That's a good question.

Many estimates can be easily checked when you have access to a data source (encyclopedia or the Internet), e.g. object heights, distances, populations etc.

Other estimates are more complicated to check (e.g. probabilities). In that case you can attempt to estimate the same thing using different techniques. This is useful for debugging and may give a general idea of your accuracy (if 3 independent estimates are close to one another, you are likely not mistaken by too much).

Also, its easier when a few people independently estimate the same thing. You can compare your results, discuss the intermediate steps and find errors. This is a great feedback, from my experience.

Is there value in the practice without feedback?

I believe there is. It's valuable as a game and simply as training. Also, sometimes any estimate is better than nothing.