1. Download part of torrent through any torrent client. I use partially downloaded huge torrent and torrent-file for one of parts of huge torrent.
2. Add this torrent to kTorrent with recheck.
3. Look at "time left" column.

Actual result: kTorrent show very low value, depends on torrent size and already downloaded %. When download started again "time left" value slowly growth but this process very slow. For example I have 100Gb-torrent downloaded on 90%, I seen "4 days" almost month ago and now there almost same value, average speed around 0.5-2 kbite/s (huge part downloaded quickly but initial seeder gone).

Awaited result: kTorrent shouldn't keep all information about torrent speed and should drop too old values. kTorrent shouldn't use any information from recheck period and try to calculate "time left" using this information.

Lain_13 wrote:1. Download part of torrent through any torrent client. I use partially downloaded huge torrent and torrent-file for one of parts of huge torrent.
2. Add this torrent to kTorrent with recheck.
3. Look at "time left" column.

Actual result: kTorrent show very low value, depends on torrent size and already downloaded %. When download started again "time left" value slowly growth but this process very slow. For example I have 100Gb-torrent downloaded on 90%, I seen "4 days" almost month ago and now there almost same value, average speed around 0.5-2 kbite/s (huge part downloaded quickly but initial seeder gone).

Awaited result: kTorrent shouldn't keep all information about torrent speed and should drop too old values. kTorrent shouldn't use any information from recheck period and try to calculate "time left" using this information.

Which estimation algorithm are you using ?

I don't know how to calculate "time left" well but current method shows me very strange values. I think problem in calculation of average speed. Can you explain in a few words how average speed calculated in 3.2dev?

Had a look at the average speed calculation in the time estimator, it seems we do not take into account the number of imported bytes. Will fix that.

In the mean time in the advanced settings change the algorithm to Window of X or moving average, this will probably give you better results.

George wrote:Had a look at the average speed calculation in the time estimator, it seems we do not take into account the number of imported bytes. Will fix that.

Ya, it always gives weird numbers when you add a torrent that already has data - especially when you have almost all of it already. I'd always figured that that was just life though since I've seen no evidence that ktorrent actually keeps track of bytes found when the torrent is created or found in additional checks later (for instance, if you get one of the files elsewhere and replace one of the downloading files with it). It's always seemed like it just treats that data as data that it downloaded. The fun thing that that does beyond the speed calculations is that it throws off your real seed ratio. You could have downloaded half of the files elsewhere, but ktorrent still counts them as downloaded and thus against your ratio.

I'd always figured that that was just how life went, but if ktorrent actually keeps track of how much data is found during integrity checks and counts that separate from the amount downloaded, then such issues could be fixed.