Ah, the amount of news will likely increase before it decreases!
I like your close examination of the model. And your perspective would work. However, in this case, I prefer the idea of having a field that is a preexisting instance that remains for the duration of the game. It is a bit disconcerting to think of the field being deleted.
Let me clarify my use of English in the state model and in the association text phrases.
In the state model I used “out-of-bounds” to mean “outside the lines marked on the field”.
Thus, a ball is out-of-bounds when it is beyond the marking lines on the field. So, the field itself actually includes (some) space outside the bounding lines.
On the class diagram I use “bounds” as a verb synonymous with “provides marking lines”.
A field bounds zero or one ball. In English bounds usually speaks of physically restraining, but here I stretch that meaning just a bit. :)