“Expected goals measures the quality of chances and how often we expect them to be scored.“

All goals count the same in soccer, but some are a lot more difficult to put into the back of the net than others. Take for example these two goals scored by CFC last season:

Gooooooal! Dalton’s Damian Rodriguez @pacharo1220 scored his first professional goal. #AlwaysForward pic.twitter.com/SCroweXBQK

— Chattanooga Football Club (@ChattanoogaFC) November 14, 2021

Young @pacharo1220 with the speed, the skill, and the assist. @Naglestad10 finishes it. 1-0 CFC in the 10’. For live play-by-play, follow @CFCmatch. pic.twitter.com/0V9DEyzkV6

— Chattanooga Football Club (@ChattanoogaFC) November 3, 2021

Both of these goals go down as a shot on target and a goal, but just looking at the two we can see that Damien’s goal was a bit harder to convert than Markus’s; if we gave them both the same chance a hundred times, we would expect Markus to find the back of the net more often. Is there some way to quantify that difference? Of course! Expected goals measures the quality of chances and how often we expect them to be scored. Now, instead of seeing only the number of shots a team has taken, we can see how often we might expect them to score with all the chances they have in a match. This helps when looking at the flow of matches, as we can now tell if a team is creating quality chances or just getting lucky with wonder goals. We often find ourselves thinking “we deserved more from that match” or “we got away with that one” and now we have the numbers to back up our statements.

So, for the two chances above, our NISA xG model (which still needs more training before the season starts) thinks that shots similar to Damien’s will go in about 1% of the time at the NISA level, while Markus’s would go in about 61% of the time. This does a much better job at summarizing chances than flatly stating that both were shots on target.

How exactly is this xG model created? How is this one NISA specific? We’ll post about that soon!