Kind of interesting this one. I was looking through some of the logs for my home-brewed spam system on my home hetwork. Not perfect, but a lot cheaper than commercial products - even as a Microsoft employee, I would still need to buy this stuff otherwise, and as an ex-Exchange dev, it's not that hard (well it is to be honest, but that's definitely for another day)! [There's a hint if there's any commercial vendors reading this ;-)] Over four months, I've averaged around a 48% spam detection with pretty good accuracy, so I'm pleased with how it's performing by and large.
Now here's the interesting bit. I ran off a report which shows the percentage of spam detected per day. Here's what I discovered based on the lightest days for spam for each month:
August: SundaysJuly: FridaysJune: SaturdaysMay: Thursdays
Of course, there's statistical error likely in all these results as I don't receive a huge amount of email compared to Microsoft for example (something like 10,000 emails in total over the 4 months). However, these results were obvious even without deep statistical analysis. I'd be interested if someone has an explanation!