LB-Hockey

Steals & Chokes Report

Moving into the second generation of LB-Hockey dashboards, we wanted to shine a light on an area that, while not necessarily neglected in the hockey analytics space, deserves deeper and more diverse coverage: goaltending. We’ve already expanded on goalies’ stylistic aspects in our Multi-Year Cards, the evolution of their skills with our Progression Tracker, and dollar valuations through our Contract Projector. But it still feels as if we, and almost anyone else, hadn’t truly attempted to capture the position’s game-breaking ability.

Introducing the Steals & Chokes Report, our first goaltender-exclusive tool! Despite hosting many lines of data, it remains a fairly simple visualization. But before diving into all the individual elements, we need to explain how we quantify its focus.

Measuring Steals & Chokes

Many times have we heard about how a goalie “stole” this game, or lost his team that game, so how do we go about classifying those? In analytical terms, a stolen game occurs when a goalie’s Goals Saved Above Expected is higher than his team’s win difference. For example, let’s say Dallas beats Buffalo 3-2 with Jake Oettinger posting a 2.1 GSAx. That would mean the Sabres were expected to score roughly 4 goals, which would have gotten them the win. But thanks to Oettinger’s performance, Dallas came out with the W. This would then qualify as a steal for the Stars’ netminder. We can easily mirror this to obtain chokes, which are when a goalie’s GSAx is lower (so deeper into the negatives) than his team’s loss differential.

There are a couple more important things for these calculations, and they stem from forcing our expected goal perspective to be from the goalie’s point of view. This means removing empty-netters from the team’s win/loss differential, since the goalie never would have been pulled in the first place if the team was winning instead of losing, and vice versa. Moreover, the actual expected goal calculations change. We remove shots that missed the net and calculate threat level given that the shot was on goal, thus increasing the xG values.

Totals & Ranks

The total number of steals and chokes is displayed for the selected goalie season, along with a line plot to show their progression through the years. The data goes as far back as 2020. To provide more context on these totals, we’ve attached “bucketed ranks” that work like so:

1. Sort all goalies by starts for that season
2. Group the top 32 as “Starters”, the next 32 as “Backups”, and the remaining few as “Callups”
   • This cut-off will change should the NHL team count change in the future
3. Sort in descending order by steals within these groups to obtain the Steals Rank
4. Sort in ascending order by chokes within these groups to obtain the Chokes Rank
   • Done this way to maintain a “higher rank = better” philosophy

Game Reports

The main attraction of this tool is the game reports section, which lists the ten most recent steals and chokes for the selected goalie season. With these, you can visualize the degree to which goalies impacted any given game in an easily digestible format. The game’s date, team, opponent, and result are shown to give a brief boxscore, supplemented with save percentage and goals saved above expected as the shot-stopping stats. The GSAx boxes are colour-coded to put into perspective how they fare relative to all goaltending performances across the league that season. Lighter colours are closer to the midpoint, while darker shades indicate more extreme outings (one way or the other).