Gaming the System?

[radium]

Recently I've made available a series of sports videos via bit torrent. Initially, seeding went very well; multiply seeders appeared with share ratios slightly greater than 1. Lately I've noticed however, that when certain peers, recognizable from their IP addresses, join the swarm, one-half to two-thirds of all the file pieces I'm seeding are immediately directed towards them. Apparently they don't share, because the time it takes to fully seed a torrent is extended proportionally.

So far uTorrent reports that these "special" peers are all using uTorrent 1.7.4, the same program and version I'm using. Typically when seeding, there are 40 to 60 peers from 5 to 10 countries, using different programs and versions: uTorrent, FlashGet, BitComet, Transmission, Azureus, KTorrent, BitComet, etc.

I tried seeding with the Initial Seeding option selected and not selected: no difference. I finally banned the special peer IP addresses; that restored "democracy" and reduced seeding times.

Though this issue has been resolved, I'm curious: what was different about these IP addresses that enabled them to preferentially suck up file pieces that I was seeding? Could it be that they have something like a super fast Internet connection, or have they figured out a way to game bit torrent?

[Switeck]

I've heard of client id spoofing for a LONG time, µTorrent even identifies some clients as fakes.

And I've heard about peers gaming the system too, though I cannot say I've seen it myself before...or even if there's much truth to it being a possible cheating client.

But what seems strange is µTorrent didn't recognize what this new client is a fake of...assuminig of course it's not something new.

This is "wonderfully" bad news if it's a new client that's really good at gaming the system AND stealing µTorrent's client id to do it!

With lots of people on ISPs that are becoming even more hostile towards BitTorrent, many are turning to whatever speeds up their download even if it's not fair to others on BitTorrent.

We need more information to confirm/disprove anything about this though.

[radium]

I've handed off my seeding responsibilities to a volunteer who has far more upload bandwidth than I. However, thinking back, I've seen this before and will doubtless see it again; next time I'll get some screen shots of the peer list. It's pretty obvious when you see it: a peer joins the swarm when the average completion is 30% to 40%, over time this new peer's completion is 80% to 90% while the rest of the swarm has only gained marginally.

[Switeck]

I've heard BitTyrant works really good at "boosting" its download speed. But it does so by uploading "just enough" to everyone to try to maximize its download speed -- NOT by being a total leech.

A total leech should only get download speed from seeds (who can't directly check to see if peers are sharing) or peers via optimistic unchoke.

I've seen a handful of posts here about possible/probable cheating clients...and for the longest time thought BitComet by default was one of them. But as it turns out, it too was being "gamed"/cheated by the true cheating clients if posts of that nature are anything to go by.

...I'll probably have to research some more BitTorrent cheat sites to find more on this, many of which aren't indexed by Google. (...though that's not terribly a bad thing for us.)