Month: July 2023

Changes will be needed for the Oilers penalty kill next season

/ sunilagni / 1 Comment

With powerplays across the league improving every year with an influx of offensive talent, it’s become more and more imperative for teams to figure out how to limit shots and scoring chances against. Goalies of course play an important role, but their own ability to stop shots is an area that coaches have little influence on. What coaches can do, and what they have more control of, is how well and how often their skaters can prevent shots and chances in the first place. This can be managed by their coaching tactics and player deployment strategy. And it’s something the Oilers have struggled with under the current coaching staff.

Since February 10, 2022 when Woodcroft became head coach, the Oilers penalty kill has allowed the 7th highest rate of shots against in the league and the 10th highest rate of scoring chances (using Fenwick, or unblocked shot attempts as a proxy). Thanks to some league-average goaltending, the Oilers penalty kill results haven’t been a complete disaster, with the team allowing 7.75 goals against per hour, which has ranked 17th in the league.

One of the key issues is the personnel Woodcroft has deployed as his first penalty kill unit, with the group posting one of the worst rates of shots against in the league when compared to the first units of other teams. On average, first unit penalty kills allow 57.78 shots against per hour, which is around 5 shots more than their teams second unit (about a 7% increase) as they’re more often up against the opposing team’s top powerplay units and their star players. The Oilers top penalty kill under Woodcroft has allowed a rate of 66.05 shots against per hour, which has been the third highest in the league. This rate of shots against per hour is also around 11 shots more than what the Oilers second unit has posted, an increase of around 17%, and double than the league average. (Please note that the forward with the most penalty kill minutes from each NHL team was used as a proxy to assess each team’s first unit. On average, the forward with the most penalty kill ice time played 38.8% of their team’s total penalty killing time.)

Woodcroft’s go-to defence pair on the penalty kill since becoming the Oilers head coach has been Darnell Nurse and Cody Ceci, with Ryan Nugent-Hopkins, along with a rotating partner, being the most common forward up front. One of these three players has been on the ice for 71.2% of the team’s total ice time on the penalty kill since Woodcroft took over (665 total minutes). As mentioned before, the Oilers performance and ability to suppress shots has been well below league-average levels among first unit penalty kills, so something will definitely have to change for next season. Holland did mention the penalty kill unit as an area of concern in his recent media availability, and they have signed forward Connor Brown who has experience killing penalties and saw good performances from Mattias Ekhom after the trade deadline. But more will likely need to be done if the Oilers want to ensure improvements next season.

One thing that the coaching staff needs to strongly consider is finding a new defence partner for Nurse, who has posted better on-ice numbers with defencemen other than Ceci.

Under the previous coaching staff prior to Woodcroft’s arrival, Nurse’s three most common penalty kill partners were Ethan Bear, Evan Bouchard and Adam Larsson. And his best performance numbers (again looking at shots against) was with the one defender who’s still with the Oilers, as he and Bouchard limited the rate of shots against to 51.34 per hour – just better than league average (~54 shots against per hour). It’s unfortunate that Nurse is the one that needs help and doesn’t always drive positive outcomes, but at this point you’re just trying to get as much value as you can from the player. The other option is just having Bouchard and Ekholm take more of the penalty kill minutes, as those two showed good chemistry together as regular partners at the end of last season, and this way it doesn’t disrupt the defence pairings at even-strength.

The other option to consider is removing Ceci altogether from the penalty kill. For pretty much his entire career, Ceci’s on-ice rate of shots against on the penalty kill has been higher than his team’s rate. Put another way – over his career, Ceci’s team’s consistently did a better job at suppressing shots against when he wasn’t on the ice. The only time Ceci’s numbers were closer to his team’s levels was in Pittsburgh, the year before the Oilers signed him – and this appears to have been an outlier season. And unfortunately for him and the Oilers, his on-ice shot suppression number continue to be poor on the penalty kill and in-line with his career levels.

If the Oilers could find a suitable alternative for Ceci, someone like Ekholm who posted solid shot-suppression numbers for the Oilers after the trade deadline, we could also see Nugent-Hopkins numbers bounce back. Under the previous coaching group, Nugent-Hopkins had a positive impact on the team’s ability to suppress shots against on the penalty kill, with the team doing better with him than without him (a -7.27 shots against relative to team). This previous season he posted the worst numbers in his career and in the league (a +16.53 shots against rel), likely because he had to play so many minutes with the Nurse and Ceci tandem. In 105 minutes with them, RNH’s on-ice rate of shot attempts against was over 71 per hour. In 66 minutes away from Nurse and Ceci (since Woodcroft too over), RNH saw his on-ice rate of shot attempts against drop down to 52.38, which would be closer to league average levels.

