What.cd changes required ratio system

[dragoi92]

Very good news. Now it easy to keep a good ratio and acc in what. Thnx staff of what.

[Funkin']

Thanks for your explanations dvdasacd(always a pleasure to read, if not somewhat confusing some of the time ) and whatman.

Your proposed system (both the seed hours and the higher base ratio) sounds a lot like what my original plan was - set the required ratio to 1-(seeding/snatched), where seeding is calculated an average for a 7 day time period (ie. your formula with the time moved). Thus, someone who seeds 100% of the time would have a required ratio of 0, and someone who seeds 0% of the time would have a required ratio of 1. Mathematically, this system is much nicer to look at and deal with than what we've implemented.

Now that sounds pretty good. As long as your not saying that everyone has to seed 24 hours of every single day to receive a required ratio of 0. As I'm sure that's just not practical for most.

I guess all we can do is sit back and wait to see if this system really does promote more leeching. If it does, and maintaining a positive ratio becomes just that much easier for everyone, then bravo for implementing something different that actually works.

[dvdasacd]

Your proposed system (both the seed hours and the higher base ratio) sounds a lot like what my original plan was - set the required ratio to 1-(seeding/snatched), where seeding is calculated an average for a 7 day time period (ie. your formula with the time moved). Thus, someone who seeds 100% of the time would have a required ratio of 0, and someone who seeds 0% of the time would have a required ratio of 1. Mathematically, this system is much nicer to look at and deal with than what we've implemented.

We decided to add the gradual required ratio multiplier and reduce the time to 72 hours for two different reasons. The multiplier was to ensure backwards compatibility and prevent outrage, as well as simply make life easier for the newbies.

Yeah I think I'd have the ratio requirements gradually rising like what.cd anyway. Although in truth it depends on what your site is. The Pedro's community is mostly fine with 0.8 across the whole community at all times, but they're just a different bunch of people, and up until recently, and even still now, have had nowhere near as many seedboxes in the community so it's been far easier to seed than at what anyway. The ideal is to work out an implementation which could work for seedboxes coexisting with normal users and everything still being fair, and I think it can be done that way - but in what.cd's case it's a matter of transitioning - what you've done is probably the best way to have scaled the algorithm for the time being. I think we're all going to see how it plays out at what and if things need to be tweaked further (to make it even fairer to download at what.cd while not making any other part of the system less fair) - I hope they will be.

So really, if you run the math, your formula and our formula are exactly the same - it's just that you consider "good enough" seeding time to be 7 days, and we consider "good enough" seeding time to be 72 hours.

Are you sure? Isn't it that mine was rewarding users small amounts for a torrent if they seed less than 72 hours in the last week and and larger amounts if they seed that torrent more than 72 hours? And yours is only rewarding people a set amount for each torrent seeded in each week, so long as it is at least 72 hours?

The 72 hour thing, while it may look a tad lenient of us, is because we consider someone who seeds 72 hours of the week to be seeding 'enough', and we don't want to place the harsh expectation of seeding 24/7 on them. It also gives them some buffer time where they don't have to be seeding - eg. formatting computer, going away for the weekend, etc. Not expecting people to seed 24/7 is almost exactly like not expecting them to maintain ratios of 1.0. While it may seem unfair for people who seed more than 72 hours per week (and while we do prefer people to seed longer), we feel that it isn't necessary to reward them with the requirement system - they'll be rewarded with the extra upload they get.

The reason I thought of my particular algorithm was because someone pointed out that seeding 10 torrents for 144 hours ought to be rewarded as much as 20 torrents seeding for 72 hours, but under the current what.cd system, it would only be rewarded half.

However, I just realised that to be even fairer and more accurate on rewarding a user's seeding contribution (which is what it's all about), you should reward them according to how many bytes of torrents they are seeding in those seedhours. Someone could download ten 20GB torrents and ten 5MB torrents and then just seed the 5MB torrents and be rewarded exactly the same amount as if they seeded that other 200GB worth of torrents that they snatched.

So an even better algorithm would be: Requirement = Current requirement*[1-(total sum of each torrent's seedhours multiplied by its size in bytes)/(total possible seedhours multiplied by the accumulative size of all snatched torrents in bytes)] where "each torrent's seedhours" is the number of collective hours seeded by that torrent in the past week and "total possible seedhours" is 24 x 7 x the number of snatched torrents - the most you could possible contribute seeding-wise. "Hours" being calculated by the minute, ofc.

I'm trying to think of how you would implement a "good enough" value (an equivalent of the "minimum bar" 72 hours thing in what what.cd algorithm) into this different algorithm which measures bytes being seeded, not how many hours being seeded. I don't think it can. There's only one variable in my latest algorithm, and thats the percentage of total possible seeding contribution of all your past downloaded bytes, in the last week. So then to set a "good enough" measure, perhaps you could just set some percentage (50%, which would be pretty dang good for someone who's downloaded a lot) and if you reach that, you are taken to 0.0. Depends what you'd consider as good seeding contribution and how much you wanted to reward the user for it.

I think doing it this way would be much better because currently if someone seeds a torrent for 71 hours they're not rewarded at all - yet just one more hour and they're rewarded the same amount as if they seeded for the whole 168 hours in the week. All you'd have to do to make the more literal approach fair, is scale it gentler by setting a lower bar and a faster upscale of how much you reward their seeding byte size contribution.

