[tor-talk] Tentative results of analysis of data on metrics.torproject.org

[Karsten Loesing]

On 06/09/14 00:51, Virgil Griffith wrote:
> Can someone suggest to me how to calculate the probability that a node will
> be selected for a circuit?  Failing that, pointing me to the spec for the
> selection algorithm might be enough.

Onionoo uses consensus weight fraction as rough approximation to being
selected for a circuit.  I'm not sure what your use case is, but this
might work for you, too.

Onionoo further calculates guard/middle/exit probabilities by weighting
the consensus weight with Wxx numbers contained in the consensus.
That's probably more accurate if you're interested in the probability
for those three positions, but it's still an approximation.

Here's the source code, but beware it's Java!

https://gitweb.torproject.org/onionoo.git/blob/HEAD:/src/main/java/org/torproject/onionoo/updater/WeightsStatusUpdater.java#l104

All the best,
Karsten