It should also help to have Connor Brown as an option on the penalty kill, but as I mentioned in a recent piece, his numbers were typically stronger when he wasn’t the first forward option. When Brown was traded to Ottawa from Toronto and became the top penalty killing forward, his numbers took a big hit with the Senators often struggling to prevent shots with him on the ice. So it’s probably best to try him out on a second unit first, monitor his progress and then gradually feed him the heavier minutes, potentially alongside Nugent-Hopkins or eventually his replacement.

The other option is to give forward Ryan McLeod more minutes on the penalty kill, as the team’s rate of shots against dropped by around 15% when he was on the ice since Woodcroft was hired. He does well against top competition at even-strength, so it might be worth giving him more responsibility shorthanded.

While the onus is on the coaching staff to implement the right tactics and player deployment strategy, it’s also on the general manager to monitor and identify the issues, and bring in the right coaching and player personnel. Prior to Woodcroft, the Oilers also struggled to suppress shots against on the penalty kill under Tippett, as the team allowed the 11th highest rate of shots, but was bailed out by some excellent goaltending that ranked second best in the league. Because of that, the Oilers were able to keep their rate of goals seventh lowest in the league. An inability to suppress shots and chances is something Tippett struggled with during pretty much his entire time in Arizona, so it shouldn’t have been a surprise to Oilers management.

Holland and his professional scouts have tried more than a few times to bring in some help for the penalty kill, but they appear to be more focused on previous results (i.e., goals against), which can be driven by the team’s goaltending, than the actual process behind the results (i.e., helping bring down the shots against). Acquiring Ekholm is a step in the right direction, as he has the skill and experience, but that was done in year four of Holland’s tenure as general manager. At this point, it might also be worth considering adding a penalty kill expert to the coaching staff who has a track record of suppressing shots and chances against – and not just goals. Better late than never.

Whatever the Oilers do going forward, it’s critical that a deeper analysis of existing flaws are done on a more regular basis. Improving their penalty kill is going to help increase the Oiler’s odds of winning games, and needs to be closely monitored.

Data: Natural Stat Trick

Connor Brown as an option for the Oilers penalty kill

/ sunilagni / 2 Comments

I find whenever Ken Holland signs or acquires an experienced player, and mentions his penalty killing abilities, it’s definitely worth investigating further. There’s been more than a few instances where the Oilers general manager and professional scouts have misread a players abilities, still sign them, only to see them struggle and eventually depart the club. The two examples that instantly come to mind are Kyle Turris and Markus Granlund. Both were expected to help with the team’s depth and be effective penalty killers when their previous results clearly indicated otherwise.

Connor Brown is the latest player the Oilers have signed who has had experience on the penalty kill, which will be needed considering their shorthanded performance and results under Woodcroft haven’t been great. Last season, the Oilers penalty kill allowed the 12th highest rate of goals against in the league (8.52), largely driven by their inability to suppress offence as reflected by their rate of shots against per hour (58.31) – which was ninth highest in the league. Holland mentioned in his most recent press conference that the penalty kill is something they’d like to improve, and that Connor Brown, among others, are expected to help with the team’s overall defensive play.

He’ll be ready to go for Training Camp and ready to go for the season. I just think he’s a good hockey player. He’s got hockey sense, he can play 200 feet, he can kill penalties, and he’s got a couple of 20-goal seasons. So, I know he’s excited and I think he’s going to be a great fit for our team.

Edmonton Oilers

Below is a summary of Brown’s penalty-kill experience and on-ice performance numbers relative to his team from the last seven seasons. I removed his first season in Toronto where he only played seven NHL games, spending most of the 2015/16 season with the Marlies in the AHL as a 21 year old.

In his first three NHL seasons with Toronto (ages 22-24), Brown was mostly on the second penalty kill unit playing the third most shorthanded minutes among the Leafs forwards, and the fourth highest rate of minutes per game. The Leafs penalty kill had pretty good results in that three year period, allowing the ninth lowest rate of goals against, largely driven by their goaltending which posted a save percentage of 87.77%, sixth best in the league. The skaters really didn’t do that great of a job preventing unblocked shots attempts (i.e., Fenwick, a proxy for scoring chances) and shots against, as the team ranked at or below league average in this time period.

But with Brown on the ice, the Leafs did see a slight drop in the chances against, as reflected by his relative-to-team numbers in the table above. For example, with Brown on the ice between 2016-2019, the Leafs allowed a rate of 50.56 shots against per hour on the penalty kill. Without him, that rate increased by about 11% to 56.12 shots against per hour. Again, Brown wasn’t consistently on the top penalty kill unit playing against top powerplays during this time period. But he performed well in a secondary role, and clearly had the coaching staff’s trust in key situations.

His next three seasons in Ottawa (ages 25-27) were a little different. He still played regularly on the penalty kill, now getting top unit minutes against the best powerplays in the league. But it appears that this increased workload, and perhaps playing in a new system on a weaker team hurt his overall performance numbers.

