Weekdays - does it matter which weekday I buy?

I analysed the daily Dow Jones Industrial for the period of January 1930 - September 2008 whether we could generate any excess profit by only buying at a certain day an selling on the next day. The result may be misleading at the first sight. As we can see in the following barchart, buying on Tuesday's and selling on Wednesday's was by far the most profitable one (1200% profit over the period) while we would have lost when always just holding money over the weekend.

2=buying on Friday's close, seling on Monday's close
3=buying on Monday's close, seling on Tuesday's close
4=buying on Tuesday's close, seling on Wednesday's close
5=buying on Wednesday's close, seling on Thursday's close
6=buying on Thursday's close, seling on Friday's close

Unfortunately this result is misleading. When looking at the chart above we could be tempted to deduce that it would be worth it to buy an equity on Tuesday's evenings and sell it again on Wednesday evening, rather than just hoding it for the entire week. When having a closer look on how robust this strategy is, we can quickly see that this is not a strategy that can be taken seriously.

When we narrow the time period to January 2006 to October 2008 the result looks completely different.

It is important to only look at results that are statistically significant. For instance from the 27 to the 28 October the Dow rose more than 10% (Monday to Tuesday), depicted in Column 3. One single event should certainly not be a determinant of our optimal strategy. That's why I test all the days on statistical significance.

When looking at the period since 2006 and apply a t-test or Wilcoxon Sum Rank test, we don't find any statistical significance in any of the days. When we look at a longer period, the result is too much dependent on what period we look at and the strategy cannot be used to gain an excess return.

When we look at the whole dataset of the dow jones since 1930 the best days were Tuesdays to Wednesdays, But since we were unable to find this effect in the last 3 years we must assume that it is either a spurious correlation or that the effect has disappeared.

Below the outperformance when buying $40 on Tuesdays and selling on Wednesdays versus holding $10 each other day of the week since 1930.

We can conclude that it doesn't matter which day of the week you buy or sell.