Hockey has long been a sport in which hard work is cherished as the ultimate virtue. You certainly won’t find me arguing that players and coaches do not need to work hard, but somewhere along the line, an undercurrent of distaste for intelligence has developed. Using analytical tools to evaluate players and teams has been met with suspicion and often even hostile disapproval.
Don’t get me wrong, I’m not talking about all intelligence being frowned upon here, just the kind that comes from the mouths of those that have never played the game at the NHL level. Intelligent analysis or strategy from “hockey guys” is of course granted a place of almost mythical honor in the sport, so I don’t think it’s fair to say that hockey rejects smartness as a whole. It’s more that hockey often demands that you have the appropriate level of street cred, or rink cred, I guess, in order to be considered to have intelligence of any actual value.
The heart of the now years long stats vs anti-stats battle lies in this mentality. Traditional analysis in hockey tends to put the most stock in experience and knowledge gained from years of playing the game. Statistical analysis in hockey tends to put the most stock in the gathering of data and analysis based there upon. “Watch the game, nerds” and “Your eye-test sucks” spring from the perceived disconnect between those two sides.
Traditional analysts see statistical analysts basing opinions of a player’s skill or value on numbers and perceive a lack of “hockey knowledge” in the results of a computation. This hockey knowledge has in many cases been built through playing the game at an elite level or being part of the game for many years. No one wants to feel like their hard earned skills are not unique or special and are easily replaced by a nerd behind a computer who didn’t have enough talent to play the game. What they usually mean when they say “Watch the Game, Nerds” is “Watch the game the way my years of experience tell me to watch it. Value the qualities that I value.”
This resentment is particularly apparent when the stats analysts say something about a player that is inconsistent with the commonly held opinion of that player gained through the eye-test. It can be very difficult to reconcile two very different views of a player as a result. Which opinion is correct? If one of the opinions is incorrect, does that mean the whole system of evaluation driving the “wrong” opinion is flawed?
Traditionalists see statisticians putting numbers on blast and yell “Watch The Game, Nerds” while statisticians see traditionalists espousing the virtues of leadership and making the safe play and yell about dinosaurs. Statistical analysis has led us to the realization that Expected Goals are the best predictor of future goals and ultimately, success. Shot metrics are the second best predictor of future goals. This is all based upon math and statistical modeling.
Traditional analysis is founded upon understanding the game of hockey and observational skills such as knowing the systems that are used, the players’ responsibilities in those systems, what the safe play is and what talent or skill looks like. When statisticians say expected goals are the best predictor of future goals (success), traditionalists say “duh, everyone knows goals win games.” When statisticians say shot metrics help predict future goals, the traditionalists say “this isn’t some big revelation, everyone knows having the puck is better than not having it.”
The problem with all of this is the lack of a nuanced understanding and appreciation for what the other side knows and takes into account in its analysis.
To even get started on the right path to creating predictive statistical models, a person must have a certain level of understanding as to what is and is not important to winning hockey games; however, having an in-depth knowledge of formations or systems or the ability to identify individual skills on a microscopic level is not really necessary to carry this out. The more knowledge you have, the deeper the understanding of the variables that impact goals, shot generation and shot suppression will be and thus the more thorough, comprehensive and ultimately, useful, the statistical model will be.
To effectively analyze the game and players through a traditional lens, you have to understand long standing hockey knowledge and be able to judge skill. The problem with both of these approaches is that a lack of knowledge or misplaced emphasis on elements of play that do not actually contribute to winning can introduce big flaws into your analysis. I’ve seen plenty of traditional analysis that completely ignores all of the lessons we’ve learned from statistics and statistical analysis that ignores the practical concerns of what happens on the ice (I’ve been guilty of the latter in the past, particularly when I first started out).
Traditional analysis on TV or radio broadcasts tends to build an emotional narrative, because that is very appealing to many fans. Most fans of a sport watch it because it is an enjoyable hobby, even during the times that their team is struggling. Emotional narratives and human interest stories most certainly have a place in the discussion of sports because they are an important part of the overall allure and culture.
