BakaBT

BakaBT => Site Feedback & Rules => Topic started by: Astara on February 24, 2012, 06:37:39 AM

Title: 'Minor' discrepancy found in upload stats...
Post by: Astara on February 24, 2012, 06:37:39 AM
Was looking over stats on my page to see which series had uploaded the
most, and seeing where that placed me on the torrent's seeder page.

Trouble is the stats on my user page, say I've uploaded 68GB for Simoun, but on
the torrent page, it says I've only uploaded about half that.  Hmmm....

How is it that there would be 2 sets of numbers and why would they be over 2x apart?

It looks like the torrent page is the one in error, -- if you compare hours or hits, I'd be at #6 or #7 -- and the folks in those position have 65 and 66GB uploaded. So the 68GB on my personal page would statistically correlate with #hits and hours almost exactly... 

I checked another one -- think it was #2 or #3 on my page -- same problem.

So why is it that the torrent page seems a bit behind -- I know it says it might be 5 minutes behind, but I don't think it's gonna catch up that much in 5 minutes... ;-)...

 
Title: Re: 'Minor' discrepancy found in upload stats...
Post by: Mirgond on February 24, 2012, 09:46:38 AM
You didn't seeded that torrent once for 3month (iirc, could be less) or more.
Your public stats are resettet then while you see the full stats on your profile.
Title: Re: 'Minor' discrepancy found in upload stats...
Post by: Astara on February 24, 2012, 06:27:32 PM
Hmmm....Must of been that time I was seeding for 6-8 months and my client wasn't registering upload/download amounts, though I still got bonus points for carrying and it still listed the torrents I carried...  That was annoying...  :-(

'side I thought it was the personal stats that got reset after 3 months?  Not the webpage stats...?   *sigh*...
Title: Re: 'Minor' discrepancy found in upload stats...
Post by: Mirgond on February 24, 2012, 07:21:50 PM
No, personal stats are reset after 30days, if you disable the keeping.
Title: Re: 'Minor' discrepancy found in upload stats...
Post by: Astara on February 24, 2012, 10:36:50 PM
No, personal stats are reset after 30days, if you disable the keeping.

Hmmm...So how is it that I would have one reset (the torrent page), which takes (???) 3 months(?) of inactivity, but not the personal stats which only take 30 days?...  I mean if I stopped seeding for the period it took for the torrent page to be reset, wouldn't it be certain to have reset my personal stats?

I.e. I don't recall a time when I was seeding that I wasn't seeding that torrent, but @ only 6k hours, that's not even a year.  Sigh...  Why keep two sets of books, er, figures?   

If one wanted to give an accurate picture, and that member hadn't zero'ed their personal page stats, then it seems like propagating their personal stats to the torrent page would give a more accurate picture of where they'd been putting their bandwidth?... 

Given that my stats on my personal page are visible to anyone who clicked on each person downloading, is there a positive reason to give some lower number on the torrent page than what would be shown on a personal page?

I'm not suggesting keeping any more data than you already do -- i.e. if my personal page was 'zero'ed', then there'd be no record of how much I'd uploaded -- but since the info is available to other users, why have the torrent page show some lower number?

FWIW -- that was one of my first torrents -- and haven't NOT seeded it as long as I've been seeding, but that doesn't mean the non-reporting of bandwidth problem I used to have didn't somehow cause that -- but as I did get 'seeding credit' for all the bonus torrents I was carrying then, (was the only credit I got for seeding during that time!), then it seems the any torrent page shouldn't have thought I wasn't seeding 'it' ("as if" each torrent page had it's own view of my seeding -- which seems unlikely!).    I mean if the bonus torrent think I am seeding them, then wouldn't all my torrents have thought I was seeding them even if my actual U/D amounts weren't changing?

(Too much I don't know about the algorithms behind all this -- but just seems that the torrent pages should have, at least as much 'GB' recorded as the person has on their personal page -- but without knowing why you'd keep two separate sets of figures, it's hard to say for sure). 

Note -- am talking specifically about a case where I haven't tried to zero out my personal stats and have left them in a default 'open access' state (which I think may be limited to logged in members)...so it's not like I'd be trying to hide what torrents I'd seeded or not... 

If what I say makes sense, correcting the algorithm to use max(personal-page, torrent page) would seem to be a good idea   ;D, but if it doesn't make sense, .. oh well!...   :'(   C'est la vie.   ::)
Title: Re: 'Minor' discrepancy found in upload stats...
Post by: datora on February 25, 2012, 07:31:52 AM
.
As noted, if you disable keeping stats on your tracker profile, data will get deleted.  But, you seem to be keeping all data by default.

