It's ratios junky.
Here is the prediction from the year 2000 scenario.

The total run size in year 2000 was 464,100.
The total count on July 8, 2000 was 308,415.
The total count on July 8, 2004 was 214,217.

To Calculate:
464,000 ÷ 308,415 = 1.50479 (the ratio of run size:count)
1.50479 x 214,217 = 322,352 (the 2004 run size at that ratio)

Another way:
214,217 ÷ 308,415 = .697504 (the ratio of 2004:2000 counts)
.697504 x 464,100 = 322,352 (the 2004 run size at that ratio)

If you knew the total run sizes from the other years you could duplicate the calculations for those years.

Finegrain had the right idea but was using the total count rather than the total run size.

Predictions by repeating the individual years scenarios are:
2000 - 322,352
2001 - 327,171
2002 - 668,338
2003 - 363,837

Obviously the year 2002 was a peculiarity.

By incorporating other trends and factors into the methodology my prediction formula indicates a final count of between 310,000 and 365,000 and a total run size of somewhere between 347,000 and 416,000 with the best guess at around 372,000.

If the counts keep coming in over 12,000 through the weekend the trend will push the estimate further towards the 416,000 level.