Stars represent a players contribution to the game.
These are not to be used to compare teams but can be useful for comparing players who are playing in the exact same position. Also, it was the analysis of stars that helped experienced managers in determining the percentages of contributions for each position (e.g. 75% scoring and 25% passing for forwards).
Stars are evil
Stars are considered "evil" because they can be misleading about a player's contribution to the team's performance. For example, an inner midfielder who produces three stars may have contributed one star to attack, one to defence, and only one to midfield; this player may do less to help your team win than if he had contributed only two stars, but all of them to midfield. Another example would be an inner midfielder who earns two stars when played defensively, but four stars when played offensively. Improper use of a players skills can lead to a poor star rating, making a very good player seem bad. A team's total star rating does not suffer from this exact problem. Rather, the issue here is that all contributions are scaled equally even though some ratings are more important than others — most notably, midfield rating. There are much better ways to compare overall team performance, such as HatStats and LoddarStats.
As long as the impact of secondary skills is carefully considered, stars are a useful tool for comparing individual players.
Saying that "stars are evil" can help newbies understand that they need to look at match ratings to understand the outcome of a match (so it reduces comments like "I should have won, I got 22 stars and he had only 20").