I've just run the following test and come up with the following results: The average gain per week in the DJIA from 1986 to 2005 is 0.205% but if you only look at weeks with expiry Friday's then that gain is 0.326% and if you exclude weeks with expiry Fridays then the gain is 0.195%. (this was posted during a week which had an expiry Friday)

This week is supposedly more bullish for equities because it is expiry week. Here is a spreadsheet that I did testing this theory: ExpiryFridayWeeksTest.zip

Very interesting indeed. I also suspect that weeks prior to the expiry week have a higher probability of being bullish. Have you already tested that?

No I haven't tested that. I assume that you would be looking for a gradient of improving bullish probability as you got closer to the expiry week and then for it to drop off the week after right? Would that imply that the week after expiry is generally bearish?

That's right, the bullish tendency should start already a few weeks before the expiry date. I'm not sure about the week after expiry, though. Apart from that, I wonder what the mechanism behind this thing is. Understanding that would make predictions more robust against factors like long term bull/bear trends or seasonal patterns.

How were the percentages calculated? I usually calculate a percentage change as (P(t) - P(t-1))/P(t-1). For example, the percentage change in cell G3 should be equal to =(E3-E2)/E2 = 0.827% instead of 0.635%

I would usually calculate them the way that you have described but I specifically wanted to try and leave out the weekend effect (if any) so I used the Opening price of the week as P(t-1). Otherwise used exactly the same formula that you would use. So G3 would have been =(E3-B3)/B3 before I converted it to values.