The Battle of Alberta via Analytics: Andrej Sekera vs. Dougie Hamilton

The Calgary Flames made a huge splash in acquiring up-and-coming defensive stalwart, Dougie Hamilton, of the Boston Bruins, for 3 draft picks. The deal shocked everyone. Apparently, the Oilers also negotiated for Hamilton, but Boston wanted Darnell Nurse, as well as 3 draft picks (#16 overall, and two second round picks). It seems Edmonton balked at trading Nurse, which to many fans is understandable. Fans, media, and management have much optimism in Nurse’s potential, which was reinforced by his stellar season and deep OHL playoff run with the Sault St. Marie Greyhounds. Somehow, though, Calgary negotiated a deal for less and the rest is history. The Oilers were desperate for a top-pairing defenseman, but they also needed a #1 goaltender. Their major draft-day trade turned out to be exchanging draft picks for Cameron Talbot of the New York Rangers. I like this deal and have high hopes for Talbot. (Based on war-on-ice computations of adjusted and high-danger zone save percentage, I had Talbot ranked quite high on my list of potentially available goalies.)

Still, free agency was coming up on July 1st and some potential top-pairing defensemen were available. On opening day, the Oilers acquired free-agent defenseman, Andrej Sekera. I was excited! In my regular Oilers Facebook group, we had discussed who we wanted for our top pair and Sekera’s name always came up. Indeed, Sekera appeared to be in high demand by other teams. Last season, Sekera was traded from the Carolina Hurricanes to help Los Angeles into the playoffs. Unfortunately, the NHL point system as it is (i.e., overvaluing overtime wins), and with LA’s bad luck in overtime, they didn’t make the playoffs. To this point, Sekera looks like the Oilers’ answer to filling at least one spot for a top-2 defender.

There appears to be wide agreement in the hockey world that Hamilton is going to be a stud defenseman, if he is not already. Sekera has been around for a few more years and is considered a solid defenseman, at least a top-4. Important to note, though, that he and Justin Faulk were the top-pair for the Hurricanes. Most would also agree that Hamilton is the bigger name–and Calary signed him to a fatter contract–, but what would an analysis of their metrics tell us? How much better is Hamilton? Or are they closer in performance than many would assume?

Shot Attempt Metrics: Quality of Competition

Unpacking David Johnson’s WOWY (With-or-Without-you) tables, which, in part, shows how each player performs against specific opponents, I used the opponents’ shot-attempt differentials as quality of competition (QualComp) indices. Opponents with higher shot-attempt differentials (SAT%; also known as Corsi) were considered tougher competition. I assumed that how Sekera and Hamilton performed against weaker competition was not as important. Instead, I focused on tougher competition, specifically, opponents with an average shot-attempt differential of 50% or greater. I divided the tougher competition into 5 levels of average SAT%–56%, 54%, 52%, & 50%–with each level having a range of +/- 1% and a sample size of about 30 opponents. This method allowed me to compare each player’s shot-attempt differential versus similar levels of competition.

In addition, I computed a metric that I call Break-Even QualComp SAT%, or Break-Even SAT%. (I introduced the metric here.) The Break-Even SAT% indicates the level of competition in which a player can hold their own, that is, maintain a 50% shot-attempt differential. The higher the Break-Even SAT%, the tougher the competition the player can handle. I graphed this quality of competition comparison below.

(Click on graph to enlarge. Use your browser’s [back] button to return to the article.)

Regardless of the quality of competition, both Sekera and Hamilton demonstrated impressive shot-attempt differentials. This is clear in their Break-Even SAT differentials with Sekera at 54.8% and Hamilton at 55.9%. Opposition at this SAT% level are typically first-line players on strong possession teams such as Pavel Datsyuk (Detroit), Sidney Crosby (Pittsburgh), and John Tavares (New York Islanders).

In terms of which player outperformed the other, there doesn’t seem to be a clear “winner,” although the Break-Even SAT% suggested that Hamilton was slightly ahead. To gain another perspective, I computed their average SAT differentials versus all opponents with a SAT% greater than 52%. I chose 52% because teams with a SAT% of 52.5% or greater have a 90% chance of making the playoffs. In a loose way, then, the following graph shows how Sekera and Hamilton performed against playoff-caliber teams.

