From
http://forums.steampowered.com/forums/showthread.php?t=876797
Okay we know the system is supposed to be random, and seems to be, however there are increased reports of the "idling" becoming less and less productive.
So here is my two cents:
(i do not count if i get a new or old gun, a drop is a drop)
Right after the update i racked a "decent but low" number of unlocks and clocked like 20 hours of tf2 in the first two days of the release.
After this i idled a while during the night and got in 7 hours ... 1 item.
At some point i decided to delete all and try to get them "honestly"
from then i got so far 6 of the guns, only 2 dupes.
So here is my theory, the system tries not to advantage players who play more from players that play less, because of this it evaluate what is your daily TF2 consumption and this determinate how often per hour you can "roll the dices".
If at one point you have a brutal increase of your daily playtime you will see an increase in drops, but this sharp increase will skew your average playtime per day and result in a slowdown of your "rolling of the dices".
So the more you abuse the system the slower it gets.
Wich basically makes farming useless because what drop in 15hours of idle farming +5hours of play could have dropped in the 5 hours of play alone.
There is also a high probability of a "cutoff value", namely a maximum drop rate with is probably around the drop speed Valve wishes for the average play time of an average player.
The first days of the update most of the big players clocked much higher than usual playtime and it appeared to reward them more, which is probably true on the short term as any system that determinate new game sessions based on statistics of past game session is bound to have some "lag" before it rectify itself.