[Request] Transfer seedtime on trump

[Roardom]

First, add a way to declare a torrent as being trumped on deletion.

On the first announce a user makes on a new torrent that trumps one or many older torrents, transfer AVG(immediately trumped torrents seedtime)/2 + AVG(immediately trumped torrents _bonus_ seedtime) to a bonus_seedtime column to the new torrent. Incorporate this new bonus_seedtime column into the bonus point seedtime calculation.

Reasoning is to encourage users to permaseed the new torrents and it will increase the user's average seedtime

[markhc]

I've been doing some math on this because it seemed to me that this new system heavily discourages the snatching of new season episodes for TV Shows, which is not something that I think should be discouraged.

If you could take a look at it and see if I went wrong somewhere that'd be great, but this is the scenario I was comparing:

A user seeding 8 torrents for a full month starts grabbing individual episodes of a new tv show, released weekly for 12 weeks. After the last episode is aired, a season pack is made that immediately trumps all older episodes.

In the old system, with all trumped torrents removed from the calculation, after 12 months the average seedtime in hours for this user is 9040 hours, with 9 total torrents counted:

(9360 * 8 + 6480) / 9 = 9040

With the new system, they now have 21 total torrents (12 of which do not accrue any additional seed time). I have included the average seedtime of these 12 torrents into the starting seedtime of the Season pack (Avg EPs Seedtime is 1152, half of that went into the pack's Bonus Time).

(9360 * 8 + 2160 + 1944 + 1776 + 1608 + 1440 + 1224 + 1056 + 888 + 720 + 504 + 336 + 168 + 7056) / 21 = 4560

While initially after the pack drops the average seedtime is indeed higher, this soon falls off as they have more torrents not gaining any seedtime. So after 12 months the average seedtime for this user would be 4560 hours with the new system, almost exactly half of what it would be with the old way.

Of course this example is a bit contrived, a real active user would be downloading other things that hopefully aren't trumped, but they probably will have a fair share of trumped torrents, at least if they watch ongoing tv shows. Spreadsheet with the calculation linked below. Thank you!

[Roardom]

I missed an important part: if the user seeds the torrent that trumps, then the trumped torrents are excluded from the seedtime average. If the user doesn't seed the trumping torrent, then the seedtime for the trumped torrents are included in the average.

[markhc]

I missed an important part: if the user seeds the torrent that trumps, then the trumped torrents are excluded from the seedtime average. If the user doesn't seed the trumping torrent, then the seedtime for the trumped torrents are included in the average.

Ah gotcha haha Yeah then this is a great system indeed. Thank you for the explanation! (and sorry about the confusion)

[Roardom]

@markhc

It should also be noted that the tracker you're commenting from has implemented its own closed source method of "trumping". Stock UNIT3D currently has no concept of trumping - it sees all deleted torrents the same. The most you can "guess" a trump to be is a deleted torrent with an approved status and hope that all non-trumps were rejected first before deletion. UNIT3D also doesn't save in the database anywhere of the two torrents involved in a trump.