As this graph shows, their respective shot-attempt differentials are nearly identical (52.1 vs 51.8). To be sure I was making a fair comparison, I  computed the average competition SAT%, which was about 54% for each player. By separating offense (SAG/60; shot-attempt generation) and defense (SAS/60; shot-attempt suppression), we have a glimpse into how their playing styles differ. Although their differentials are similar—difference of 4 shot-attempts per 60 min–Sekera is a “higher event” defenseman than Hamilton. Compared to Hamilton, Sekera’s teammates not only generate more shot attempts with him on the ice, they also allow more shot attempts. At this level of analysis, it seems that Sekera and Hamilton are similar in their ability to handle tough competition. However, quality of competition is only one side of shot metrics. The other side is a player’s quality of teammates.

Shot Attempt Metrics: Quality of Teammates

Using the WOWY tables, there’s two measures we can examine on the influence of teammates in relation to a player. First, there is the player’s impact on his teammates: How much does a teammate’s shot-attempt differential change, and which direction (positive or negative) does it change, when paired with the player? Second, there is the player’s dependence on his teammates: How much does a player’s SAT% depend, positively or negatively, on his teammate’s influence? This can be a complicated relationship. For instance, just because Player A’s SAT% improves a ton with Player B doesn’t mean Player A’s SAT% is overly dependent on Player B. That’s because Player A may also be improving Player B’s SAT%. They are better together than apart. Isn’t this the ideal of close relationships we strive for in life? As is true in life, it is true in hockey.

In the tables below, I present both players’ SAT dependency and impact on teammates. I focused on key teammates, namely, the 1st and 2nd line centers and the player’s primary defensive partner. I treated the centers as a proxy for the first and second lines. I further divided the measures into shot-attempt generation (SAG; also known as Corsi-For) and shot-attempt suppression (SAS; a.k.a Corsi-Against). There are many numbers, which can be confusing. The important numbers in bold, which I have labelled as “totals.” In fact, the numbers in bold are averages of the 3 teammates. If you only wanted to know the one number that would give you the most information, “total” (i.e., average) SAT/60 +/- would be it.

We would expect that when playing with the team’s top centers, or their primary D-partner, that a defender’s shot metrics would improve. (All the shot attempt rates are presented in 60-minute units.) On average, Sekera’s shot-attempt (SAT) differential improved by 9.7 shot attempts with the Hurricanes, and by 6.52 with the Kings.

As an side, as a member of the Hurricanes, Sekera’s shot attempt suppression metric remained stable (SAS/60 +/- = -0.68). In contrast, with the Kings his SAT suppression metric improved (SAS/60 +/- = -4.45). The reason? One plausible answer is that these are teams with different systems. In particular, the Kings appear to be more focused on (and perhaps better at executing) defensive systems.

Mainly because of Bergeron’s influence, Hamilton’s SAT differential improved by 13.35 shot attempts, on average.  Does this mean Hamilton is more dependent than Sekera for his higher SAT differentials? Not necessarily. Because the relationship is complicated, we also need to assess the other side: The player’s impact on his teammates.

Sekera had a positive impact on his centers in Carolina and Los Angeles, as well as his D-partner, McNabb, for the Kings. However, his D-partner on the Hurricanes, Faulk, performed worse with Sekera. Faulk’s SAT differential dropped by 10.3 shot attempts (per 60 minutes) when on the ice with Sekera. In the previous table, we can see that Faulk returned the favor by knocking down Sekera’s SAT differential by 8.27 shot attempts. Seems like that relationship was not working out too well; a bit dysfunctional even. Overall, though, Sekera’s impact on his key Carolina teammates’ SAT differential was positive (+2.31), as was his impact on his key Los Angeles teammates (+9.11).

Hamilton also had a positive impact on each of his teammates, both in terms of shot attempt generation (SAG/60 +/- = +9.8) and suppression (SAS/60 +/- = -.6.96). In fact, his impact total (+16.76) exceeded his dependency total (+13.35) by just over 3.4 shot attempts per 60 minutes. Continuing with the relationship analogy, Hamilton gave more to his teammates than he received.

