PSxG — Post-Shot Expected Goals

Definition

Post-Shot Expected Goals (PSxG) extends the standard xG model by incorporating information about the shot's actual trajectory — where on the goal frame the ball was aimed, its height, speed, and placement. While xG measures the quality of the chance at the moment the shot is taken, PSxG measures the quality of the shot itself.

History & Origins

PSxG emerged as a natural extension of xG models in the mid-2010s, driven by the recognition that two shots from the same location can have vastly different probabilities of being goals depending on where they're placed. A weak shot straight at the goalkeeper from 10 meters has the same pre-shot xG as a perfectly placed top-corner strike from the same spot, but wildly different actual scoring probabilities.

StatsBomb was among the first major providers to publicly integrate PSxG into their data offering, using it to separate shot creation quality from finishing quality. Opta / Stats Perform also developed their own PSxG models.

The concept gained mainstream visibility when FBref began displaying PSxG data alongside standard xG, allowing fans and analysts to compare them directly.

How It Works

1. Start with all the features used in a standard xG model (location, body part, game state, etc.)
2. Add shot placement features: the x,y coordinates of where the shot crossed the goal plane (or where it was heading), shot speed, trajectory angle
3. Train a model on this enriched feature set

The difference between PSxG and xG for a single shot reveals finishing quality:

• PSxG > xG → the shot was placed better than the average attempt from that position
• PSxG < xG → the shot was placed poorly (e.g., straight at the goalkeeper)

Key Applications

• Goalkeeper evaluation: the difference between PSxG faced and goals conceded measures shot-stopping ability. A keeper consistently conceding fewer goals than their PSxG faced is performing above average.
• Finisher evaluation: cumulative goals minus cumulative xG shows overall finishing, but PSxG isolates the shot placement component from the chance creation component
• Separating creator and finisher credit: if a player consistently receives high-xG chances (good creation) but converts at a low rate, PSxG reveals whether the problem is placement or bad luck

Limitations & Debates

• Requires trajectory data: not all providers capture shot placement, limiting availability compared to standard xG
• Goalkeeper positioning not always included: some PSxG models don't account for the GK's position at the moment of the shot, which matters significantly
• Sample size: PSxG differences are noisy at small samples. A player "outperforming PSxG" over 10 shots means almost nothing; over 500 shots, it's meaningful
• Blocked shots: PSxG is typically only calculated for shots that reach the goalkeeper or go in — blocked shots are excluded because their trajectory is interrupted

Relationship to Other Metrics

• xG (see xG - Expected Goals) → pre-shot probability, PSxG adds the shot trajectory
• Goals - xG → total finishing over/underperformance (includes both placement skill and luck)
• Goals - PSxG → isolates luck/GK performance from finishing placement skill

---

Tags: #football #analytics #xG #PSxG #metrics #finishing