But I actually have very limited experience with TB. Wandb's way (running avg) works very well for all ML researchers I have encountered so far

Just for the reference, smoothing implementations in wandb

Hi @<1536518770577641472:profile|HighElk97>

Is there a way to change the smoothing algorithm?

Just like with TB, this is front-end, not really something you can control ...
That said you can report a smoothed value (i.e. via python) as additional series, wdyt ?

As for the TB you are mentined, they have a debiased version

have absolutely zero experience with .ts , sorry about that (
not sure, if this will fix the issue for me, though:
still will need the simple average and the gaussian smoothing (useful for uneven x-spacing between points).

Different series should have different smoothing intensity. This is impossible to tell in advance of looking on charts.

I realize this is the frontend , so maybe it is impossible to change right now. Really don't understand how to use clearml as a dashboard to compare runs without reliable smoothing.

Interesting, do you think you could PR a "fixed" version ?
https://github.com/allegroai/clearml-web/blob/2b6aa6043c3f36e3349c6fe7235b77a3fddd[…]app/webapp-common/shared/single-graph/single-graph.component.ts