Statistical analysis strives for objectivity by relying upon numbers and banishing reliance upon human memory because of its tendency to be inherently flawed by confirmation bias and other human failings. Objectivity in analysis can often trend toward a cold or dispassionate view of the game that most of us were initially attracted to on an emotional level. It doesn’t make objective analysis any less true, but it may definitely make it less appealing to fans. Perhaps the trick for broadcasts to satisfy both sides is finding a way to build emotional narratives around objective information.
Traditional knowledge in hockey is founded upon years of watching the game, analyzing what you have observed and acting in accordance with those observations moving forward. What you look for on the ice plays a big part in shaping your analysis of the game. If you feel strongly that will, effort and tenacity are the most important traits of a good hockey player, those are the things you will look for while watching the game. If you value puck skills over all others, that is what you will look for and notice the most. We all view the game through different lenses and that goes a long way toward coloring our overall observations.
While under heavy pressure in his defensive zone, Player A gains possession of the puck and dumps it into the neutral zone. His team is able to change some of the forwards and a defenseman before going back on defense. After another few seconds, the last two players are able to change for fresh legs.
Traditional: Safe play, no goal was given up, line change was completed.
Statistical: Conservative play, possession was conceded. The attacking team gets possession and regains the offensive zone 70% of the time when the puck is dumped out.
Progressive: Review the decision making process leading to the dump out and reinforce that option as a last resort because of the low statistical likelihood of recovering possession and going on the offensive attack. Adjust the breakout scheme for end of shift plays in the defensive end to provide close support on the strong side thereby increasing the likelihood of exiting the zone with the puck under control even if it is dumped into the offensive zone during the line change.
Player A dumps the puck out of the defensive zone. A forward who just jumped over the boards on a line change skates hard into his offensive zone and causes the regrouping team (who was just on the attack moments before) to turn the puck over. The forward gains possession of the puck and an offensive attack is pressed.
Traditional: Safe play (dumping puck out) turned into offensive opportunity through hard work and tenacity of forechecking player. If the team/player works hard, this strategy will be successful.
Statistical: Conservative play (dumping the puck out) turned into an offensive opportunity; however, such occurrences are only successful in bringing about entry into the offensive zone about 9% of the time and thus not an effective strategy.
Progressive: Adjust breakout strategy for end of shift dump outs to mandate a specific forward (likely based upon area of play) gives chase to the dumped out puck to put pressure on the attacking team and buy time for the line change. This may also increase the likelihood of a turnover and thus recovery of possession of the puck.
When we combine traditional and statistical approaches to hockey, we can help eliminate our inherent biases through the use of numbers and improve our systems and strategies based upon years of experience and knowledge. Essentially, we take the best of both worlds. When the numbers we are seeing don’t seem to match up with what we observe, we can delve further into the reasons they don’t match, whether it is the result of our own confirmation bias (seeing what we are looking for or what we expect to see) or a player’s numbers being buoyed by other players on the ice at the same time or the way they are used during the game.
Using both approaches provides a system of checks and balances. The statistical side eliminates our tendency to rely too heavily on small samples and unsustainable levels of success or failure, while the traditional side keeps the numbers grounded in reality and practical use.
For example, veteran players tend to make predictable plays that in the eyes of statistical analysis are considered bad plays or mistakes that often fall in the category of making the safe play in the eyes of traditional analysis. Rookies and other young players tend to make plays that are unpredictable and considered mistakes, irresponsible plays or plays that fall into the category of too risky in the eyes of traditional analysis. Coaches, at least the vast majority of them, fall in the traditional analysis category so it is not difficult to recognize why they would prefer the veteran “mistakes” to the rookie “mistakes”.
Most coaches at high levels of the game do not have very high levels of acceptable risk which also puts the younger players and creative players at a disadvantage from the get go. If we combine the two approaches, we see there is a big difference between irresponsible play and taking a risk. In the NHL today, the level of acceptable risk seems to be at very low levels. As a result, much of the creativity that landed players in the NHL is coached out of them in favor of making the “safe” play.
