Incorrect average benchmark scores

Forums Benchmark Discussion Incorrect average benchmark scores

This topic contains 1 reply, has 2 voices, and was last updated by Admin 5 months, 3 weeks ago.

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #2365

    thermal
    Participant

    I have just provided benchmarks for my Mac Mini 2018 6-core i5 at various buffer settings. I noticed that one other user has provided a similar benchmark score to mine. Mine were around 130-135 and his was 124, IIRC. The 8 vs 16 GB RAM difference and/or different audio interface probably accounts for that.

    However, there were two other submitted scores which must be incorrect. One was 75 and one was 20. I assume the one at 20 was supplied by the admin just to create an entry for this model of Mac Mini. The problem is that this score, and the one at 75, skews the averaged total benchmark score for the 2018 i5 Mini, ending up at a little above 100 when it should be closer to 130. If the same issue applies to other Mac models as well, the benchmarks overview cannot be trusted.

    #2368

    Admin
    Keymaster

    Seen and removed! I saw that there are quite a few models for which there are scores of 20 tracks, which seems unrealistic for most models, and I believe it was spam because 20 is the lowest possible score you can select and therefore the default score selected on the “Submit a Score” page. I guess these scores were added by bots.

    So I added a user-login feature through which you can now add scores to avoid that. That meant that all previously submitted scores had to be assigned to my account and it displays my username everywhere. I didn’t add any scores myself, some models don’t have any scores at all: https://logicbenchmarks.com/apple-mac-model/macbook-air-core-i5-1-6-ghz-13″-late-2018/
    (they’re not displayed in the chart but can be found through the drop-down search menu)

    As for the incorrect average, this is something I have to correct. Instead of having the arithmetic mean displayed (which is the case right now), I’d like to have the median shown, which is less sensitive to outliers. It’s a loss of precision, but very robust to extreme values, and I think that’s more important.

    • This reply was modified 5 months, 3 weeks ago by Admin.
Viewing 2 posts - 1 through 2 (of 2 total)

You must be logged in to reply to this topic.