Between 2019-2022, the Senators penalty kill allowed the ninth highest rate of goals against in the league (7.71), largely because the team struggled to suppress shots and chances against. As a group, they allowed the seventh highest rate of unblocked shot attempts (79.06) and the fifth highest rate of shots (58.09). Brown appears to have been part of the problem, as the team struggled to suppress offence against especially with him on the ice. Without Brown, the Senators allowed a rate of 56.70 shots against per hour. But that number increased by 7.5% to 61.32 shots against per hour when Brown was deployed.

Playing on the top penalty kill unit like Brown was, it’s expected an individual’s on-ice rate of shots against will increase as you’re up against the top units and top players in the league. On average, top penalty kill units see an increase of five shots per hour compared to their team’s second penalty killing units. But posting anything above this level, which Brown did in his first two seasons in Ottawa as a number one penalty killing forward (i.e., +5.89 in 2019/20 and +11.39 in 2020/21), is pretty alarming. Brown did post better numbers in his final season in Ottawa, but this could be an outlier just based on what we know about player aging curves and Brown’s injury history.

Knowing Brown has experience killing penalties, and how passionate he is about the role, and his desire to leverage this season in Edmonton to get a much bigger contract next summer, I’m sure he’ll get plenty of opportunities to secure shorthanded ice time. It’ll just be important for the Oilers coaching staff and management to temper their expectations, have Brown in a secondary role on the penalty kill and closely monitor his on-ice performance before considering increasing his workload. It’d also be in the Oilers best interest to continue looking for potential solutions on the penalty kill, whether it be adjustments to tactics or player deployment. Powerplays are getting better across the league, and it’s imperative the Oilers keep pace with the increased offence.

Data: Natural Stat Trick

Using generative AI like ChatGPT to build an NHL roster

/ sunilagni / Leave a comment

It’s going be interesting to see how artificial intelligence continues to grow within professional hockey leagues like the NHL. The different barriers like computer hardware, internet connections and data warehousing and analysis have been minimized, really paving the way for more and more AI integration into all facets of life.

ChatGPT has been a neat tool to get a sense of where AI is at and how generative AI can be used to create new artifacts and information based on the data that’s available. For my own interest, I was curious to see how AI would construct an NHL roster. This is information most people would know and it’s been published before. But I wanted to see what ChatGPT could come up with. And what the starting position would be if AI was actually in charge of developing an NHL roster, or at least assisting with the work.

Below is the response I got when I asked ChatGPT how it would construct an NHL roster:

Constructing an NHL roster involves several key considerations, including player positions, skill sets, team chemistry, salary cap constraints, and overall team strategy. While the specifics may vary depending on individual team preferences and circumstances, here are some general steps to consider when constructing an NHL roster:

  1. Assess team needs and goals: Identify the areas where your team needs improvement, whether it’s goaltending, defense, or offense. Consider the team’s short-term and long-term goals, such as contending for the playoffs or rebuilding for the future.
  2. Establish team strategy and style of play: Determine the style of play you want your team to adopt. Are you aiming for a fast-paced, offensive-minded team, or a defensively focused, physical squad? This strategy will influence the type of players you target.
  3. Allocate salary cap space: Understand the salary cap limitations and budget available to construct the roster. Allocate funds strategically to secure core players, address critical needs, and maintain financial flexibility.
  4. Build a strong core: Identify and acquire or develop core players who will anchor your roster. This typically includes a mix of top-line forwards, a solid defensive unit, and a reliable starting goaltender. These players should be talented, dependable, and capable of leading both on and off the ice.
  5. Fill out supporting roles: Surround your core players with supporting cast members who complement their strengths. Seek skilled depth players who can contribute offensively or defensively, depending on team needs. Consider players who excel in special teams (power play, penalty kill) to bolster those areas.
  6. Balance player skill sets: Strive for a well-rounded roster by balancing offensive firepower, defensive responsibility, physicality, and speed. Ensure you have players who can generate offense, shut down opponents, provide leadership, and contribute in various game situations.
  7. Develop prospects: Invest in player development to groom young prospects and build a pipeline of talent for the future. Evaluate when to promote prospects to the NHL level and provide opportunities for their growth while considering their long-term development.
  8. Consider team chemistry and culture: Assemble a roster with players who can work well together, share a common goal, and fit into the team’s culture. Building a cohesive and supportive team environment can contribute to success on and off the ice.
  9. Continually assess and adapt: Roster construction is an ongoing process. Regularly evaluate the team’s performance, identify areas for improvement, and make necessary adjustments through trades, free agency, or player development to optimize the roster.

Remember, roster construction is a complex task that requires careful evaluation, analysis, and decision-making. It often involves a blend of experienced management, scouting expertise, and understanding the team’s unique circumstances and objectives.

