1. Home
  2. Learn
  3. Understanding pG
Our Methodology

Understanding pG (Predicted Goals)

Our core metric: a machine learning prediction of how many goals each team will score in an upcoming match.

6 minute read

What is pG?

pG (Predicted Goals) is our machine learning model's forecast of how many goals a team will score in an upcoming match. For every fixture, we generate two numbers:

🏠

Home pG

The predicted number of goals the home team will score.

2.1
Example: Liverpool at home
✈️

Away pG

The predicted number of goals the away team will score.

0.8
Example: Burnley away

These two numbers are our only prediction output. We don't tell you what bet to place or who will win — we provide the data, and you decide how to use it.

pG vs xG: What's the Difference?

People often confuse pG with xG. They're related but fundamentally different:

xG (Expected Goals)

  • Measures what already happened
  • Calculated from actual shots taken
  • Available after a match
  • Answers: "How good were the chances created?"
Example:
"Liverpool had 2.8 xG vs Burnley's 0.5 xG"

pG (Predicted Goals)

  • Predicts what will happen
  • Generated by machine learning model
  • Available before a match
  • Answers: "How many goals will likely be scored?"
Example:
"We predict Liverpool 2.1 pG, Burnley 0.8 pG"

The Relationship

Historical xG data is one of the key inputs our model uses to generate pG predictions. We analyze how much xG teams typically create and concede, then use this (along with 240+ other features) to predict future goals. Think of xG as the historical foundation that makes accurate pG predictions possible.

How We Calculate pG

Our prediction model analyzes 240+ features for each match. Here's a simplified view of the process:

1

Gather Historical Data

For both teams, we collect performance data: recent form, xG created and conceded, home/away records, head-to-head history, league position, and more.

2

Engineer Features

Raw data is transformed into 240+ predictive features. For example: "home team's xG per match over last 5 games" or "away team's clean sheet rate on the road this season."

3

Run Through ML Model

Features are fed into our LightGBM model (trained on 15,000+ historical matches). We use two models: one predicts home goals, one predicts away goals.

4

Output pG Predictions

The model outputs predicted goals for each team. We also calculate a confidence score based on data quality and model certainty.

Key Features We Analyze

Recent form (last 5-10 matches)
xG created and conceded trends
Home vs away performance splits
Head-to-head historical results
Attacking and defensive strength
Goals scored and conceded patterns
League position context
Rest days and fixture congestion

Reading pG Numbers

pG values are continuous numbers that represent expected goal output. Here's how to interpret them:

Example Prediction

Liverpool
2.1
Home pG
vs
Manchester United
1.3
Away pG

What This Tells You

  • We expect Liverpool to create chances equivalent to ~2.1 goals
  • We expect Man United to create chances equivalent to ~1.3 goals
  • Total pG: 3.4 — suggests a relatively high-scoring game
  • Liverpool favored (higher pG), but not overwhelmingly
Low-scoring indicator
< 2.0
Combined pG
Average match
2.0 - 3.0
Combined pG
High-scoring indicator
> 3.0
Combined pG

Understanding Confidence Scores

Every prediction comes with a confidence score (0-100%). This reflects how certain we are about the prediction:

What Affects Confidence

Higher Confidence
  • • Teams with lots of historical data
  • • Clear favorite vs underdog matchup
  • • Consistent recent form
  • • Good head-to-head data available
Lower Confidence
  • • Newly promoted teams (limited data)
  • • Evenly matched opponents
  • • Inconsistent recent form
  • • Limited head-to-head history

Important Note

High confidence doesn't mean the prediction is guaranteed to be accurate. It means the model has strong historical patterns to base the prediction on. Football is inherently unpredictable — even high-confidence predictions will sometimes be wrong.

What pG Doesn't Tell You

pG is a powerful tool, but it has limitations. Our predictions don't account for:

Not Included

  • Player injuries and suspensions
  • Tactical changes or manager decisions
  • Team motivation (relegation battles, etc.)
  • Weather and pitch conditions
  • Random events (early red cards, etc.)

What You Should Do

  • Check team news before matches
  • Consider context pG can't capture
  • Use pG as one input, not the only input
  • Combine with your own analysis
  • Accept that surprises happen

Key Takeaways

  • pG is our core output — predicted goals for home and away teams
  • pG predicts the future, while xG measures the past
  • We analyze 240+ features including form, xG history, and head-to-head data
  • Confidence scores indicate data quality, not guaranteed accuracy
  • pG is a tool — combine it with your own analysis for best results

Continue Learning