Which makes it even more likely that time is actually moving faster. Y'see, in 2000, it felt like April was taking forEVER, but on November 3, 2010, I almost put the date in my The Notebook for English class as SEPTEMBER, and of course, now the last 24 days before Christmas felt like only 7 days.

Let x be the year + the month / 12 (for instance, July 1993 is 1993 7/12) and y(x) be the ratio of how much time seems to have passed at a given point in the month x to how much time has really passed (for instance, if time appears to be moving at twice its normal speed, then y = ½).

Then:

y(April 2000) or y(2000⅓) > 1 (let's just say y(2000⅓) = 31/30, even though I doubt that April 2000 seemed only one day longer to me than it really was)

y(September 2010) or y(2010¾) is between 0 and 29/63. Let k = y(2010¾).

y(December 2010) or y(2011) = 7/24

If k = 19/63 (the lowest multiple of 1/63 that is greater than 7/24), linear regression for these points creates a function with a zero between January and February 2015, implying that time will reverse there. Increasing k to 29/63 moves the zero of the function to between October and November 2016, so time has to reverse somewhere between January 2015 and November 2016.

But wait - since y(2000⅓) is probably more than 31/30, the value of y is probably decreasing faster than the linear regression equations I found predict, so time will reverse even sooner than that! :O