Open any team page on the site and you will find two expected-goals shares that usually agree and sometimes do not. Last season the Detroit Red Wings were 25th in one and 20th in the other. That is not a bug. The two numbers answer different questions, and knowing which is which makes both more useful.
The same model underneath
Both start from the same place: our expected goals (xG) model, which looks at every unblocked shot attempt (where it came from, the angle, the shot type, what happened just before it) and estimates how often a shot like that goes in. Add those probabilities up for a team and you get how many goals its chances were worth, independent of who finished them or who was in net. If you want the basics first, advanced stats explained walks through it.
Where the two numbers differ is in which shots they count and what they correct for.
Five-on-five xG%
This is the classic analytics number. It counts only five-on-five play with both goalies in net, and expresses the team's share of the expected goals in those minutes: 55% means the team generated 55 cents of every dollar of expected offence while it was at even strength.
The one correction we make is for scorekeeping. Every arena's off-ice officials log shot locations a little differently, and a few rinks consistently record shots closer to the net than everyone else does. We measure each rink's tendency from games played there against the same teams' games elsewhere and take half of it out.
Why five on five only? Because it isolates the part of the game where both teams have the same number of players. A great power play or a leaky penalty kill can swing a team's record, but it tells you little about whether the team controls play when the ice is even.
BlueLine xG%
BlueLine xG is our attempt at the whole game. It counts every situation, power play and penalty kill included, and it makes three more adjustments:
- Empty nets. Shots at an empty net are valued with a separate model built for them, rather than left out.
- Score state. Teams that are trailing push for chances and teams that are leading sit back, so a trailing team's shots are a little less telling than they look and a leading team's a little more. Each attempt is weighted for the score when it was taken.
- Home ice. Teams generate slightly more at home, and that is weighted out as well.
It is reported per game rather than per 60 minutes, because it covers all the minutes.
A third number, goalie-adjusted xGA, goes one step further on the defensive side: it scales each team's expected goals against by the regressed track record of the goalie who was in net, from seasons before this one. It is no longer pure expected goals, which is why it sits in its own column.
Does adjusting help?
We checked against the thing xG is meant to predict: goals. Over 2023-24 to 2025-26, with the weights fitted on earlier seasons, the raw all-situations model tracked a team's goals for and against with correlations of .753 and .699. The adjusted BlueLine version raised those to .786 and .747. Adding the goalie adjustment held goals for steady and nudged goals against to .754.
Both shares also track standings well. Across the 32 teams in 2025-26, five-on-five xG% correlated with points percentage at .819 and BlueLine xG% at .835. Plain shot-attempt share (Corsi) came in at .570, which is why we lead with expected goals rather than shot counts.
When they disagree
The gap between the two numbers is usually special teams, score effects, or both. A few clubs from 2025-26:
| Team | 5v5 xG% (rank) | BlueLine xG% (rank) | Points % rank |
|---|---|---|---|
| Detroit | 47.3% (25th) | 49.3% (20th) | 15th |
| Edmonton | 51.0% (12th) | 52.1% (7th) | 14th |
| Utah | 52.9% (6th) | 51.2% (9th) | 15th |
| Los Angeles | 51.6% (7th) | 50.5% (13th) | 20th |
Edmonton is the cleanest example. Its power play converted 30.6% of the time, which five-on-five numbers ignore by design and BlueLine xG picks up. Los Angeles went the other way: a good five-on-five team whose 17.0% power play and 74.6% penalty kill gave a chunk of it back, which is roughly where its points percentage ended up too.
Which one to use
- To judge who controls play, and to compare teams on equal terms, use five-on-five xG%. It is also the version most public analytics sites report, so it is the easiest to compare with numbers you see elsewhere.
- To judge overall team strength, special teams included, use BlueLine xG%. It is closer to how games are actually decided.
- To judge a team's defence with its goaltending, look at goalie-adjusted xGA next to plain xGA. A big gap means the goalie is carrying, or sinking, the team.
You can put any of these side by side in the team rankings table: pick the columns, sort either way, filter to a division or the top ten, and share the result as an image.