probabilistic approach to tinderboxing

[Bjoern Michaelsen]

On Tue, Jun 12, 2012 at 04:45:06PM +0200, Bjoern Michaelsen wrote:
> After sending off the tinderbox with just this one patch, do:
> 
>  p-innocent(patch)/=p-innocent(testcase)
> 
> for all patches in each tinderbox run with this patch (once for each run)

... and this should probably then be: 
>  p-innocent(patch)/=0.5*(number of patches in patchset-1)/(numbers of patches in patchset)

as we assume p-innocent(dirtypatch) ~ 0.5 and this reverts half of the 'guilt'
from tinderbox runs with both patches in it.

Best,

Bjoern