice versa). Thus, the league average PDO will always be 100. The most simplistic interpretation of this statistic is that if a team has a PDO below 100, they are getting bad luck, and we can expect them to improve as their numbers regress back to 100. However, if the team has a PDO above 100, they are getting lucky, and we can expect the opposite to happen.

Currently, Cornell has a shooting percentage of 10.4% (13th in the country) and a save percentage of 92.0% (10th). As such, their PDO is a staggering 102.4, well above 100. This does not necessarily mean that Cornell has played poorly, but it does seem to indicate that their results have over-performed their play.

So what does this mean? Is it time to panic? Is Cornell going to crash and burn in the ECAC and/or NCAA tournament?

Not necessarily. While advanced stats can be useful for making predictions, it is not the end of the conversation. Who knows, Cornell could continue to carry their high PDO, and maybe even improve their Corsi while they’re at it. Even if they don’t, this is still a talented team with a very, very strong defense that could carry them far. And it must be noted that the best teams, the champions, both in the NHL and NCAA, are usually both good (high Corsi) and lucky (high PDO). Only two of the Stanley Cup champions listed above had a PDO below 100, and both of the last two Frozen Four champions had super-high PDO’s as well as their good Corsi numbers (Quinnipiac and North Dakota carried astronomical values of 103.1 and 103.9 respectively).

All in all, it is yet to be seen if advanced stats will truly take enough of a hold in the NCAA to impact coaching the way it has started to in the NHL. As long as teams continue to have success employing stifling defensive systems like Cornell’s, they may continue to be seen as inconsequential. Honestly, maybe they are, and maybe Cornell could become the next Omaha to advance deep into the tournaments despite a subpar Corsi number.

But if Cornell once again disappoints, it may be worth it to take a closer look at advanced stats and the effects they may have on the NHL, the NCAA, and even Cornell.