Quantifying how estimates compare to a gold standard

This may be too basic of a question but I would greatly appreciate any guidance.
Say you have a dataset where a series of people have visually estimated the number of dots in a series of samples (say in a series of boxes). Almost everyone overestimates, it turns out.
And you count the actual dots for each sample, establishing the 'gold standard'.
How would you come up with a coefficient/correction factor to describe the overestimation factor and validate it? So that you can more accurately predict the real number in future visual estimates... how big of a dataset would you need?

Does that make sense?
Thank you so much!


Less is more. Stay pure. Stay poor.
You could fit a linear regression with the outcome being the number of dot and the predictors being variables of interest. The output will give you predictions and tell you which attributes may be associated with under or over predictions. You can then intervene or use this info in the future to help guide the process.