A few analytics writers have recently cautioned about comparing players from different teams using these relative metrics. They argue there are so many other things going on with the teams (e.g., systems, line-combinations) that a straight across comparison would not be meaningful. I tend to agree with this line of thinking. Nonetheless, I think it’s still useful to note the general patterns, specifically, that although each player is dependent on major teammates to boost his SAT differential, each had positive impacts on his teammates’s SAT differentials.

WOWY Diagrams

Another way to express these dependency and impact relationships is through a diagram. Fortunately for Twitter followers of hockey analytics blogger, Micah McCurdy, we’ve been treated to elegant diagrams of teammate shot-attempt WOWY metrics. McCurdy created two types, spider and WOWY diagrams. Spiders are explained here and WOWYs here. To be frank, if you really want to understand how to clearly interpret the diagrams, then I do strongly recommend reading the explanations.

What spider diagrams provide, which I did not in the above tables, is how SAT differentials change with more than 2 players together. To keep things simple, I’ll break down the diagram to its essential points. The main pattern you want to see in spider diagrams is teammate combinations moving toward the bottom-right, below the red line. This bottom-right half indicates a positive shot attempt differential (i.e., SAT% > 50%).

Combinations above the red line are negative shot-attempt differentials (i.e., SAT% < 50%). The combinations in blue are the defender’s average SAT differential and his SAT% with his primary defensive partners. To the right of the diagram, player numbers are listed along with their names and a percentage, which represents the proportional time-on-ice the player shared with each teammate.

Sekera’s Spider WOWY Diagram – 2014/15

(Click to enlarge. Use your browser’s “back” key to return to the article.)

Hamilton’s Spider WOWY Diagram – 2014/15

(Click to enlarge. Use your browser’s “back” key to return to the article.)

Most of Sekera’s and Hamilton’s teammate combinations are below the red-line. Sekera’s combinations with the Staal brothers, Eric & Jordan, are particularly dominating. In fact, their shot-attempt differential was very good (approximately 61% SAT%) when both brothers were on the ice. Interesting to note that that the Faulk-Sekera pairing actually does quite well when playing with either Staal brother. There is a small cluster of combinations that move diagonally upward to the right (still below the red line). This would show little, if any, change in shot-attempt differential, but an increase in both shot attempt generation and suppression (i.e., more events). This fits with what we found earlier in my analysis, that Sekera appears to be a higher event defensemen compared to Hamilton.

For Hamilton, the general trend of player combinations is downward (more shot suppression) and toward the right (more shot generation). Any combination with Bergeron was exceptionally strong. By my estimation, the 5-man combination of Smith, Bergeron, Marchand, Chara, and Hamilton had an impressive shot-attempt differential of 65%.

Quality Shot Analysis

So far, I’ve looked at the most blunt measure of shot metrics, shot attempts, which include blocked shots, missed shots, and shots on goal. Shot attempt metrics are useful and informative, but limit any analysis to broad brushstrokes. This past season, the bloggers at war-on-ice stepped up their measurement game and created metrics to account for higher quality shots, specifically, scoring chances and high scoring chances, which are mainly scoring chances from the slot area.

At this level of analysis, a difference between the two players is emerging. When Hamilton was on the ice, the Bruins generated significantly more scoring chances (54.7%), especially high-scoring chances (57.3%), than their opposition. Moreover, Hamilton ranked 7th in the league among defenders in terms of high-scoring chance differential and 21st in general scoring chance differential. Finally, Hamilton also had strong scoring chance differentials relative to when he is on the ice to off the ice.

Sekera’s high-scoring chance differential was in the average range (50%), relative to the league. When he was on the ice, he and his teammates allowed as many high scoring chances as they generated. His scoring chance differential was a better (51.4%), ranking him 71st in the league among defenders.

High & Medium Probability Zones

Above, I observed differences between the players in shot attempts generated and suppressed. Although their respective SAT differentials were similar, Sekera was a higher event player than Hamilton. In particular, it appears Hamilton and his teammates were better at suppressing shot attempts, whereas Sekera and his teammates were better at generating shot attempts. This loosely suggests that Hamilton may be stronger defensively. To get a clearer picture of these offensive and defensive differences, I once again used data from war-on-ice using the “Player-Hextally” option. Here war-on-ice breaks down shot locations to 3 major zones: high (the slot), medium, and low probability zones. I focused on the high and medium-probability zones because these are the areas where the action happens. For instance, shooting percentages are almost 5 times higher in the slot than from the low-probability zone.