So, the stats on your tracker profile are what bakaBT has recorded in its database of your total download/upload for any torrent on your profile, no matter if active or inactive or if it keeps getting switched between them.  It is an absolute cumulative.

You can pause torrent activity, even stop it, in your client.  But, so long as the torrent is never deleted from your client, you can restart it and the BBT tracker page for that torrent will keep track (for the most part) of your cumulative.  Delete the torrent from that client &/or recheck it on that client &/or begin seeding it with a new client and the stats on the offer page will reset (including setting your download to zero), but your personal tracker profile will still reflect the cumulative total according to your personal hash.  I suspect if you have to reset your hash that you might lose your statistics, but I'm not sure.  BBT would still track your total upload/download ratio, but the stats for each individual offer (I THINK) would get wiped.

What is shown on the offer page for a torrent is the most recent "session" you have been active in.  This is the last time that you either downloaded a torrent, or began a reseed event where you actually had to recheck it with a client.

Examples (numbers are invented, not actual) --

Download a torrent 3 years ago.  You never delete from your client, it runs on that client the full time (Stops, Pauses & Starts don't count) ... the stats on your profile and the stats on the offer page will match (mostly).

Let's say you lost a hard drive a year ago and had to rebuild, including reinstallation of your client and then you pulled the torrent files out of archive and reseeded them.  Your profile will still reflect your download, your upload and your ratio stats.  However, the offer page will now show you as only seeding from the point when you reseeded the torrent ... total hours running, total idle time, your download will be set to zero, your upload will begin from zero again, ratio displayed, etc.

If you have two computers with clients running, and you seed the same torrent from each computer (NEVER do this at the same time!), the offer page will display the stats for the currently active client, and it can get very messy if you turn the torrent on and off several times on each client.  However, your tracker profile will still (mostly) keep accurate statistics recorded against your personal hash.

I've got a few dozen torrents I've been through this with and have watched the stats carefully enough to observe all this.  I have not had to reset my hash ... so, don't know what that will do.


Final thing to remember: very often there are minor variations in the recorded stats between your personal profile and your client.  So, if your torrent is downloaded and never rechecked/reseeded, your personal profile might show slightly more or slightly less upload than what your client shows.  I've got a lot of stuff like this on my profile.  Very typical is to see something like a ratio on my client of 13.785 while my profile shows 13.453.  Most of the time it's far closer or nearly exactly the same, such as 15.387 on my client and 15.403 on my profile.

There's just a bit of noise in the system, which isn't usually visible at lower ratios, such as 2.154 or something.  I usually don't observe discrepancies until ratios exceed about 7 or 8, and often they can be exact up to 20 or more.  Also, noise like that seems to increase on highly active torrents, such as new swarms when several hundred people are all seeding & leeching simultaneously for a week or so.


Finally, sometimes the system does just brain fart.  I've D/Ld a 6 GB torrent and had only 2.5 or 4.2 GB registered as D/L on my profile.  More rarely, I've seen a few hundred extra MB extra D/L added.  I've had download credited to me on freeleech a couple times, and I've received an effective "freeleech" about twice now.  Out of some 750 or torrents, this last set of weirdness has only happened about 10 or 12 times total, and usually on stuff that's 5 or 10 GB in size or larger.


tl;dr - the BBT tracker isn't perfect and some small discrepancies are normal.  Very rarely (1 out of 100 large downloads) there might be a large discrepancy.  What is displayed on your tracker profile is the cumulative for all time for your stats; what is displayed on an individual offer page is only since the most recent re-check or reseed event on the client you are currently running.

That help at all?
Title: Re: 'Minor' discrepancy found in upload stats...
Post by: Astara on February 25, 2012, 09:30:23 AM
It answer the question of how it arises, it doesn't answer the question of 'why would you want inaccurate stats *IF*, you have the ability to have more accurate stats?

Is there a benefit in showing the less accurate stats?   What's the point in showing some fraction of what the person has seeded, if you *know*, they've seeded more?


Obviously, if the person has deleted their private stats, then the information is gone, and that would be reflected in the torrent pages, but if someone bothers to keep their stats around, then since they are official stats from baka, why shouldn't they be used?

Knowing that more accurate stats are available and not making use of them seems somehow less than honest?  I.e. my main question -- is why use inaccurate statistics if baka has more accurate ones.  I'm only referring to ones on baka's own servers, not client stats.