We haven't talked much about xPoints, and I'm not going to explain it or try to sell you on it, except it is loosely like Pythagorean baseball in that it calculates a team's most likely point value based on other underlying stats. It also works best over large data sets so teams can do significantly better or worse than they "should" in 34 game seasons. But 10 game samples are really, really small, and right now 3 MLS teams are doing better than predicted by amounts greater than any team has ever maintained over a full season, and 6 of the top 7 are 2021 teams with data going back to 2013.
The number is the amount of extra points per game they have earned over what the model predicts:
2021 LA Galaxy 0.90
2021 Orlando 0.65
2021 New England 0.62
2014 DC United 0.56
2021 Seattle 0.55
2021 Columbus 0.51
201 Colorado 0.48
Outside of 2014 DCU, the most any team exceeded its xG per game over a full 34 game season was by 0.43. The point is, those six 2021 teams are probably getting points at a pace that they cannot possibly sustain for 34 games. The Galaxy number is in another, well, galaxy. They are earning almost a full point per game than the model says they should. The Orlando and Revs data is good news for NYCFC. NYCFC has a negative per game variance of -0.34 so far. That's low, and will probably improve, but it's not so bad it could not possibly be sustained or even get a bit worse (not likely but possible). Also, the Crew have a mediocre record and probably have not played well enough even to deserve that. They are arguably the most overrated team of pre-season 2021.
On the other end, the current 4 worst single season teams all time are also from 2021:
2021 Chicago -0.93
2021 Toronto -0.76
2021 LAFC -0.61
2021 DC United -0.51
This is again a function of very small sample sets of roughly 10 games. Chicago and Toronto are not exactly good, but they are better than their record. Chicago's number is as unsustainably bad as the Galaxy's is good. Expect LAFC and DC to show improvement and move up into playoff range. Nashville not quite as extreme (-0.44) but also should get better results than it has to date.
All time, NYC has a positive variance of +0.1 per game, which is second best over the years NYC has been in the league. As you see, the variances get smaller with more games in the sample set.