In the following table, I present a team’s shot rates relative to the league, for and against, when the player is on the ice,. As well, I include these shot rates relative to the league and to the team. The shot rates are then divided into 2 zones: the slot area (high-probability) and the medium-probability zone (i.e., immediate area around the slot). Green boxes indicate strong numbers, whereas red boxes indicate weak numbers.

Consistent with Hamilton’s high scoring chance differentials above, he and his teammates generated 28% more shots from the slot compared to the league average. Moreover, with Hamilton on the ice, the team produced 25% more shots from the slot compared to him off the ice. From the slot, Sekera and his teammates were less productive than Hamilton and company, but still 5% above league average. Shot rates generated in the medium-probability zone were more similar, with Sekera having a slight advantage. Hamilton and his teammates had a shot-rate of 15% above league average, whereas Sekera and his teammates were at 17% above league average. Both players, then, appear very good in helping generate shots from the slot and medium-probability zones.

What about their respective defensive metrics? One difference is very clear: Hamilton and his teammates were far superior in suppressing shots from the slot compared to Sekera and his teammates. In fact, Sekera and company were below average, that is, they allowed 14% more shots from the slot relative to the league average.  Indeed, with Sekera on the ice, his team allowed 12.3% more shots compared to him off the ice. Although this doesn’t look good for Sekera’s defense, we also see that he and his teammates are solid at suppressing shots within the medium-probability zone–6% below league average–, whereas Hamilton and his teammates are no different than league average.

Net Goals

Now that we’re seeing differences between them more clearly, how does this translate into expected goals? Again, war-on-ice provides this information by dividing expected goals into scoring zones and then applying conversion rates to the shots to compute Net Goals per 60 minutes. (The Average Conversion figures suggest that these tables use shot attempts, not shots on goal.) I then computed Total Net Goals using the player’s time-on-ice.

The simplicity of Net goals is that it condenses many of the shot metrics into one number. When Sekera was on the ice, his team was expected to score 4 goals more than their opposition. Hamilton and his teammates were expected to score 11 more goals than their opposition. Hamilton’s advantage over Sekera is two-fold: He not only seems to help generate more goals from the slot (3.09/60 vs. 2.01/60 for Sekera), but also appears to help suppress more shots from the slot (-0.85 vs. +2.49 for Sekera). So despite their similarities at the level of shot attempt metrics, including the ability to handle the league’s toughest competition, when I went deeper into the data to account for shot quality, Hamilton appeared to be the strong player, both offensively and defensively.

New Passing Metrics

To this point, I’ve indirectly shown how each player contributes to his team’s scoring. All the above metrics are group measures: They indicate how the team is doing while the player is on the ice. But what about directly measured contributions? Fortunately, Ryan Stimson and his small army of volunteers have painstakingly collected such data! It’s called the Passing Project. Recently, one of his colleagues, Spencer Mann (Twitter: @SpenceIce), created graphs of the project’s main metrics. The graphs provide a clear snap-shot of a player’s strengths and limitations in terms of their offensive contributions. The percentile scores tell us how a player ranks compared to other defensemen.  For example, a score in the 80th percentile means the defenseman is better than 80% of other defenders on this metric.

To explain the metrics, I’ll quote Stimson (bold and italics added):

“CC% [SAC%] and CC/60 [SAC/60] are for Corsi Contribution [Shot Attempt Contribution] (individual shot attempts, primary passes leading to shot attempts, and secondary passes leading to shot attempts) percentage and per sixty minutes. These tell you how much offense goes through that player while on the ice and also how often they contribute.

Composite SAG and SG represent the total number of shot attempts and shots a player generated from both primary and secondary passes per sixty minutes. SAG/60 is solely for the player’s primary passing contributions.

Entry Assists represent the number of controlled entries a player assisted on. This is determined by the number of passes in transition (prior to entering the offensive zone) we recorded for each player.