All in all, this entire concept is merely the same old system we've had, but now with an extra seed bonus system on top which the user can make use of if they find they need it (which many sure do). It's a very smart automatically-managed bonus system which all you have to do is scale it to how much you want it to help the user.

[WhatMan]

Actually, the way we explained the 72 hour average is oversimplified and doesn't quite reflect the actual math. If a user was seeding for less than 72 hours, it's still counted - however, when we calculate the average, we still divide it by 72. So if someone was seeding 10 torrents for 36 hours, the site calculates them as having seeded 5 torrents for 72 hours.

The seed byte idea sounds quite interesting, and I can see what you're trying to do with it. But the goal of that system is different from the goal of the one we've implemented. Our goal is to guide the user towards correct behaviour - downloading, seeding their snatches for at least 72 hours per week, and gradually letting their ratio rise over time. We don't consider someone who's seeding rosetta stone to be contributing any more, or exhibiting any more correct behaviour, than someone who's the sole seeder on a rare 1950's single. However, the seed bytes system would.

[dvdasacd]

Actually, the way we explained the 72 hour average is oversimplified and doesn't quite reflect the actual math. If a user was seeding for less than 72 hours, it's still counted - however, when we calculate the average, we still divide it by 72. So if someone was seeding 10 torrents for 36 hours, the site calculates them as having seeded 5 torrents for 72 hours.

Ahh now I see that my previous "original" algorithm was indeed the same as yours - excellent - I think the actual way it works should be communicated somewhere on the site, as some people think it's a literal minimum 72 hours per torrent.

So I take it that you currently award the user according to the total seedhours per week divided by 72, whose resulting value is divided by the total *possible* seedhours per week divided by 72 which can be written as 168N/72 where N is total number of snatched torrents...and THAT resulting value is multiplied by your "normal" required ratio (after being scaled according to how many GBs you've downloaded, of course) to get the actual required ratio.

So if you seed half your downloaded torrents for 144 hours in the past week, you're considered to have been seeding ALL your torrents "forever" in that week. So if you've downloaded 100 torrents and seed 50 of those torrents for exactly 144 hours each, the required ratio at the end of the week (before adjustment according to how much you've downloaded) is (50x144/72)/(100x168/72) = 100/233.33 = 0.43, I think...in other words, the raw amount is simply, in this case, 72/168 or 0.43.

Hmm so you must then scale how much you reward that raw amount - according to how much the user has downloaded (the results of which is the table on the ratio rules page). You scale it lower for newbies so that they're rewarded more, but higher for the higher classes until the very top for which they don't earn anything for seeding...and 0.428577551107873 must be the lowest "raw" amount honored by the algorithm before the scaling is applied. Is that all correct? (My terminology is probably a bit un-technical.)

The seed byte idea sounds quite interesting, and I can see what you're trying to do with it. But the goal of that system is different from the goal of the one we've implemented. Our goal is to guide the user towards correct behaviour - downloading, seeding their snatches for at least 72 hours per week, and gradually letting their ratio rise over time. We don't consider someone who's seeding rosetta stone to be contributing any more, or exhibiting any more correct behaviour, than someone who's the sole seeder on a rare 1950's single. However, the seed bytes system would.

As for correct behaviour, I think it is good behaviour to seed the large torrents you've downloaded and not just the small ones you've downloaded . But yes in principle it is no different.

Actually, rewarding even more for a torrent with not many seeders is an idea I've thought about for some time and would be another way to make sure there's good seeding tracker-wide, while still rewarding every user according to how many bytes they are seeding. You could really experiment around with the algorithms in this way in accordance to what you want in your community. Anyway I'd say large torrents need more seeders than small torrents in general and that's why packs/large torrents are freeleech at many trackers.

Well keep up the pioneering work, and I'd like to take the opportunity to thank you for coming up with this great hybrid ratio system which I hope will be also adapted in some form or another at other places where seeding is also hard or at least downloading is to be particularly encouraged.

[kukushka]

The seed byte idea sounds quite interesting, and I can see what you're trying to do with it.

from what i understand, the beauty of the system is that it doesn't add or steal traffic which is the main currency from the system... rosetta stone problem - can't see no big deal about it in currently adopted system, but if it will be modified towards seed bonuses, then maybe it would be wise to use some simplier modifications of what chinese HD trackers are using, basically - no rewards for well seeded torrents (maybe only for 3 oldest seeders or smth). there's a lot of things that could be implemented in making seedboxes less dominant or rewarding uploaders.... don't know if there's enthusiasm to implement new ideas

ps oh, and seeding percentage on current system definitely should consider seeding size cause there's an easy way to leech a lot of ebooks - to seed them forever while not seeding on bigger torrents

[nsk]

Required ratio: 0.00

[7th]

does it means that the tracker has become almost ratioless for those who seed things forever?

I'm not talking about abusive seeding (like limiting things to 1kbps) but instead, seeding with what you have (when you're upload connection is not great).

and I read someone questioning the possibility of being banned in case you stay away for one week for example... this could be easily avoided if the tracker offers the users an option to park accounts for a time.

[ca_aok]

It's dependent on your download amount. Therefore it helps newer members get a base of torrents to seed, while forcing heavy users to maintain ratio. After you download 100GB your required ratio stays at 0.60 regardless.

[7th]

It's dependent on your download amount. Therefore it helps newer members get a base of torrents to seed, while forcing heavy users to maintain ratio. After you download 100GB your required ratio stays at 0.60 regardless.

Not in my entire life I would download 100gb of music
I just like listening to music, I'm not a fanatic