The estimate would probably be unreliable. For example, the type of question (in terms of difficulty) matters more than the number of questions that you answer correctly. And were your problem practice sessions timed and were they in blocs of 38? Did you sit at a desk and prop the book vertically? Did you take small breaks in between some questions? And was this your first attempt at the questions (the OG 12 and OG 11 contain many repeats)?
These types of questions, and many more, cast doubt on the value of any estimation based on one's OG performance. Without any further information, the best that we can do is state that (approximately) 75 percent on the OG is good, but not great.
The best indicator of one's performance and the most realistic simulator of the actual test is the GMATPrep software. The Manhattan CAT's are usually ranked (a distant) second place.
Hope this helps.