1 year ago
The goal of advanced sports metrics is to more accurately predict future performance. Aside from the xG model, which you have probably heard about, hockey has its own pool of advanced stats. All of these stats can be implemented both on the team and player level, but in regards to DFS-hockey, teams’ stats are used more often. In this article, I will cover the main terms from the world of advanced hockey analytics and will try to explain how to use them to your advantage in DFS.
Before we begin, we should make some things clear. You don’t have to be a hockey expert to know that there are different situations in hockey, such as even strength, power play, penalty kill. In order to provide proper context for performance, advanced metrics are used only in evaluating play during certain situations. As about 80% of the total minutes in the NHL are played at even strength, indicators of advanced metrics when 5 on 5 are the most useful for us.
Secondly, in order to truly measure scoring efficiency, we should make a «per 60» adjustment. This is a standard method when evaluating efficiency over big sample sizes, where the amount of ice time played always differs, thus estimating just raw counts will always yield skewed results.
Corsi is the fundamental concept in advanced hockey statistics. Corsi For (CF) includes shots on goal, missed shots on goal, and blocked shot attempts towards the opposition's net. Corsi Against (CA) includes the same shot attempts directed at team’s own net.
Let’s look at the CF/60 and CA/60 stats of the NHL teams after the first half of the season:
As we can see, Montreal and Toronto were the best teams in generating shot attempts. Rangers and Ottawa were the worst in allowing them.
By combining Corsi For and Corsi Against, we get Corsi For Percentage (CF%). This is the ratio that shows the percentage of shot attempts created by a team relative to the total shot attempts in a game during 5-on-5 situations.
CF percentage of 52% and above indicates that a team is very good at controlling the flow of a game (such teams are presented in the upper right corner of the chart). 50-52% CF percentage is solid, 48-50% is weak. Everything below 48% is poor, and you can find these teams in the lower-left corner of the chart.
Overall, Corsi is a good indicator of the team’s ability to «drive play», but it’s obvious that all shots are different and outshooting the opponent doesn’t necessarily mean that the shots fired are of good quality. These problems are addressed by the following two metrics.
High-danger chances are a more precise metric to evaluate the quality of chances created. Some of you are familiar with the «big chance» term in football, and this is basically a hockey equivalent. High danger scoring chances for (HDCF) are the team's shots from certain areas on the ice that have a high likelihood of the puck going into the net. High danger scoring chances against (HDCA), same as CA, are the same shots attempts directed at team’s own net.
As far as DFS-hockey is concerned, we have more interest in the teams’ HDCA rather than their HDCF. In general, we should pick players that are playing against the teams with the worst HDCA/60 numbers, and try to avoid picking those facing teams with solid HDCA/60.
This concept is used not only in hockey and is easy to explain. The xG model assigns a particular weight to every unblocked shot, based on the likelihood of the shot resulting in a goal. A shot from within the face-off circle may have an xG value of 0.2, while a wrist shot from the point might only be worth 0.02.
The xG model helps us to estimate how many quality chances each team creates and allows. Expected goals and High danger scoring chances are closely correlated and can be used almost interchangeably.
The worst NHL teams in terms of HDCA/60 and xGA/60 over this season:
|1||New York Rangers||2.72||45.38||11.22||12.72|
|4||San Jose Sharks||2.37||47.81||10.79||11.48|
|5||New York Islanders||2.26||48.32||11.13||11.33|
|10||Toronto Maple Leafs||2.31||50.70||10.73||10.84|
* Source: naturalstattrick.com
Using CA/60, xGA/60, and HDCA/60 in tandem, will help you to identify the teams you most likely should attack in the slate. It should be mentioned, that in order to get more representative results, you should restrict your sample. I prefer analyzing the data which relates only to the last month.
The purpose of PDO is to determine if a team has either been lucky or unlucky in the previous contests. PDO is a sum of a team’s shooting percentage (the percentage of shots on net taken by a team that resulted in a goal) and it’s save percentage (the percentage of shots on team’s own net stopped by goalie), the NHL average will be right at 1.000.
A team that has a high PDO percentage can indicate that the team is overperforming (getting lucky) and is ready to regress towards the mean. A team that has a low PDO percentage can indicate that the team is underperforming (getting unlucky).
We can see the teams’ PDO in the following chart:
Colorado (1.027) and Boston (1.022) have the highest PDO numbers after the first half of the season. The reason for that is excellent goaltending and high shooting percentage. San Jose (0.974), Detroit (0.968), and Los Angeles (0.976) are the unluckiest ones. San Jose struggled from extremely poor goaltending, while the Red Wings and the Kings were bad in regard to shot conversion rate.
It should be said, that PDO is used less often in regard to DFS. The model is more long-term by its nature, and the goaltending quality affects the numbers significantly as well. However, what we can expect is some sort of regression in the teams’ shooting percentage towards the NHL average.
In this article, we covered the basic terms from the advanced hockey analytics. We discussed them on the team level, but you can do your own research analyzing these metrics in regard to individual players and the lines. The methodology should be the same. What you should remember as well is that hockey is a very random sport by its nature. Things we talked about today will not make you successful in DFS-hockey overnight. However, most players don’t consider them when creating their lineups, and that might give you an edge in the long run.
Khan, a Ph.D. student in Economics, loves studying advanced sports analytics and using it to his advantage in DFS. Having started to play daily fantasy sports in 2017, he currently has a net profit of over 70000 euros. Sits in the top-5 of DFS players in Europe, finished 3rd in FanTeam's WCOFF 2018 and WCOFF 2020.