If I have a keyword search query and I get 2 documents one with score 90 and other with 91 (and assuming that the best match we can get on the currently indexed data for any query is 100), using minmax normalizer gives the score for document with score 91 as 1 and the score for document with score 90 as close to 0 (1e-45). Firstly this creates a huge gap in the scores for both results, and secondly even though they both have decent scores in keyword search the document with lower score suffers become of this normalizer.
I see that L2 normalizer is mentioned in the docs in 9.0.3 version but when I checked elasticsearch's repo I couldn't find that normalizer in this release. If there any update on this new normalizer? Or better yet, did anyone face this issue with minmax normalizer and have any solutions for it?