SC Contribution% and SCC/60 are the exact same thing as CC% and CC/60, but represent only the scoring chances a player was involved in. I combined our passing data for scoring chances with War-on-Ice’s scoring chance data to arrive at the total number of scoring chances a player contributed to. SC SAG/60 represents the number of scoring chances set up from a player’s primary passes.

Source: Ryan Stimson, In Lou We Trust (SB Nation), Twitter: @RK_Stimp

Source: Ryan Stimson, In Lou We Trust (SB Nation), Twitter: @RK_Stimp

Based on these graphs, the main differences between Sekera and Hamilton are their respective strengths in zone entry assists and generating scoring chances. Hamilton was especially strong in his contribution to scoring chance ranking almost at 90th percentile compared to other defensemen. This is consistent with the indirect measures I presented earlier (i.e., Scoring Chance% & High-Scoring Chance%).  Sekera was not as strong as Hamilton, but still above-average ranking at about 60th percentile. I was disappointed to see that Sekera appeared to be weak (about 25th percentile) in setting up scoring chances from primary passes.

But when it came to assists that lead to controlled offensive zone entries, Sekera excelled with a ranking over 95th percentile. In contrast, Hamilton was surprisingly quite below average (30th percentile).  Reading from different sources, and from the few Los Angeles games I watched, Sekera’s passing stood out. His overall passing metrics easily support that observation. I’m excited about seeing Sekera’s passes lighten the load for the Oilers’ puck-lugging forwards, like Taylor Hall.

Summary

Hamilton and Sekera have both handled the league’s toughest competition and held their own. Indeed, their shot-attempt differentials are almost identical against this competition. A main difference that began to emerge was that Sekera was a higher event defender than Hamilton. That is, more shot attempts happened both ways when Sekera was on the ice. When examining each player’s impact and dependency on teammates, they both depended on their top 2 centers to bolster their SAT differentials, but at they also reciprocated by positively impacts on their centers’ SAT differentials. In addition, Hamilton had the benefit of playing with Chara, although Chara’s SAT% also improved with Hamilton. Sekera’s defensive pairing in Carolina, Faulk, was not a particularly good pairing, although they did well when on the ice with either of the Staal brothers. Despite limited time, Sekera appeared to work well with his defensive partner in Los Angeles, McNabb, as well as his centers, Kopitar and Carter.

Looking deeper into the data, in particular, by comparing scoring chances, as well as quality shots based on location, Hamilton came out on top both offensively and defensively. In terms offense, he was much stronger than Sekera in generating scoring chances, whether through primary passes leading to shots or taking shots himself. Hamilton also appeared to be better than Sekera in suppressing shots from the slot. Sekera’s main advantage over Hamilton, and most other defensemen in the league, was his ability to execute passes that lead to controlled zone entries. Sekera appeared to be especially strong in contributing to shot attempts and shots on goal, but the shots tended to be of poorer quality than those of Hamilton. This data suggests that Sekera is not a strong playmaker, yet it also suggests  that his passing will help the Oilers in their transition to offense, which is an area of weakness for them with their current blueline. (This link takes you to Sunil Agnihotri’s post on what the Passing Project data reveals about the Oilers’ defense.)

When Sekera was asked why he chose the Oilers, he hit on all the right notes. “I looked at the roster and I saw the team they had, the coach and the management,” Sekera said. “When I saw what kind of players they had there, it made my choice very easy. They have a lot of skill, a lot of speed and a lot of smart players. They have a good coach, a good GM and a good goalie. It was a good place for me to play with my style of hockey, so that’s why I chose Edmonton” (Source: NHL.com). Skill, speed, and intelligence is what Sekera wanted to match his style (and a 6th year on his contract didn’t hurt either).

I am excited about the upcoming season. McDavid alone may be worth the price of admission some nights. But to see a more complete team, one with a very competent defenseman like Sekera at the helm, gives me even more hope about the Oilers future. The time is coming soon for the rebuild to have a playoff-worthy structure and I see Sekera as a key piece.

Thanks for reading. There was a lot of information packed into this post, but I hope I made it understandable. Some of this information I’m presenting for the first time, so it’s entirely possible I’ve made errors or been unclear. If you have any comments or questions, I’d like to hear from you.

Walter Foddis
Twitter: @waltlaw69