People who hate xG can skip this post. For the rest, let’s discuss xGoal Differential and “Goal Differential Minus xGoal Differential.” xGD measures how well a team is performing measured by adding up both quantity and quality of shots taken and allowed without regard to whether those shots went in. GD-xGD measures how much a team’s actual goals scored and allowed performance differs from its xGD. A positive number is good or lucky, negative the opposite. My personal view is that GD-xGD is not 100% luck, but it probably is >50% luck.
Over sufficiently many teams, leagues and seasons, composite GD-xGD will approach zero. That’s how the system is built. In smaller numbers, even an entire season for one team, you can have pretty big variances, but outliers rarely repeat year-over-year, because IMO, it’s mostly luck.
- The NYCFC coach with the best per game GD-xGD is Jason Kreis at +0.25 per game. Keep that in mind if you are tempted to believe that good coaches find a way for their teams to outperform their xG. Kreis has the worst GD and worst xGD but his actual GD was significantly better than his xGD. And it’s not because Villa was a great finisher (nor that Saunders an amazing shot-stopper). Vieira and Torrent also had DV and Torrent’s numbers improved the year without him. PV also had a G-xG magician in Frank Lampard.
- Here is the per game xGD and GD-xGD for the rest of the NYCFC coaches:
- Vieira +0.17 / +0.20
- Torrent +0.36 / +0.13
- Deila +0.59 / +0.08
- Cushing +0.42 / -0.58
- 2022 breakdown xGD / GD-xGD
- 2022 Ronny +0.81 / +0.42
- 2022 next 8 +0.87 / -0.25
- 2022 last 10 +0.05 / -0.85
- Until the injuries hit, Cushing kept the team xGD up very high, but NYCFC and opponent finishing took a wrong turn. Team still went 4-1-3. Then the xGD over the last 10 games dropped – by a lot – to mediocre. Not good, but far from terrible. The team should probably have 8-10 more points than it does in that period as the GD-xGD fell off a cliff.
- For comparison, Torrent’s breakdown is +0.17 / -0.21 in his first 25 games and +0.54 / +0.42 in his last 28.
- Cushing has had 18 games, roughly half of those dealing with major injuries.
Disclaimer – yes I do see things worth criticizing in how Cushing has the team playing even with all the allowances. But he’s very much not had a fair shake yet and did fine before the injuries went exponential. I expect the team sees this as well, and that unless we miss the playoffs he returns in 2023. If they don’t qualify, I’ll be curious to see if they follow the Kreis rule or not.
Things what I noticed while looking at this:
- Is Philadelphia lucky or good this year? Both. Per game xGD +0.72 / GD-xGD +0.70
- Austin is very, very fortunate. Per game xGD -0.20 / GD-xGD +0.70
- Only 2017 Atlanta finished a full season at +0.70 or better on per game GD-xGD.
- Counting both 2022 Union and Austin, only 11 teams have had a full season of +0.50 per game or better GD-xGD. This is based on 228 team seasons from 2013 to present on ASA.
Leave a Reply