The majority of the analytics on this site are derived from my Expected Goals (xG) model, and this plot is no different. By focusing on shot quality over box score stats, xG provides a more stable metric by drawing information from every shot attempt rather than only those that become goals. For example, if you were assessing the performance of a team on a given night, xG takes into account the quality of the ~30 shots instead of just the ~3 goals. By expanding this to include the opponent's xG, we can calculate an xG For Percentage (xGF%), a metric that identifies which team earned the greater share of the expected goals.

xGF% = xGF / (xGF + xGA)

This is “expected” in the sense that we’re attempting to answer “What did we expect to happen?” given all of the information we have for each shot attempt by each team. We can contrast that to “What happened?” by plotting it against the actual goal share.

GF% = GF / (GF + GA)

If the two metrics align (“Bad Process, Bad Results” or “Okay Process, Okay Results” or “Good Process, Good Results”), then we may have more confidence that the results are indicative of the process. In other words, luck or unmeasured skill appears to be a smaller factor in the team’s outcomes than the measured process.

If the two metrics conflict (“Bad Process, Good Results” or “Good Process, Bad Results”), then we should have less confidence in what we are seeing. We'd expect the GF% to move towards the xGF% in the long-run, unless the model is failing to capture something in the process that yields reliable results.

As with most statistical analysis, the date range you choose should alter your interpretation of the plot. Over a short range (ex. a single game or a week), the sample size is small so you are more susceptible to noise. This is especially true for GF%, because there are so few goals scored. Over a long range (ex. multiple seasons), the metrics are more stable, but teams will be more likely to drift to the middle (“Okay Process, Okay Results”) and you’ll also compromise the context as it will include roster turnover, coaching changes, etc. There is no “perfect” range, so sitting in the middle (ex. a few months of games) and/or comparing multiple ranges to each other is the best way to find the real signal amongst the noise.