There are only so many times that a young player can be punished for taking a risk before he starts to ignore his gut when he sees the opportunity for creativity arise. Personally, I think if we were better at recognizing the difference between irresponsible play and risk taking, we would see more of the reward the latter creates and stop stifling the creativity that makes the game so enjoyable. Raising the level of acceptable risk requires more preparation to ensure the players know how to respond and react to one of their teammates taking a risk in case things go wrong.
Coaches can instill confidence in their players by teaching discipline in carrying out the systems he’s implemented and preparing them for the situations they will face. This helps them avoid irresponsible play and know what is expected of them. Building up the trust between player and coach is all about decision making on the ice. The more the players know what the coach wants from them in a given situation, the more the coach can trust the players to make good decisions. If the coach can trust a player to make good decisions, he can also encourage that player’s creativity. Creativity is the antithesis of predictability. Predictability is much easier to prepare for and defend against than creativity.
Players can help themselves in this regard too. If a player wants to follow his gut and take a risk to create something, he better give maximum effort to mitigate the negative impact on his team if it doesn’t work out, every single time. If the coach can’t trust his players to work their tails off on the backcheck, he can’t trust his players to take risks.
Another practical example of combining traditional and statistical approaches to maximize efficiency and effectiveness:
Situation: The defense has been pinned in their own zone for a while and the players are getting tired.
Traditional approach: the player that can get possession of the puck should try to dump it out of the zone.
Desired result: get the puck far enough out of the zone to complete a line change and get fresh legs on the ice.
Reward: While the attacking team will likely get possession of the puck and relaunch their attack, fresh legs will be there to defend against that attack and take another shot at getting possession of the puck so they can go on the attack. A player with fresh legs might be able to cause a turnover as the new attack starts and turn the tables for his team.
Risk: Icing the puck thereby negating the ability to change on ice personnel, getting caught in the line change and scrambling to defend against a fresh offensive attack in your defensive zone, accomplishing the line change and leaving the new players to defend against the offensive attack and again try to get the puck out of the zone.
Progressive approach: the player that can get possession of the puck should try to get the puck out of the zone under control.
Desired result: even if he simply dumps it into his offensive zone for the purpose of completing a line change, the puck is as far away from the defensive zone as possible thereby buying more time for the line change.
Reward: the line change has more time to happen safely, potentially a greater chance of regaining possession in the team’s offensive zone if a forechecking player can jump over the boards quickly and force the regrouping team to make a mistake in their own defensive zone.
Risk: Controlled zone exits take more time to carry out, greater chance that possession will be lost during this attempt, the line change will not happen if the puck is still in the defensive zone, tired players may make mistakes that lead to a goal against or take a penalty.
Both of these approaches carry risk with them, so how do we decide which one is better?
If I’m a coach and with my analyst, I go over the numbers for dumping the puck out of the zone, I find that when my team dumped the puck out, we were immediately defending against an attack roughly 70-72% of the time. My team was able to go on the attack between 28-30% of the time after dumping the puck out. When my team exited the zone with the puck under control (carry out or a tape to tape pass), we were able to go on the attack about 88% of the time. So, if we know that dumping it out leads to defending against another attack 70% of the time and controlling the zone exit leads to defending against another attack about 12% of the time, it is obvious that the controlled exit is preferable.
When we combine what the numbers say with the traditional knowledge based upon experience, we can redefine what making the “safe play” really looks like. There are times during a game, where it is simply impossible to carry out a controlled zone exit so obviously we cannot simply do away with dumping the puck out of the zone; however, the definition of “impossible” needs to be the same for everyone. By combining traditional and statistical analysis, we can create a checklist for players that guides their actions when it comes to dumping the puck out.
Player has been defending in his own zone for a while and knows it is time for a line change. He gains possession of the puck: