Tuesday, May 26, 2015
Authors Posts by David Johnson

David Johnson

Toronto Maple Leafs' new head coach Mike Babcock laughs during a press conference in Toronto on Thursday, May 21, 2015. THE CANADIAN PRESS/Darren Calabrese

There are potentially a number of impacts that the Mike Babcock signing could have on the Toronto Maple Leafs, including off-ice impacts such as free agent recruitment.

We already heard yesterday that Cody Franson believes that Babcock makes Toronto potentially more attractive to free agents, including himself, as a result of the credibility he brings to the team as well as the relationships he has built with players over his years in the NHL and with Team Canada. The question I will try to answer today: What will his impact be on the ice?

DALLAS, TX - JANUARY 23: Tyler Bozak #42 of the Toronto Maple Leafs takes a faceoff against Tyler Seguin #91 of the Dallas Stars at the American Airlines Center on January 23, 2014 in Dallas, Texas. (Photo by Glenn James/NHLI via Getty Images)

A significant portion of modern hockey analytics revolves around Corsi (or SAT% as defined by the NHL), which is really nothing more than looking at which team takes more shot attempts.

If you can out shoot your opponent, the theory is that it goes a long way to driving success in terms of out scoring your opponent and ultimately winning games. There is a lot of evidence to support the case that Corsi is a major component of on-ice success. While I believe many people put too much weight on Corsi statistics, I do accept that it is a major component of success.

TORONTO, ON - DECEMBER 14: Tyler Bozak #42 of the Toronto Maple Leafs battles for the puck with Brayden McNabb #3 of the Los Angeles Kings as teammate Jonathan Quick #32 defends the goal during NHL game action December 14, 2014 at the Air Canada Centre in Toronto, Ontario, Canada. (Photo by Graig Abel/NHLI via Getty Images)

Last week, I took at look at zone start changes from Randy Carlyle to Peter Horachek and found some interesting differences.

Among them was a flip in zone starts by David Booth, going from a heavy defensive-zone player under Carlyle to a heavy offensive-zone player under Horachek. Another significant observation was the massive improvement in the first line’s Corsi from near league-worst levels (low 40s) to being more than respectable, particularly for Tyler Bozak, who had a CF% of over 53%. Booth also saw a massive improvement in his CF% from Carlyle to Horachek. This week, I look at WOWY stats for these two players to see if it shines any more light on the situation.

Tyler Bozak

This first chart takes a look at who Bozak played the most ice time with under Carlyle and now under Horachek.


As expected, the majority of Bozak’s ice time is with JVR and Kessel, and while there has been a slight drop off in ice time with JVR, there isn’t a huge difference here. A more significant change is the defensemen he has been on the ice with. He has spent far less time on the ice with Phaneuf (partly due to injury) and Robidas under Horachek, and more with Gardiner, Polak and Rielly.

Now let’s take a look at the change in CF% from Carlyle to Horachek:


Aside from Lupul (who has a tiny sample size under Horachek), Bozak’s CF% has improved dramatically with everyone, which is not unexpected but is a clear indication it is the top line that is driving the improvement in CF%, not due to playing more with good puck movers like Rielly and Gardiner. The top line is doing something right from a possession perspective, but thus far that hasn’t translated to winning the goal scoring battle.

David Booth

Here is David Booth’s ice time-with-teammate chart:


Not a lot of change, really. Less ice time with Phaneuf (again, partly due to injury) and Rielly, and more ice time with Polak and Franson. Among the forwards, he had a little less ice time with Smith and a bit more with Santorelli (not shown due to limited ice time with under Carlyle).

How about Booth’s CF% WOWY comparison?


Big CF% jumps with Smith, Gardiner, Phaneuf, Clarkson and Franson, with only Rielly seeing a significant CF% drop.

While Bozak’s improvement appears to be largely a change in playing style by the top line, it is quite likely that Booth’s improvement in CF% is likely in part due to more ice time in offensive roles. When he played with offensive defensemen like Gardiner, Phaneuf and Franson, he saw significant boosts in CF%, which is almost solely driven by generating more shot attempts (under Carlyle Booth had a CF60 of 42.8 compared to 58.2 under Horachek). This is the opposite of Bozak, whose dramatic improvement in CF% is in large part due to reducing shot attempts against (although the top line has improved shot attempt generation as well). Bozak’s CA60 was 71.0 under Carlyle compared to 54.9 under Horachek.

The two things that seem to have changed the most from Carlyle to Horachek? David Booth is being given a significantly larger offensive role and the top line of Kessel, Bozak and JVR have dramatically reduced the rate at which they give up shot attempts against. Clearly this hasn’t yet translated to success on the score sheet, so as of yet neither of these adjustments has worked out, but there are certainly a number of people in the Maple Leaf organization who are hoping that the changes will eventually pay dividends.

Randy Carlyle Peter Horachek
GLENDALE, AZ - NOVEMBER 04: Head coach Randy Carlyle of the Toronto Maple Leafs looks on from the bench during third period action against the Arizona Coyotes at Gila River Arena on November 4, 2014 in Glendale, Arizona. (Photo by Norm Hall/NHLI via Getty Images)

Under Randy Carlyle, the Maple Leafs were a test case as to whether a team can win despite being dramatically out shot largely through controlling shot quality, both for and against.

There is ample evidence to suggest that under Carlyle the Leafs were in fact an above average shooting percentage team and may even in fact have been able to suppress shot quality against to some degree. Unfortunately, their ability to be significantly out shot outweighed any ability to have a positive impact on shot quality, and they were at best a mediocre team.

Randy Carlyle and Mike Babcock walk out into "The Big House' - 2014 Bridgestone NHL Winter Classic on January 1, 2014 at Michigan Stadium in Ann Arbor, Michigan.

The other day, over at my hockey analytics blog, I wrote an article on the relationship between Possession/Corsi (CF%) and Shooting Percentage (Sh%) in 5v5 close situations.

I figured I’d piggy back on that analysis a little and take a look at the Leafs over the past three and a half seasons spanning the Randy Carlyle era.

In my HockeyAnalysis.com article, I showed that, while some elite level teams or truly bad teams can break the trend, there is generally a strong negative correlation between a teams CF% and shooting percentage. Recall that CF% stands for Corsi For Percentage, which is the percentage of all shot attempts taken, for and against, that the team itself took (or shot attempts for divided by shot attempts for plus shot attempts against). So, what this is saying is that better possession teams generally take a hit in terms of shooting percentage. There are some teams that are reasonably good at both (Pittsburgh, Chicago) and there are some teams that are terrible at both (Buffalo), but generally speaking good Corsi teams are weak shooting percentage teams.

I wanted to take a look at how Corsi and shooting percentage compared during the Randy Carlyle era, so I produced the following chart to show just that.


The chart above is for the past 4 1/2 seasons using all 5v5 data up through Saturday’s game against the St. Louis Blues. The black lines denote breaks between seasons, the red line shows when Carlyle was hired, and the green shows when he was fired. This is a rolling average of 1000 Corsi (for and against) events.

There is in fact a weak negative correlation between the two statistics (correlation is -0.24) which supports my previous findings. It should be noted that there has been fairly significant roster turnover during this time (Grabovski, MacArthur, Kulemin, etc., out, Clarkson, JVR, Winnik, Santorelli, etc., in) so the correlation may not be as strong as if there were more roster stability.

Here is a summary of the last 4 1/2 seasons pre-Carlyle, under Carlyle and post-Carlyle.


Everything points to the Leafs being a weaker Corsi team under Carlyle but a better shooting percentage team. When Carlyle took over, the team’s CF% immediately started dropping and their shooting percentage immediately started improving. The Leafs shooting percentage dropped off last year from 2012-13 levels, but that is likely a result of losing good on-ice shooting percentage players in Grabovski and MacArthur. The Leafs have also been a significantly better possession teams under Peter Horachek and their shooting percentage has tanked. Yes, the sample size is incredibly small and a 3.6% shooting percentage is unusually low, but one should probably not expect it to rise back to Randy Carlyle levels. All evidence suggests if the Leafs improve their possession game it will come at the cost of shooting percentage. The hope is the improved possession game results in more consistency and provides more benefit than the cost of the hit to shooting percentage, but Leaf fans might want to get used to more low scoring games like we have seen the past week or so. What it also means is that real significant improvement will likely require roster changes, not just coaching and playing style changes.


NEW YORK, NY - MARCH 05: Tyler Bozak #42 of the Toronto Maple Leafs (R) celebrates his goal at 1:51 of overtime along with Phil Kessel #81 (L) to defeat the New York Rangers 3-2 at Madison Square Garden on March 5, 2014 in New York City. (Photo by Bruce Bennett/Getty Images)

The last three games have really highlighted the Leafs defensive woes, but the problem has been an ongoing one and a significant portion of the blame must be given to the Leafs top line.

On Puckalytics.com, you can find what I call “Percent of Team” statistics, which are nothing more than the percentage of all events by the team that the player was on the ice for (only counting games in which he played in). So, for example, we can see that Ryan Suter leads the league with a %ofTeam TOI of 45.44% this season, which means Suter has been on the ice for 45.44% of his team’s 5v5 ice time. The Leafs leader in %ofTeam TOI this season is Roman Polak at 36.55%, whereas a guy like Trevor Smith is at 16.36%.

TORONTO, ON - DECEMBER 9: James van Riemsdyk #21 of the Toronto Maple Leafs skates the puck away from Kris Russell #4 of the Calgary Flames during NHL game action December 9, 2014 at the Air Canada Centre in Toronto, Ontario, Canada. (Photo by Graig Abel/NHLI via Getty Images)

I didn’t have any particular topic I wanted to write about this week, so yesterday morning I asked my twitter followers what they wanted me to write about as far as a statistical analysis of the Leafs is concerned.

I’ll tackle a few of those questions today.

DENVER, CO - NOVEMBER 6: Tyler Bozak #42 of the Toronto Maple Leafs skates against Jamie McGinn #11 of the Colorado Avalanche at the Pepsi Center on November 6, 2014 in Denver, Colorado. (Photo by Michael Martin/NHLI via Getty Images)

In the summer, I wrote about Bozak’s statistically-improved 2013-14 season.

In that post, I observed that Bozak had significantly improved his 5v5 offensive production last season, and in particular his assist rates, while noting that his power play production had been relatively unchanged. Here is my concluding paragraph from that post:

TORONTO, ON - NOVEMBER 8: Peter Holland #24 of the Toronto Maple Leafs celebrates his goal against the New York Rangers during NHL action at the Air Canada Centre November 8, 2014 in Toronto, Ontario, Canada. (Photo by Abelimages/Getty Images)

The Maple Leafs just surpassed the quarter pole of their schedule with a win over the Detroit Red Wings in game #21 of the 2014-15 season. Last week I looked at how individual players are performing. This week, let’s take a look at how the team is performing as a whole.

It has been a bit of a roller-coaster season to say the least, especially the past 8-10 days. A mediocre start —  2-3-1 in their first 6 games — turned for the better in a big way as the Leafs went 7-2-1 in their following 10 games. Then, everything went haywire: They lost their next three games, two of them in an embarrassingly bad way, including a 6-2 loss to the league’s worst team in the Buffalo Sabres and a 9-2 loss to the Nashville Predators  (who haven’t scored more than 4 goals in any other game). In true inconsistent Maple Leaf fashion, they followed up those two embarrassing losses with two big wins against a pair of good teams in Tampa Bay and Detroit.

Despite the season being all over the map in terms of consistency, after 21 games the Leafs sit at a respectable 11-8-2 record. This is behind their pace of last season after 21 games (13-7-1), but it still puts them in the middle of the playoff hunt.

What about the underlying numbers, though? How do they look? We saw that there weren’t too many surprises at the player level in my post last week, so does the same hold true at the team level? To investigate, let’s look at some charts comparing the first 21 games of the 2014-15 season with Toronto’s full-season results from the past three seasons. Let’s start by looking at 5v5 GF/60 and 5v5 GA/60.


Both GA/60 and GF/60 have risen from last season, although the increase in GF60 is more significant, meaning the gap has closed. That said, the Leafs still have a negative goal differential at 5v5, which is troubling. How about Corsi?


This chart is all good news for Leaf fans. The Leafs have, thus far, reversed a trend of increasing shot attempts against and decreasing shot attempts for. Bringing in some good Corsi players like Winnik and Santorelli and getting rid of some poor ones like McClement and Bolland are probably the most significant factors here. All that said, the decrease in shot attempts against has not resulted in decreased goals against rate. Is this due to bad luck or is something else happening here? More on this in a bit; let’s look at GF% and CF% first.


Here we see the Leafs overall goal and Corsi differential expressed as a ratio. In both cases the Leafs are below 50%, leaving ample room for improvement, but in both cases the situation has improved from last season. All in all, there are some positive signs thus far this season.

Next, let’s look at the percentages,  which is where things may get a little more troubling for Leaf fans, particularly with save percentage.


Although I expressed doubts that the Leafs would be able to maintain an above average shooting percentage, thus far they seem to be doing so, even showing a bit of an improvement over last season. That said, in last week’s article I identified several players (Winnik, Komarov, and Santorelli specifically) who had on-ice shooting percentages well above their career averages; I still wonder if their current shooting percentage is sustainable. Time will tell.

With respect to save percentage, the teams overall 5v5 save percentage has dropped significantly this year to 91.43% from 92.84% last year. Is this unusually low or should we expect a bounce back? To answer that, let’s look at this shocking chart comparing changes in CA/60 with changes in Save%.


The chart is shocking due to the incredibly high correlation (r^2=0.99). Over the previous 3 seasons, the Leafs saw their team 5v5 save percentage increase in lockstep with their corsi against rate. Now, as their corsi against rate improved this season, their save percentage has declined in lock-step. I have charted a tonne of hockey stats over the years and it is incredibly rare to ever see a chart so ‘perfect’ as this one above. Hockey analytics is typically messy and incredibly uncertain, so when I see a chart like above it really stands out.

So, what is this telling us? If we assume that goaltenders are not a factor (they are, but let’s assume not for now, or at least we’ll assume the quality of goaltending talent hasn’t change significantly for the Leafs over the past 4 seasons), it is telling us that it is possible to give up more shots but of lower quality (easier to save, thus a higher save percentage). It seems that teams can impact shot quality against.

Let’s look at the opposite side of the game and see if the Leafs’ shots attempts for is in any way correlated with their shooting percentage.


It’s difficult to see here, but there is in fact a bit of a negative correlation (with r^2=0.13).That is not all that surprising given the Leafs forward group has undergone a significant overhaul the past four seasons, affecting the underlying shooting skill. There may be nothing to any of this, but we can find similar trends for other teams, especially those with relative stability over the years.

For the Kings, the r^2 between CF60 and Sh% is 0.57 and the r^2 for CA60 and Sv% is 0.55. The Bruins are 0.79 and 0.32 if you include the past five seasons. The Rangers are 0.83 and 0.54 over the previous four seasons, excluding this season. There may be a dynamic between shot quantity and shot quality that needs more exploration, but certainly the chart of the Leafs CA/60 and Sv% above is interesting.

On a final note, in the summer I did an investigation into rush shots where I identified a shot being rush shot based on whether an event (faceoff, shot, hit, etc.) had recently occurred in the opposite end or neutral zone. I determined that rush shots are generally more difficult shots: Players scored on 9.56% of rush shots and just 7.34% of shots that we cannot conclusively determine to be shots on the rush from the game logs.

What is interesting is that the percentage of the Leafs’ shots against that are rush shots has skyrocketed so far this season. The past three seasons saw 22.8%, 19.3% and 23.8% of all shots against coming off the rush. This season, that has sky rocketed to 28.9%, and they’re on pace to give up about 13% more rush shots this season despite reducing the overall shots against. This likely means Leaf goalies have faced a higher quality of shot thus far this season which likely explains (at least in part) their drop in save percentage this year.

To summarize, we can definitely observe some positive improvements in the Leafs first 21 games this season, most notably reducing the shots attempts against and boosting the shot attempts for, although there is definitely still ample room for improvement. On the downside, it seems the Leafs are still giving up a number of higher quality rush shots against, indicating they are still far from a defensively-reliable team.


COLUMBUS, OH - OCTOBER 31: Phil Kessel #81 of the Toronto Maple Leafs skates with the puck as Tim Erixon #20 of the Columbus Blue Jackets skates back in defense during the second period on October 31, 2014 at Nationwide Arena in Columbus, Ohio. (Photo by Jamie Sabau/NHLI via Getty Images)

The Maple Leafs are now 18 games into the 2014-15 season, a decent enough sample size to take an early advanced stat look at the Leafs performance so far.

In the off season, the Leafs overhauled much of their front office and jumped on the hockey analytics bandwagon in a big way, although they left the general manager and head coach in place. Player wise, the majority of the third and fourth lines also changed along with a pair of their defensemen (Polak, Robidas). The question is: Has anything changed in terms of playing style, tactics, or results? This week I will take a look at the individual players before breaking down the team more generally next week in an attempt to answer that question.

Jake Gardiner
Photo: NHLI via Getty Images

In my last post, I looked at which Maple Leaf players are producing their fair share of the offense. In the comments to that post, a few readers wanted me to take a look at the players defensively to see who is performing up to par in that aspect of the game. Today, I’ll look at the defensive side of the game in conjunction with their offensive performance.

It may be worth while to go back and read it now if you haven’t yet read the previous article, but I’ll remind you of the process below: 

by -
David Booth of the Toronto Maple Leafs scores the winning shoot-out goal on Anthony Stolarz of the Philadelphia Flyers for a 3-2 victory in pre-season game action, Monday September 22, 2014 in London, Ontario. THE CANADIAN PRESS//Dave Chidley

If you are not yet aware, I have recently launched a new hockey statistics website at Puckalytics.com. Long-term this will be the replacement for stats.hockeyanalysis.com, but not all features are implemented yet; for now, both will remain operational. There are some new statistics available at Puckalytics.com and one of those features is the ability to find the percentage of a team’s overall statistic that the player was on the ice for. I call these ‘% of Team Stats’ and they can be found here.

Toronto Maple Leafs captain Dion Phaneuf, right, reacts after scoring the game winning goal as Tampa Bay Lightning defenceman Victor Hedman, left, look on during over-time NHL hockey action in Toronto on Thursday, April 5, 2012. THE CANADIAN PRESS/Nathan Denette

In May of 2013, I wrote an article on Dion Phaneuf and his play when the team is leading vs when they are trailing. It presented a pretty compelling case that Phaneuf is a significantly better player when the Leafs are trailing, a situation in which he is given a more offensive role.

Toronto Maple Leafs v Boston Bruins - Game Two
BOSTON, MA - MAY 4: Phil Kessel #81 of the Toronto Maple Leafs celebrates with teammates Nazem Kadri #43 and Ryan Hamilton #48 after scoring a goal in the third period against the Boston Bruins during Game Two of the Eastern Conference Quarterfinals during the 2013 NHL Stanley Cup Playoffs at TD Garden on May 4, 2013 in Boston, Massachusetts. (Photo by Jim Rogash/Getty Images)

My first post here at MLHS was an Introduction to Advanced Hockey Statistics, which I recommend you go read if you have not done so yet. In it I introduced the concept of WOWY statistics and WOWY analysis in player evaluation. In this post I am going to look into this deeper with examples and explain where and how you can get the data yourself.

What are WOWY statistics?

WOWY stands for With or Without You and essentially looks at pairs of players and how they perform together and how they perform apart. WOWY statistics typically consists of three groups for every pair of players. If we have players P1 and P2 then the three groups are statistics when P1 and P2 are on the ice together, statistics when P1 is on the ice without P2 and statistics where P2 is on the ice without P1. The statistics one looks at can either be goal-based statistics (i.e. goals for/against, goals the player scored, etc.) or corsi/shot based statistics (shot attempts for/against, shots taken by the player, etc.). In theory any statistic could be looked at in a WOWY analysis but the statistics that are commonly available are GF20, GA20, GF%, CF20, CA20, and CF%.

Why use WOWY Analysis?

WOWY is important because good players should be able to make their teammates better. So, if the statistics show that your teammates consistently perform better when they are playing with you than when they are not it is probably a good sign that you are doing something well when you are on the ice. Conversely, if they are performing consistently worse with you than apart from you it is a pretty strong indication that you are holding back your teammates.

Ken Hitchcock, coach of the St. Louis Blues, has mentioned several times that he uses WOWY or WOWY-like statistics to help him identify matchups and players who work well together. In a recent ESPN article Hitchcock mentioned using WOWY statistics (though he did not specificy use the terminology WOWY, that is essentially what he was using).

My [focus] is in matchups and combinations and chemistry,” said St. Louis Blues coach Ken Hitchcock. “I get information on — ‘Is a matchup working or not? Is the chemistry working?’ I get it where you’re starting guys on the ice, is it the right place to start? I don’t get in on all the puck possession stuff. I take pride in understanding that stuff, but I use it on matchups. I use it between periods.

“Chemistry is huge. It tells you basically how you’re coaching,” Hitchcock said. “For instance, last year there were three players I played together that I thought had good chemistry. The data showed me otherwise. When I looked back at their shifts, the data was right. I kept putting it back together thinking it was working or would work, but it didn’t work.”

I don’t know where Hitchcock gets his WOWY stats from (my site or internally produced databases) but it is good to see NHL coaches understanding the value in it.

WOWY can be used in player evaluation too. Recently Sportsnet televised a half-hour show called A Numbers Game which they looked at Jake Muzzin of the Los Angeles Kings and utilized WOWY to show that he is probably more important to the Kings than most people might believe. Muzzin is a perfect example of advanced statistics identifying an undervalued and under rated player who probably deserves a lot more credit for his teams success.

Where do I find WOWY numbers?

For this post I am going to use new Leaf Roman Polak as my sample player. I think he is a good example of where a WOWY analysis may have given Leafs management second thoughts when making the trade for him.

The first thing we need to do is figure out how to get Polak’s WOWY statistics. The only (publicly available) source for WOWY statistics is my site stats.hockeyanalysis.com. You can only access WOWY pages through each player’s player page. The easiest way to get to a player page is to click on the ‘Players’ button at the top of the page (see red circle in image below). This will give you a master list of all the players and you can search for. Click on their name to get to their player page.


The alternate method to get to a player page is to conduct a player statistics search using the search options found in the green ellipse in the image above. When the search is complete, you can click on any player names within the search results to get to their player page.

Once you get to Roman Polak’s players page, you will find a boatload of statistics; to access WOWY data, you are only interested in the table of links at the top of the page as shown in the image below in the red box:


As you can see, there are a lot of links there. I have made WOWY statistics available for 5v5, 5v5 Zone Start Adjusted, and 5v5Close Zone Start Adjusted situations and they are available for every single season or multi-season combination for each of the past 7 seasons. Clicking on 2013-14 under 5v5 with take you to Roman Polak’s 2013-14 5v5 WOWY stats page. At the top of the page you will see the same links to all the WOWY stats as well as a link below it to take you back to the player stats page we just came from. Below this there are actually 3 tables on but the WOWY stats are in the first one, so let’s take a look at that (I’ll get to the second and third tables in a bit).


I have put colored boxes around each section to make it easier to explain what each section is. Click the image to get a larger version of it or simply go to Polak’s 5v5 WOWY page for last season to see the real thing.

The red box contains the names and position of every player Polak played with for at least 5 minutes last season sorted by the number of minutes they played together. Note that clicking on any of the players names in this table will take you to their player page, which in turn you can go to their WOWY pages.

The light green box shows Polak’s individual statistics when playing with each particular player. The first row in the table is Polak himself so these are Polak’s overall numbers during 5v5 play last season. He had 4 goals, 7 assists, 11 points on 77 shots on goal and 149 shot attempts.  The second row are his stats when playing in front of Jaroslav Halak, while the third row are his statistics with his main defense partner Ian Cole.

While the light green box are individual statistics, the rest of the table are ‘on-ice’ statistics. The blue box are on-ice statistics when both Polak and his teammate are on the ice together. The dark green box holds Polak’s statistics when Polak is on the ice without his teammate. The black box contains his teammates statistics when they are on the ice without Polak.

Let’s take Ian Cole as an example since he was Polak’s main defense partner. The first row shows that Polak played 1044:29 of 5v5 ice time last season. The third row, Ian Cole’s row, shows that Polak and Cole were on the ice for 413:22 together. That means Polak played 631:07 apart from Cole and Cole played 224:05 away from Polak.

So, how did they perform together and apart? Well, together they posted a CF% of 49.1%. When Polak was playing without Cole on the ice with him he had a CF% slightly better at 49.4%. When Cole was on the ice without Polak his CF% was 49.3%. Okay, this is a pretty bad example because they essentially played the same with and apart from each other.

If we jump down several rows to Polak’s second most frequent defense partner Barret Jackman we see a slightly different story. Together they had a CF% of 49.9%, Polak without Jackman had a CF% of 49.0%, while Jackman without Polak was a much better 54.1%. One player does not make a trend, but this would suggest that Jackman is a better defenseman since Polak has better statistics with him and Jackman has significantly better stats without Polak than with Polak.

As I said though, one player does not make a trend. What we want to do is look up and down the list to see if Polak is consistently pulling down his teammates statistics. There are 16 forwards or defensemen who played at least 100 minutes with Polak last season. Of those 16 players, only Chris Stewart and Vladimir Sobotka had a better CF% when playing with Polak than playing apart from Polak. Although I’d still be cautious drawing conclusions about a player based on just one season of CF% WOWY data, this is not making Polak look very good. For a complete WOWY analysis one would probably want too look at multiple seasons and look at GF% data as well (unfortunately for Polak -and Leaf fans- what we see in 2013-14 data is pretty similar to previous seasons as well).

You may have noticed a ‘Visualize this table’ link at the top of the table. This takes you to a series of charts that chart teammates with Polak vs without Polak statistics and I think is a quick way to get a feel of Polak’s impact on his teammates. Here is the CF% chart for Polak’s 2013-14 WOWY stats to show you visually the data just discussed.


I have added in the diagonal red line to make the discussion easier, but unfortunately you have to visualize this 1:1 line on my site. The 1:1 diagonal line indicates the point where a players statistics are equally good (or bad) with Polak and apart from Polak. If players have a CF% better with Polak than apart, their bubble will be below or to the right of the diagional line. If the players have a CF% worse with Polak than apart, their bubble will be above or to the left of the diagonal line. The size of the bubble indicates the relative amount of ice time they played together (large bubbles indicate more ice time with Polak). As you can see, the majority of the players are to the upper left of the line, indicating they have better statistics without Polak than with Polak. To see that this is a trend continuing from previous seasons, have a look at 2012-13 and 2011-12 charts.

Words of Wisdom

WOWY’s are a useful tool, but one has to be cautious in using them as player usage and other factors may come into play. Some points to remember are:

  • Goal WOWY stats (GF20, GA20, GF%) are highly variable due to the relative infrequent nature of goal scoring. Use goal WOWY’s carefully and definitely don’t draw any conclusions from a single season WOWY for goal data. Look for longer term trends. Using 3 or 4 year data can make the goal based WOWY stats significantly more useful and will provide you with much greater confidence in what they are showing.
  • Some players play multiple significantly different roles and that can have huge impacts on WOWY’s. Bozak as discussed in an earlier post is an example of this. When Bozak isn’t playing with Kessel he is often just put out there for a defensive face off where if he wins the face off he heads to the bench and if he loses the face off he would typically face shots against. The result is he has very little opportunity to generate shots for and ample opportunity to be on the ice for shots against. The result is his away from Kessel statistics are outright terrible. There are probably not a lot of players who have this significant of an impact but one must take it into consideration. This is of particular interest for face off specialists such as Bozak (or at least perceived face off specialist).
  • If every teammate has better stats with a player than apart from that player, it is a good sign that the player is having a positive impact on his teams results.
  • Conversely, if every teammate has worse stats with a player than apart from that player it is a good sign that the player is having a negative impact on his team’s results.

The Other Two Tables

I mentioned earlier that there were two more tables on the players WOWY stats pages. These two pages are very similar to the WOWY stats I just described above but are for opponents as opposed to teammates. The first table is for opposition forwards and the second table is for opposition defensemen. These tables are best used to tell you whom a player lines up most frequently against but also show how the players performed when playing against each other. It’s probably less useful than the teammate pages but interesting nonetheless.

TM and Opp Stats

If you browse around stats.hockeyanalysis.com you will see stats with a TM or an Opp in front of them. For example TMGF20 or OppCF20. The TM stats are like a combined WOWY or a summary WOWY stat. TMGF20 is and average of a players teammates ‘apart’ statistics weighted by the time the two players played together. So, for Roman Polak’s TMCF20 for last season it is a weighted average of all of Polak’s teammates apart CF20 (CF20 found in the black box in the WOWY table image above) weighted by the TOI with Polak (TOI found in the blue box in the image above). This is essentially what his teammates do when apart from Polak. Last year Polak’s TMCF% was 0.533 (or 53.3%) which is significantly better than his CF% of 0.492 (or 49.2%). This is consistent with the analysis above that suggested most teammates had a better, sometimes significantly better, CF% apart from Polak than with Polak.

As you might expect, the Opp statistics are similar but are a weighted average of the players opponents statistics. Both the TM and the Opp statistics can be used as a high level overview of the players WOWY and ‘Against You’ statistics and when compared with his own statistics one can quickly assess whether the player is having a positive or negative impact on his teams results.

 In Conclusion…

I hope this provides you with an overview of what WOWYs are, what they are used for, and where to find the data. I understand all these advanced statistics can feel a little overwhelming at first, but if you stick with it they will soon all make sense. They really aren’t that complicated, but they do have a bit of a learning curve. Oh, and definitely don’t be afraid to ask questions.


NEW YORK, NY - MARCH 05: Tyler Bozak #42 of the Toronto Maple Leafs (R) celebrates his goal at 1:51 of overtime along with Phil Kessel #81 (L) to defeat the New York Rangers 3-2 at Madison Square Garden on March 5, 2014 in New York City. (Photo by Bruce Bennett/Getty Images)

In my first post here at MLHS, I wrote an introduction to advanced statistics which generated some good discussion in the comments. A number of the comments were with respect to WOWY analysis and some more specifically with respect to Kessel and Bozak.  I will delve into that more here in my second post.