Thanks for reporting, we'll try to reproduce and come back to you.
Yours, Steffen

Hi, can you provide us with an exact query to reproduce this ?
I used the Example1 with all defaults, and got different result scores. yours, Steffen

Hi Steffen,
Thank you for your reply.
I do not have access to my data until January 8. I will send you the data used to reproduce the request as soon as possible.
Yours, Bénilde Bonnefille

Hi Steffen,
As discussed, you will find attached the data entered for the query for which I obtained the above results. The search was carried out using the default parameters for an analysis in positive ionization mode with an ESI source (all MS levels). The data are presented as pairs m/z / intensity ratio without any pre-selection of the experimental data.
Yours, Bénilde
MS2_oxazepam.txt

@meier-rene or @Treutler or @sneumann or others - any progress here?
We have just received emails to check progress; an update would be great. Thanks!

I guess its my pleasure to fix this. I already found the code location and its more complicated than I thought in the first moment. The program actually calculates a score and its not only 0.999, other results do also appear. Its just not correct or at least not identical to the results we could obtain with the old version. I have to trace were the numbers come from and I need to compare them with a few examples I calculate by hand following the original MassBank publication.

Hi, I think we could fix this problem and its online available on our development server https://msbi.ipb-halle.de/MassBank/ . Could you please test there and report your findings?

Minor issue found: The 'Instrument Type' selection is not taken into account.

Hi,
First, thank you four the feedback.
I just did the requested test, and it seems much better! (see search results below)
However, the results are different from those obtained with the previous version: the highest score is not the same as on the results initially obtained (first example in the first message of this discussion), and the best "match" does not correspond to the same MassBank spectrum although the research data are identical. Unless these spectra have been removed from MassBank, there may still be a problem there?

Concerning the 'Instrument type': the results does not show any problem of instrument type selection, I only get results in ESI, as requested (for the first 20 results).

I wrote too quickly: I confirm the 'Instrument type' selection problem. I requested only (LC-)ESI-QTOF results (always with the oxazepam data) and obtained all the ESI Instrument type results.

We have deployed a new code base on our test system. Please try! We are quite confident, that the results of the similarity score and the number of hits are now equal within the accuracy of the algorithm compared to the old version. Nevertheless you will find some minor differences in the floating point numbers, but that's not very unusual after a refactoring of code, a change of java version, different operating system and different hardware. Floating point operation have always a limited accuracy.
Nevertheless there is a difference in number of matched peaks for your example (fig. below) which we can not explain. I'm wondering how you could obtain a hit count of 86 for the second hit in the old Massbank if the query spectra has only 82 peaks. There is something strange with this numbers. 82 is definitely the max with your attached spectra. Maybe the old MassBank was giving incorrect values for this field...

)

Thank you very much for your efforts @meier-rene
Look forward to hearing the feedback from @BBonnefille to see if this indeed resolves the issues they experienced

Hi, I performed the test as requested and I get results in agreement with those obtained before the update of MassBank (I get the same results that the ones you attached).
Thank you for all the work you have done to correct the issues we have encountered.
I hadn't noticed the problem of the number of hits that seems to be solved with the corrected version you propose. Perhaps this problem was due to the difference in the acquisition mode of the reference spectrum data versus the submitted data (low resolution in opposition to high resolution).
Thank you again,
Bénilde