David Bal's goal had a xG of 0.07, but a xOnTarget of 0.04 according to the A-League xG tweet I saw, cos I noted at the time how stupid that is. A shot can't be more likely to go in than it is to hit the target - tells you exactly what the xG stat is worth.
xG in its current format is a poorly understood statistic that is actively misleading when applied to small sample sizes and really shouldn't be used to analyse single matches. It provides a better picture when applied over a season to identify trends, but the algorithm behind it isn't detailed enough to accurately identify how likely any given shot actually is to go in. They've done a reasonable approximation based on historical data but it is very prone to putting strange percentages on routine shots.
xG in its current format is a poorly understood statistic that is actively misleading when applied to small sample sizes and really shouldn't be used to analyse single matches. It provides a better picture when applied over a season to identify trends, but the algorithm behind it isn't detailed enough to accurately identify how likely any given shot actually is to go in. They've done a reasonable approximation based on historical data but it is very prone to putting strange percentages on routine shots.