So you're saying that by November, people will be paying you to take their bitcoins? I think you're smarter than that, so stop repeating that line. "Long term" doesn't mean "1-4 months", which is the only timeframe for which there is a approximately linear fit of $4/month down. It's not true at >4 months: it's actually a flat line for 5 months, and the best linear fit for >6 months is positive slope, not a loss.
Linear drop doesn't even work in the short term of the past 2 weeks, either. Prices have been flat-line in the $4.5 to $6.5 range for two weeks now, and have not dropped $2 more to $2.50 since Sept. 10th.
So... though a linear fit might work for an arbitrary length of time , it's obviously not going to be linear long-term. My vote is for inverse exponential based on the inflation rate until there's some serious additional new interest in Bitcoin. The graph appears to be flattening out at $5-$6. Hmm... perhaps this might also work to bring us from $31 down to $6-ish in 4 months?
http://www.wolframalpha.com/input/?i=graph+31%2Fx^%281.03%29%2C+x+from+1+to+10
Not perfect, but I'll bet you $1000 it models bitcoin value more closely than your function come November.