xG and Expected Goals - Is It Already Priced Into Markets or Is There Still Edge?

[Klaus]

Specific question with a specific timeline that matters.

I started incorporating xG data into the Bundesliga model in 2016.

At that point xG was available to serious analysts but hadn't entered mainstream football discourse. Match of the Day wasn't displaying it. Casual fans had no idea what it was. Operators were incorporating it inconsistently.

The edge I found from xG-based analysis in 2016-2018: meaningful. Teams systematically over or underperforming their expected goals created predictable regression patterns the market hadn't fully priced.

2019 onwards: xG enters mainstream broadcast. Opta and StatsBomb data becomes standard operator infrastructure. The Bundesliga model's xG component produces diminishing returns.

2021-2024: the basic xG regression trade is largely closed in top flight markets.

My question for the forum: does genuine xG edge still exist anywhere, or has the metric been fully absorbed into market pricing.

 

[SharpEddie47]

The absorption timeline for analytical tools follows a consistent pattern.

Academic discovery. Small community of serious analysts using it. Edge exists. Tool becomes commercially available. Operators adopt it. Edge compresses. Tool enters mainstream discourse. Edge largely gone in liquid markets.

xG has followed this pattern precisely.

The NFL equivalent was DVOA and similar efficiency metrics from Football Outsiders. Generated genuine edge when the analytics community was small. By 2015 the major US books had hired the analysts themselves.

The question isn't whether xG has been priced in at the top level. It has.

The question is how far down the leagues the absorption has traveled.

Championship? Probably mostly priced.
League One? Less certain.
National League? Possibly still generating signal.
Comparable leagues in smaller markets? Potentially significant remaining edge.

The tool hasn't been eliminated. The frontier has moved.

 

[FadeThePublic]

The public misunderstands xG in a specific way that might still create edge.

The mainstream version of xG tells fans whether their team was unlucky. Three goals scored from 2.3 xG: overperformed. Two goals conceded from 0.8 xGA: unlucky defensively.

The public application: "we deserved better" or "that result will regress."

The betting application: fade teams the public thinks are unlucky.

The public's xG reasoning produces consistent market distortions around narrative.

A team that loses 1-0 while generating 2.5 xG becomes a public narrative of unlucky deserving team.

The public bets them heavily next week.

If the market hasn't fully adjusted for this narrative distortion: fade the narrative.

The tool is priced in at the operator level. The public's misapplication of the tool creates new distortions the operator may not fully accommodate.

 

[oli_sussex]

The exchange perspective on xG adoption is specific.

The exchange price reflects the aggregate of all participant models.

2016-2018: xG-informed positions were visible as distinct from non-xG-informed positions through their timing and direction.

Early xG traders were positioning ahead of casual participants who were still using basic form and results.

By 2020-2021: the xG-informed position was no longer distinguishable from the aggregate because xG had become a standard input for most serious participants.

When a tool is adopted by the majority of sophisticated participants simultaneously it stops differentiating them.

The edge was in being early to xG. Being current with xG generates no edge over other current users.

 

[CoachTony_Bets]

The coaching adoption of xG is relevant here.

Most professional clubs now use xG internally for player recruitment, game planning, and performance analysis.

I use a simplified version in my own coaching work.

When coaches use xG to understand their own team and opponent tendencies, that information circulates in networks before it appears in public data.

The coaching xG application produces soft information that isn't in any public dataset.

Not fixing. Not inside information in the problematic sense.

Just a different layer of applied xG understanding that exists in professional circles before it's visible in match data.

Whether that layer creates betting edge for someone positioned to access it is an open question.