#xGextraliga: Lessons from disassembling xG by types of attack

We know that xG (expected goals) metric is a very precise represenatation of team strengths on the ice. It is time to look at components of the metric. We can approach this in various ways – by location, by shot type, by different score, by dangerous possessions or by types of attack. The last option is enabled by tracking 2020-21 xGextraliga data and I select that for this article.

These are components of total xG tracked (75 games, 432 xG, 437 goals):

  • 5v5 quick attacks = also called „rush attacks“ where the attack is counted just a few seconds after entering the offensive zone, odd man rushes are included here
  • 5v5 slow attacks = team enters the offensive zone, holds onto the puck a attacks from the pressure
  • 5v5 OZ steal attacks = team wins/steals the puck in the offensive zone and attacks from that possession
  • 5v5 OZ face-off attacks = all possessions that started with the offensive zone win, if a team leaves the zone a new possession is counted
  • PP attacks = all power-play offence
  • SH attacks = all shorthanded situations offence
  • EN situation attacks = all offence created with a goalie pulled by one team
  • 3v3 situation attacks = usually an overtime hockey

First a visualization of how many goals were expected to be scored from described types of attack:

As we might already know more than a quarter of all goals are scored on a power-play (25.5% of xG and 28% of all actual goals in my sample). The PP vs SH xG ratio is 7.5:1 which alligns nicely with a special team research from the NHL (8:1 in Matt cane´s piece). At 5v5 quick attacks are the main sources of offence and have twice as much volume as slow attacks and OZ steal attacks. 5v5 OZ face-off attacks have a lower volume overall but still contribute to almost 6% of all expected goals in hockey.

I want to go a step further and similarly to finding out which metrics are significant in hockey I checked selected metrics that are more significant overall for its correlations (r) to winning (goal differentials). Displayed below are game by game correlations with an increasing sample size.

First thing to notice is how wildly dependencies on „winning“ change over the sample. At 40 games mark we have a very different results that we obtained at the end of tracking. Early games were often in favor of forechecking teams (good control of 5v5 OZ steals) and 5v5 slow attack offence gained its respect in later games. Two metrics that have oscillated with solid r values uncluding its final form are PP attacks and 5v5 quick attacks. These are two types of attack with a relatively high volume (first graph). This says: „Control the important type of attack and you should do alright in many games.“ 5v5 face-off attacks do not correlate strongly with winning but it is to be expected for their relatively low share on total goals scored in hockey.

Game by game correlations are nice but which types of attack work long-term? I compared my metrics in 10-game sample by all 14 teams in Extraliga and their 10-game goal differentials. The correlations are as follows:

+0.81 for PP attacks
+0.53 for 5v5 slow attacks
+0.46 for all 5v5 quick attacks, OZ steal attacks and face-off attacks

This is interesting. To me the main takeaway for 5v5 values is this:

There is no clear proof that teams playing a specific style have an edge over their opponents. You can focus on rushes, forechecking or pressuring from cycle. At the end it comes down to a quality of your players, your team and your gameplan.

The strongest correlation of +0.81 is assigned to PP attacks which only supports that claim in my eyes. More quality teams and players are able to show their strength on a power-play and be difference makers in a game by game scenario (+0.49) or long-term (+0.81). Lastly, it is interesting to see that importance of slow attacks (and face-off attacks to some extent) raises in a long run (from +0.24 on game by game basis to +0.53 on a 10-game sample)


These are quite few information being thrown at you. Here is to highlight main takeaways:

  • More than a quarter of goals are scored on a power-play, cca 65% at 5v5 with 31% of them being scored from quick (rush) attacks.
  • There is no unique style that would bring you closer to success, your team needs to be strong in various areas of the game.
  • Still a strong power-play is an essential weapon for having a success, also importance of controlling slow attacks battles seems to gain its significance in a long run.


Zanechat odpověď

Vyplňte detaily níže nebo klikněte na ikonu pro přihlášení:

Logo WordPress.com

Komentujete pomocí vašeho WordPress.com účtu. Odhlásit /  Změnit )

Facebook photo

Komentujete pomocí vašeho Facebook účtu. Odhlásit /  Změnit )

Připojování k %s