Average P/E without filters doesn’t make sense because P/E is highly nonlinear. You’re adding apples and oranges. Here’s an example: You have one company in your 2000-stock index with near-0 earnings for the year. If it’s $0.00001, then P/E will be effectively positive infinity, which will dominate your average calculation, which will be some very big positive number (positive infinity divided by the number of stocks in the index).
On the other hand, if it was $0.00001 losses, it’d be effectively negative infinity. The difference of a fraction of a penny around zero swings your “average P/E” from a very big positive number to a very big negative number.
Thus, average P/E without filters is essentially noise: the more near-zero-earnings companies you try to average P/E for, the less actual information the result carries.