It was lvl 40.
Directly: There’s a few situations here. Let’s consider a hypothetical book was submitted with a LVL 30 estimate, but a consensus level of around 20.
Situation A: Graders during temporary phase grade it as a lvl 20 book
If the graders consistently grade it as a lvl 20 book, the initial grade doesn’t have any impact. A book only moves out of the temporary system after there’s a good variety of grades (=, >, <), which in this case only occurs when it’s being compared against books ~ lvl 20. As comparisons in the temporary phase do not impact the grading of books they’re compared against, there is no ultimate impact.
The book enters the db as fully graded at lvl 20.
Situation B: Divergent graders during the temporary phase
This is the case that causes issues. Let’s say the graders in the temporary phase judge it at lvl 27 and it comes into the db as fully graded at lvl 27. Then it will take awhile to reach the consensus level of lvl 20. Additionally, it injects a lot of points into the global system, so all gradings are inflated.
However, you would expect divergent grades to randomly come into the system, sometimes higher, sometimes lower… so over the course of the entire db it hopefully has a relatively small impact and the notion of a ‘lvl 20’ book for instance stays mostly constant.
Outside of those injected points into the global system, this divergent initial grading does not have a lasting impact on the book. Once it reaches the consensus lvl of 20, it will just bounce around that level a little (maybe 19-21), but stay put.
Indirectly: You are right that situation B is encouraged somewhat by a bad initial estimate. Going back to our example, the first few grades you would be prompted for the lvl 30 estimated book would be against lvl 30 books you’ve read. There is of course a higher chance of divergent gradings if you’re prompted against books of a drastically different level to the consensus, rather than a similar level.
But I think this is still pretty well mitigated against in the new temporary rating system… it was much worse in the old one. And, in general, I think people attribute bad grades to poor estimates too much. It’s much more likely that someone genuinely thought a book was harder in the initial phase and situation B occurred.
The major caveat to this is high level manga submissions. High level manga are more likely to be compared against novels as there’s fewer of them, leading to a larger chance for situation b as obviously people have different opinions when comparing manga vs novels.
That’s perhaps a situation we could better mitigate against.
