LB-Hockey

Playoffs are in full force, and so is the work at LB-Hockey. Over the last three weeks, I’ve been hard at work fixing bugs and implementing new features across the site’s arsenal. Moving forward, starting with this round of changes, I aim to provide a recap for you all of what’s been optimized, fixed, and completely changed throughout the dashboards. This should give you a glimpse into what’s going on behind the scenes here, and more importantly, keep you up to date with what’s being worked on. So without further ado, let’s dive right in.

Season-Dependent Player Headshots & Logos

The Single-Season Player Cards allow you to rewind and see players’ past years. Previously, switching between different seasons for any given player did not affect the displayed headshot, only showing the current one. Now, not only will the headshot reflect the selected season, but the team’s logo at the time (as displayed above by the Coyotes’ logo evolution). This comes with an improvement in the picture quality of all logos used in every tool.

Accurate Former Contract Info

Sticking with the Single-Season Cards, proper contract & age tracking has been implemented for previous individual seasons, rather than simply being estimated. For this dashboard, player ages are calculated relative to September 15th at the start of the selected season. In all other tools, it will remain to be the player’s current age at the time of the most recent data update.

Net-Front Play & Compatibility

Moving over to the Multi-Year Player Cards, we have the biggest change of the batch. Some will remember the Impact on Opposition/Teammates being at the bottom of the Checking & Teamplay categories respectively. Although these two measures of “With or Without You” analysis are still useful info, they are more measures of overall player impact rather than providing insight into a player’s toolkit. Therefore, they have been kept in the skater SPAR calculations but have been replaced on the dashboard by two new metrics.

The first is Net-Front Play, which was already present in the Development Matrix & in badge form here. For forwards, this score is calculated based on their ability to establish position in front of the opposing team’s net and generate chances (through rebounds, tips, etc). As for defensemen, this measures how efficiently they can suppress chances close to their goalie by being an imposing presence through boxing out & tying up sticks. For example, some of the defenders who currently rank at the top of this metric are Radko Gudas & K’Andre Miller. Meanwhile, the 100th percentile forwards for this season are Brady Tkachuk, Zach Hyman, & Chris Kreider.

Compatibility looks to capture a player’s ability to make their teammates’ lives’ easier by succeeding at sporting a style of play that effortlessly meshes with their linemates’ expected strengths & weaknesses. This is measured by first selecting all skaters around the league at the same line level based on 5v5 TOI. All these playing styles are combined to build the profile of the ideal linemate within this range. When it comes to some of the best forwards this season at leveraging teammate strong suits & covering their weak spots, we have Aleksander Barkov & Adam Lowry. Whereas on D, Devon Toews & MacKenzie Weegar lead the pack.

With these changes, the skills progression section of the Progression Tracker has been updated. This change expands the pool of available skills from the 25 seen in the main cards to now additionally include PowerPlay Offence, Penalty Killing, Impact on Teammates, & Impact on Opposition.

Comparison Dashboard SPAR Projections

We’ll move over to the Comparison Atlas where we see another big change, notably at the bottom with its SPAR Chart (Standings Points Above Replacement). Firstly, it has been modified to accommodate up to 5 years, which will be the maximum now moving forward. However, the more important update relates to the projection. This part likely had the least optimized and most difficult-to-follow code section across all my scripts. It has now been cleaned up & overhauled, and so has the projection method. The prediction logic is now more robust by making better sense of players’ progression given their recent year-over-year improvements, decreases, and age. Now, younger players on the up will be rewarded properly while the opposite will hold for older players that are falling off.

Father Time Badge

For the first time in almost a year, a new badge enters the arena. Since there are already a couple of badges dedicated to younger improving players such as Breakout Candidate & Most Improved, it felt necessary to add one that rewards the ageless wonders of the league. This one highlights the 35+ players who perform above a SPAR threshold, with an added multiplier for the skaters in this selected range who outperform their expected age’s performance.

New Badge Selection Method

Since there are only six badge slots in the cards, a decision must be made on which badges are chosen to be displayed for skaters with more than six. Previously, it simply boiled down to sorting by level (platinum, gold, silver, bronze) in order from oldest to the most recently implemented badge. This led to the older badges being displayed more often than others. For example, say a given player has 1 gold, 3 silver, and 5 bronze badges. Using the old method, the 2 bronze badges chosen defaulted to the ones added to the cards the earliest, which resulted in certain badges like End-to-End Rusher or Scoring Threat popping up more often than others. Now, the sorting is done first by badge level, then by ranking within it (gold 1, gold 2, gold 3, etc). As a result, a player’s high-gold badge will be prioritized over a different badge that ranks lower within the gold tier. This also ensures more badge diversity between all players in the dataset.

More Badge Updates

A few more, albeit minor, changes have been made regarding badges. Most notably, some of the more recent badges have been implemented into the goalie cards such as Most Improved, Bang For Your Buck, B E E F, and Father Time. In addition, some badges have had their calculations improved to better represent what they intend to capture. Wrecking Ball is the one which saw the most changes as it is now far more consistent while End-to-End Rusher, Shutdown Stalwart, Coach’s Dream, & Well-Rounded Worker also got slight adjustments.

More Diverse Ideal Linemates

As some of you have pointed out, certain skaters tended to appear rather often in the ideal linemates section of the cards. Guys like Aleksander Barkov & Adam Fox popped up a lot which made sense given that these two especially have very high compatibility scores and sport the Puzzle Piece badge, but it was still occurring a little too much for my liking. Therefore, the category weights have been tuned in the ideal linemates script, and more importantly, a linear penalty (formerly exponential) is now used for the TOI & SPAR difference between any two players. This ensures that higher-level players will not be matched as often with more middle-of-the-pack guys.

Various Tuning

Some additional less-interesting changes have been made that don’t quite deserve an entire section so I’ll rattle them off quickly here:
– A shot-on-goal regression threshold has been added in the single-season SPAR calculation for goalies, which will reward starting goalies and regress smaller-sample ones.
– The Clutch score for skaters has been adjusted to better capture playoff performance.
– Progress has been made towards the Team Cards!
– Some bugs have been eradicated relating to different player name spellings.
– Added some quality-of-life improvements on the coding side of things (better progress tracking, slightly smaller data outputs, JupyterLab update).

Off-Season Plans

The goal for this off-season is to roll out two different team-oriented visualizations. The first one, as previously mentioned, are the Team Cards which will offer team-level data with both single and comparison views. I also hope to finally finish the Roster Builder which has been in the works for well over a year now and will far-and-away be the most interactive of all the dashboards up to now.

I’ll continue to optimize the site’s dashboards and let me know if any of you run into issues throughout the site. Cheers!