I think there is a weirdly linear (in a log-log sense) relationship between the visitor count (number of monthly visitors) and the global Alexa rank of a website.
This is called a power law.
It is easy enough to remember - a little like Moore's law, or Ohm's law :)
|Relationship between Alexa rank and monthly visitors|
To halve your ranking, double the visitor count.
If you dropped to 1/3 of your usual site visitors, your new ranking will be old ranking x three.
And so on.
Here's the data I used:
|Site||%Traffic Reach||Est. Monthly Visitors*||Alexa rank|
*The monthly visitors numbers are estimated by assuming jobberman has 1million visitors per month (I think that corresponds to a ranking around 5k) and that the traffic reach percentage (data given by alexa for each site) is simply number of visitors for this site / a fixed number corresponding to all traffic. That is, take the traffic to be proportional to the traffic reach percentage.