5v5 Expected Goals Share (xG%)

Overview

This pillar analyzes the flow of play in hockey games by measuring shot quality dominance at even strength. It identifies which team is statistically likely to win based on chance generation rather than current luck or goaltending performance.

What It Does

It calculates the ratio of Expected Goals For (xGF) versus Expected Goals Against (xGA) specifically during 5v5 play. This metric filters out the noise of power plays and penalty kills to reveal which team consistently controls the puck and generates dangerous opportunities in neutral game states.

Why It Matters

In hockey, actual goals are rare and subject to high variance (luck/goalie variance). 5v5 xG% is a more stable predictor of future success than goal differential because it relies on a larger sample size of shot data, effectively predicting regression to the mean for 'lucky' or 'unlucky' teams.

How It Works

The model ingests shot location and type data for both teams, assigning a probability of scoring (xG) to every unblocked shot attempt. It aggregates these values while the game is at 5v5, applies score-adjustments (accounting for leading teams playing defensively), and compares the two teams' rolling averages.

Methodology

Formula: (5v5 xGF / (5v5 xGF + 5v5 xGA)) * 100. Analysis utilizes a 10-game and 25-game rolling weighted average. Data is adjusted for score states (removing 'garbage time' effects) and venue bias. Probabilities are derived from historical conversion rates based on shot coordinates, angle, and type (rebound, rush, set play).

Edge & Advantage

Provides a trading edge by identifying value on 'unlucky' teams (high xG% but low actual win rate due to bad PDO) and fading 'lucky' teams (low xG% but high win rate due to hot goaltending), as these variances tend to normalize over time.

Key Indicators

• 5v5 xG% Share

  high

  The percentage of total expected goals a team captures at even strength.

• High Danger Chance Share (HDCF%)

  high

  The ratio of shots taken from the high-probability slot area.

• Score-Adjusted Fenwick

  medium

  Unblocked shot attempts adjusted for score effects (teams turtle when leading).

Data Sources

• NHL RTSS API

  Real-time shot location and coordinate data.

• MoneyPuck Models

  Public xG probability models for shot types.

Example Questions This Pillar Answers

• → Will the Carolina Hurricanes beat the New York Rangers on the moneyline?
• → Which team will have the most shots on goal in Period 1?
• → Will the Toronto Maple Leafs cover the -1.5 puck line spread?

Tags