The pattern syntax for the calculated grade by fields is
Specifying the <value> part is optional and allows you to retrieve for calculation only occurences of the specified value. For most uses you can leave that empty.
Specifying the <pattern-of-supporting-field-which-returns-a-numeric-value> is required. In most recent 2.7 and 2.8 versions, the calculated grade is supported by the ratingmdl, select and radiobutton fields.
With the ratingmdl field you need to use the aggregate patterns, e.g. fieldname:max, fieldname:avg, etc.
With the select and radiobutton field you can use fieldname:key. You can also use fieldname if the field options are numeric. The fieldname:key pattern will return for calculation the selected option index, 1 to number-of-options. The fieldname pattern will return the option label if numeric and that could be any number you put there.
So, assuming you have a ratingmdl field named 'evaluation' and you want to assign a calculated grade by max rating, the formula would look something like
No equal sign required.
If you want to factor by 5, the formula would look something like
##:evaluation:max## + 5
With the select or radio button fields it would look similar but with their applicable patterns.
Note that the above applies only if users have only one rated entry. If users have more than one rated entry, these patterns will return to the formula a comma delimited list of values from all the user's entries, and you need to aggregate that in the formula. So for instance, with the 'evaluation' field and more than one rated entry per user, and you want to take the average between the max rating of each entry, and factor by 5, the formula should look something like:
average(##:evaluation:max##) + 5
Goes without saying (or not), you can use in the formula more than one field. For instance, you can add to an entry more than one grading field, say three, to create a grading rubric. Suppose these are ratingmdl fields named 'evalstyle', 'evalargument', 'evalcomplexity'; the fields scale is 100; the score is weighted 20%, 50%, 30%. The formula should look something like:
##:evalstyle:max## * 0.2 + ##:evalargument:max## * 0.5 + ##:evalcomplexity:max## * 0.3