First off, I’d recommend setting all settings the same as below including creating tags to catch the skipped matches & performers - it’ll make it easier to go back and correct issues.
`
Now for the disclaimer: If your library contains content from any of the major studios (Think Brazzers, BangBros, Mofos, TeamSkeet, RealityKings etc etc) then you will get incorrect matches due to contaminated fingerprints.
Back in May I nuked my 32000 scene stash instance and started again. Decided I’d use identify to rebuild because ‘It can’t be that bad, surely?’. For the most part where fingerprints are reasonably clean then the scene will match 100% no problem. Auto accepting potentially erroneous matches is always going to be weak link in the process. Conservative estimate is that overall incorrect matches were still under 5% but a good proportion of them were from the studios above. I posted some examples of questionable matches on Discord at the time
The real weakness of Identify is performer matching. Obviously single named performers will always be problematic so that goes without saying. As it currently stands Identify works on a ‘first match is a good match’ scenario. If the scene performer is Mary Smith and you don’t have a performer called Mary Smith but you do have Jemima Jones who has an alias of Mary Smith then identify will apply Jemima as the correct answer on all affected scenes. Multiply that scenario by the larger number of scenes and it’s a well hidden draw back.
Long story short: After still finding issues appearing, I once again nuked my stash instance on October 11 & am in the process of correctly assigning what is now 46000 scene collection by using tagger and verifying the results.
As always it’s horses for courses and I wouldn’t discourage anyone that really really wanted to experience the entertainment of thinking you were going to watch one thing and it turned out to be something completely different.