2:25am:

With more precincts in, Webb now leads by 7,800 votes. My crude model now predicts an eventual 8,200 Webb lead, which if the last four hours is any indication, is a significant under-estimate.

1:52am:

With 99.6% of the precincts in, the Webb lead is growing even faster than my crude model is predicting. Webb's lead is now 4,700 votes, but my model (see below) is now predicting an eventual 5,700 vote margin for Webb. It will probably be nontrivially more.

1:45am:

With 99.47% of the precincts in, the very rough estimate (see below) is that Webb's lead of about 1500 will grow to about 3400.

1:14am:

With 99.26% of the votes in, Webb leads by less than 2,000 votes. If missing precincts exactly match reporting precincts, then the Webb lead should eventually grow to a rough estimate of 3,850 votes. That should be enough to withstand a recount.

It is interesting how such a simple model could correctly predict that a 13,000 vote Allen lead with over 95% of the vote would eventually lead to a small Webb lead when nearly all the votes came in. It appears that the precincts that were being withheld were even more likely to go toward the county's majority than the earlier reporting precincts.

12:25:

Both MSNBC and FOX are showing that Webb is leading by 2,400 votes with 100% of the precincts reporting.

12:14am:

The four localities with more than 2 precincts outstanding are ALL in localities that are voting Democratic: Fairfax county, Loudoun county, Prince William county, and Richmond City. When 100% of precincts are in, it should be much closer or Webb should take the lead.

11:45pm:

Even though Allen still leads by a few thousand votes with 97.7% of the vote in, the trend in Webb's direction continues. If the missing precincts match the reported precincts in each VA county, Webb would win by 399 votes.

11:27pm: Welcome Instapundit readers. You will note that I am not predicting that Webb will win by 550 votes. Obviously, the race is too close to call. What I have been doing is computing the final results IF the missing precincts vote exactly like the nonmissing precincts in the rest of that county.

10:54pm:

With 95.58% of the vote in, Allen leads by 13,000 votes. If every missing precinct is exactly like the other precincts in that county, then my computations now predict a Webb victory of 550 votes.

A recount is almost certain.

10:40pm:

I ran the estimates (using the assumptions earlier in this post) with almost 94% of the vote in and came up with an Allen lead of 136 votes. Five of the last 6 Arlington precincts just came in (I haven't analyzed them yet). The last 3 Fairfax precincts have been still out, with nothing coming from that county in well over an hour.

10:30pm:

With 93.21% in, Allen leads by 18,000 votes (0.86%).

10:20pm:

Most of the Arlington County votes came in about an hour ago (at about 9:17), but not a single Arlington precinct has come in for the last hour. From my experience in watching Chicago election returns, very long gaps in reporting often lead to huge surges for the more popular candidate in that county.

I have been running county by county numbers in Virginia, on the assumption that the precincts not reporting are of the same size and vote the same way as the rest of that county's precincts vote. When about 85% of the votes were in, this method projected a Webb win by a few hundred votes, while with about 90% of the votes in, this method projects an Allen win by a few hundred votes.

Unfortunately, that means that the election may be close enough that voting irregularities or litigation might determine the outcome.