It's right, you should not assume that BPM databases are accurate. In fact, it's impossible to be 100% accurate, no matter how hard you try (not to mention songs with tempo changes, or those without a steady tempo, etc.).
But it's wrong to say all services rely on the exact same algorithm. BPM detection is done by detecting relevant beats, that's the only thing such services have in common. Such tools can be made in JS, C#, Python, etc. The logic behind them is the same, but the way it is done isn't.
For example, we use Artificial Intelligence to work our magic. Basically this is done by training a neural network, feeding it with as many different songs as possible (needless to say you need to prepare such datasets first, which is time consuming). In short the more the NN trains, the more accurate becomes our algorithm. And it has nothing to do with the one Traktor uses (even if some bpm-finding services use the same open-source scripts).
For this reason we update the script behind our tools each time it trained. In the end results you get today won't necessarily be the same next year.
Of course, we can't update the whole db (that's just for new entries), but we encourage users to report bad tempos (it's a built-in feature since the beginning). We also perform manual checks (songs are then tagged as "verified"), but we don't do this a lot, we are just 2 working on gsbpm, and we prefer to focus on developing the site.
Hope this helps. Any help is greatly appreciated btw, feel free to can contact us if you're interested in verifying tempos (or even creating a MusicBee plugin thanks to our public API).
cheers
The largest database of beats per minutes in the world. Access to millions of songs tempo.
