Why linemate and competition metrics may not be as simple as we think


Competition Histograms by Eric Tulsky from NHL Numbers, Sep 23, 2012


We know that linemates have a larger impact on results than competition on the average. This has caused many to change player deployment chart axis from QoC to QoT metrics.

However, it’s not quite that simple.

The area of contextually nuanced studies with numbers like competition and teammate metrics is still well in its infancy. We have a general idea of what’s going on but there is a lot of information in the details.

We show here that a 1 percentage point change in teammate and competition Corsi% has an equal but opposite impact on observed output, but there are some differences. The distribution in the NHL is much smaller with competition. However, unlike with competition Corsi%, teammate Corsi% impact is not the same for all players.

Continue reading

Hockey Graphs Top 50 NHL Players

Before the season begins there’s two lists that seem to cause a stir within the hockey community: the Top 50 Players in the NHL. The Hockey News puts out one in its annual season preview yearbook while TSN has an entire hour-long broadcast dedicated to it. Both lists are compiled in the same way (which is why they tend to be similar), and that’s via a poll of people inside hockey. That may be where some of the controversy lies.

It’s not necessarily that those guys are wrong about who’s the best of the best – it’s their job after all – it’s that their opinions tend to be moulded by a few biases that cloud their judgement. From looking at the list every year (and how it changes) it’s shaped a lot by recency bias, reputation and a winning pedigree.

I wanted a different take on the debate so I enlisted some of hockey’s top nerds doing work in the public sphere to share their opinions of who they think will be among the 50 best players in the league in this upcoming season.

Continue reading

2015 Hockey Graphs Standings Predictions

As a loyal reader of Hockey Graphs you may be aware that today marks the start of the 2015-16 NHL season. This is an uncertain time in many hockey fans’ lives, and you probably have questions: Will my favourite team make the playoffs? Who is going to win the Stanley Cup? Are the Leafs really going to be that bad again?

As standings-list-compiler-in-chief of Hockey Graphs (we almost went with Listicleditor in Chief but thought it was too wordy), my job is to answer those questions for you. I have made the process of knowing who is going to make the playoffs, win their division, and play for the cup easy! So easy in fact, that even a Habs fan could get it (the process, not the Cup, they’re definitely not going to win the Cup). Simply consult the sharply presented table below and you (probably) won’t even need to watch the games! Isn’t that simpler and more fulfilling?

Continue reading

Hockey Graphs live video (Pod)cast #1: Talking about RITHAC and Manually Tracking Hockey Games!

For those who missed it, below is the archive from tonight’s Hockey Graphs Live Video Podcast #1, featuring Garik16, Ryan Stimson, Ben Wendorf, and DTMAboutHeart!  We talked about the upcoming Rochester Institute of Technology Hockey Analytics Conference (RITHAC), Neutral Zone and Pass Tracking, and how we go forward with deciding what to track and what to look at with such data.  Give us a listen and if you have any thoughts for what we should talk about next cast, please leave a comment!

Practical Concerns: How I do video

Video is the best teaching tool there is.

Video is the best teaching tool there is.

Preparing and organizing game footage is one of my main responsibilities working for the McGill Martlet hockey team, and has become something that I enjoy quite a bit over the course of the past two seasons. Having played for coaches who use video analysis to various degrees in both hockey and tennis growing up, I think seeing one’s self play sports on video is the best way to correct deficiencies and identify areas for growth.

Continue reading

Hockey Graphs live video (Pod)cast #1 : Tuesday 10/6/15 at 9PM EST!

This Tuesday at 9PM EST, we’ll be holding our first official Hockey Graphs live video podcast. This is a live event, so if you want to chat with some of the Hockey Graphs guys or ask some questions to us, you can do that and we’ll try and answer! (If you’re on mobile, you can chat via the youtube ap, apparently). And if you have some topics/questions you’d like us to discuss that you know of in advance, please feel free to leave a comment to this post and we’ll try to discuss in the actual live cast!

The stream will be featured here: http://www.youtube.com/user/wraithlead8/live

We did a preview stream for about an hour on Saturday, so you can see what that looked like below:

Hope to see you then!

Neutral Zone Score Effects: How does the game score affect play in the Neutral Zone?

This is the type of entry most likely to be made by a leading team – an odd man rush Carry-In against a d-man conceding the zone (the legendary Andrew MacDonald). Image courtesy of http://www.BroadStreetHockey.com

We know a decent amount about score effects in the game of hockey at this point. We know the obvious: teams play “conservatively” and shell up when leading and become more aggressive when trailing, resulting in the leading team taking less shots and the trailing team taking more.

Less obviously (at first glance anyhow), We also know that the leading team’s shooting % increases substantially when compared to when that same team is tied (Graph courtesy of new Hockey-Graphs writer Petbugs), while the trailing team’s shooting % either increases barely or stays the same as it does when tied.

As detailed earlier on this site this results in the trailing teams tending to score more than leading teams, especially in late.

But what we haven’t really detailed is some of the mechanics of “How” this happens. Some have suggested its because teams may change their systems with a lead or deficit. Others, such as Justin Bourne , have suggested the effect is mainly the result of psychological effects – no one wants to take the risky play when leading and possibly cause the opponent to get the tieing goal as a result of a risk, so players stop taking as many risks when leading even though the coach is preaching the same scheme. But little attention has been paid to seeing if any of these things are true, whether via a statistical analysis or via some systems breakdown.

So what I’m going to do in this post is attempt to break down how score effects change how the neutral zone is played. And by doing so, we can by extension, try to get a better picture of how the play in the other two zone is affected by score effects.

Continue reading

Practical Concerns: Fixed vs. growth mindsets in hockey analytics


The vast majority of scientific research has indicated that people with growth mindsets tend to be higher-achieving than those with fixed mindsets. Growth-oriented people are the ones who are more interested in furthering their education, or picking up a new hobby, or getting out of their comfort zones to experience new things.

Continue reading

Expected Goals are a better predictor of future scoring than Corsi, Goals


This piece is co-authored between DTMAboutHeart and asmean.


Expected goals models have been developed in a number of sports to better predict future performance. For sports like hockey and soccer where goals are inherently random and scarce, expected goals models proved to be particularly useful at predicting future scoring. This is because they take into account shot attempts, which are better predictors of a team and player’s performance than goal totals alone. 

A notable example is Brian Macdonald’s expected goals model dating back to 2012, which used shot differentials (Corsi, Fenwick) and other variables like faceoffs, zone starts and hits. Important developments have been made since then in regards to the predictive value of those variables, particularly those pertaining to shot quality.

Shot quality has been the subject of spirited debate despite evidence suggesting that it plays an important role in predicting goals. The evidence shows that shot characteristics like distance and angle can significantly influence the probability of a certain shot resulting in a goal. To date, no formal attempt has been made to account for shot quality in an expected goals model that we know of.

In Part I, an updated expected goals (xG) model will be presented that accounts for shot quality and a number of other variables. Part II will deal with testing the performance of xG against previous models like score-adjusted Corsi and goals percentage.

Continue reading