Not bad! Pretty concise summary that would do okay in a simple business proposal document. It has all of the core concepts of building out an NHL roster, and we know some NHL clubs are better at this than others. I think from here, you’d be able to dig into each of the nine concepts and eventually have AI carry out the work to accomplish each one. A lot of work, with investment in people, processes and technology needed. But if you’re chasing championships, why wouldn’t you consider using every tool available including generative AI?

The challenge for any team of course would be first feeding the AI system the related data and information and knowledge to build solutions from. For example, to assess the team’s needs, you’d have to give it as much player and game data, along with historical information for it to know what the right metrics are, what’s an acceptable range for each metric, and how successful team’s have been built in the past. That itself is a massive lift, and would need constant refining as more and more data, such as player tracking and new player attributes, become available. The key for any team starting out with AI is identifying what your actual goals are and then prioritizing specific areas you would want to apply AI to.

The reality is that as AI technology gradually improves and works out its existing flaws, it is going to play a massive role in professional sports as teams will look for every competitive edge possible to build championship-caliber rosters. It’ll be interesting to see how owners and their executives embrace the technology available and integrate it into their overall business operations, especially in the NHL where some clubs still don’t the infrastructure to receive, store and utilize player tracking data. Those teams are likely going to struggle and we could see a noticeable performance gap develop between the top end teams and the rest of the pack.

Managing declining assets

/ sunilagni / 7 Comments

One of the issues that’s come up for the Edmonton Oilers this off-season is their lack of cap space and inability to make significant improvements to the roster. Misreading the market, poor decision-making and plenty of overpays will do that. And it’s a difficult cycle to break unless you have some creativity and courage in your front office.

Now plenty of forwards and defencemen and a goalie are on heavy, long-term deals. But that shouldn’t be a reason for the club to play things conservatively this off-season. There’s plenty of room for improvement, and management needs to find ways to add talent and depth, and gain any competitive advantages – regardless of how small the margins or gains might be.

One concern I have heading into next season is the top six, which no doubt has some excellent pieces. The issue is that it’s getting older, with Zach Hyman, Ryan Nugent-Hopkins and Evander Kane all over 30 years old now and struggling to produce in the recent playoffs. And while all three bring specific skillsets to the team, their production is more likely to drop-off as they age. When that happens is anyone’s guess, and is based on a number of factors. But for one player in particular, there are indicators that it could happen sooner rather than later.

Looking at how Evander Kane has performed relative to his team’s overall performance at even-strength, we see a declining trend with his team’s gradually doing better without him than with him. If you’re getting paid and deployed like a top six forward, and spending significant time with two of the top players in the world, the team should be seeing a bump in productivity with you on the ice. But that hasn’t been the case for Kane. Last regular season, the Edmonton Oilers posted a Corsi For percentage of 50.93% at even-strength with Kane on the ice, a drop from the 52.55% Corsi For percentage without him on the ice. The team’s shared of expected goals, which factors in shot quality, was an even bigger drop off, as indicated by the lines in blue. Without Kane on the ice, the Oilers posted a 54.70% share of the expected goals (Source: Natural Stat Trick). With him on the ice, that share dropped to 48.08%, with the Oilers getting out-chanced more often. Kane’s overall decline appears to have started in his first full season with San Jose (2018/19) when he was 27 (right after when offensive players tend to drop off), and it’s hard to imagine things getting better considering he’ll be 32 years old before the next season begins.

Whether it’s his performance, or the injuries he’s sustained, the Oilers coaching staff has made adjustments to how Kane has been deployed. Just over 25% of his total ice time at even-strength last season was against elite-level competition, which is the second lowest proportion of ice time against this group in his career. For context, in his first two full seasons in San Jose, Kane spent around 35% of his ice time against elite-level competition (Source: Puck IQ).

Kane is definitely a player to watch this upcoming season. The hope is he can bounce back and be a solid supplementary player in the top six. But it’s becoming more and more likely that he’ll lose another step, possibly paving the way for one of the younger prospects in the system – someone like Dylan Holloway, or maybe even Xavier Bourgault, to make the jump and play more of a feature role in the top six. It’ll be critical for the coaching staff to closely monitor the performance levels of Kane and the top six, and make swift adjustments when necessary.

This issue also has to be on the general manager’s radar. Kane’s value remains high across the league because of his experience and past results. This might be a good time for Ken Holland to find some much-needed creativity and courage and start creating a trade market for the player, maximizing a potential return. The team’s cap situation is a mess right now thanks to him, with some good young players like Bouchard and McLeod needing long-term contracts, and more youngsters on the way. Clearing cap space is going to be important for long-term sustainability, something Ken Holland has struggled with pretty much his whole career. So it’ll be interesting to see how he navigates this one.

Data: Natural Stat Trick, Puck IQ