So the short version is that if the things we want to see passed are passing tomorrow (74, 502, Inslee) then the historical and mathematical trends are that unless there is an act of God, they'll finish passing.

@1: Basically, yes.

@2 this is why God hates statistics, Nate Silver, and gay marijuana.

With all due respect to Barreto & crew ...

1) Every single county will provide an estimate of how many votes they have remaining to count, but these estimates are NOT notably accurate on election night - and in some cases are out of whack right up to the last votes are counted;
2) Historically, election data has shown that the Yes/No or Dem/Rep vote within each county ARE NOT very stable from election night to the final tally;
3) a disproportionate share of [ballots not yet counted] come from King County, but this is NOT a dominant share of such ballots;
4) Given the estimates of 81% turnout by Secretary of State Sam Reed, we are factoring another unreliable number into the model.

Cranking and crunching these numbers myself in close races, (countywide and statewide), I have seen a number of cases where the early returns so treated - and the late returns as well - produced wobbly projections.

Still, I'm glad I can shun requests for such analysis, and just tell anxious cliffhangers "go look here" instead.

So that is why so many people are voting for free at the Red Square ballot location for King County Elections.

Either that or the HotDawgs ...

This mostly works ok unless there are late developments in a race that would make later ballots skewed one direction or another (ex: Sanders' state supreme court loss). But for very, very close races, the number of outstanding ballots matters: counties only report a) ballots on hand (lots more arrive after election day) and b) their estimate of what the total turnout will be (which usually doesn't have better than +/- 2% accuracy). It will be nice to have the projections available, but they won't be any more accurate than the statistics feeding the model. And not (much) more accurate than the past mathsy slog commenters' projections using the same limited information.

@5:

I'll go farther. In recent elections, late-arriving ballots have been more Democratic than the earlier ballots statewide.

I tracked results for a few days in 2010 (thankfully, I didn't have to repeat the epic series of reports I wrote in 2004). There was a very consistent increase in Patty Murray's percentages, and it was by no means limited to King County. Here's something I wrote on Friday, Nov 5, 2010:

Based on the results reported on the Washington Secretary of State's website through Wednesday, and extrapolating each county's current Murray and Rossi percentages to include the county's waiting-for-processing ballots, my projection for the final result has risen to 52.0% to 48.0%. Compare that to yesterday's projection of 50.9%-49.1% (narrower than today's actual margin) and the Tuesday projection of 50.3%-49.7%. Yes, these have been two excellent days for Patty Murray!

The final result, BTW, was even better than my projection in that report (52.4% to 47.6%).

Patty's King County percentage through Wed had been 62.6%, but she took in 68.3% on Thu. Similar outcomes were seen in Snohomish (51.5% Tue-Wed to 52.8% Thu), Whatcom (51.0% to 54.9%), and even Klickitat (41.4% Tue-Wed, but 52.9% on Thu). All in all, Murray's percentage on Thu topped her previous county-specific percentage in 16 of the 22 counties reporting that day.

It wasn't just Murray either. Rick Larsen trailed John Koster Tuesday night (49.6% to 50.4%); Stephen Colbert even included Larsen in a piece about defeated Reps whose districts he'd "better known". But Larsen got 53.2% of the Wed-Thu ballots. And only one or two King County precincts were in the old WA-02.

Some LD races also flipped (always R to D) between Tue night and the final result -- Luis Moscoso (1st), Roger Goodman (45th), Rodney Tom (48th). King County, to be sure, though most of LD-01 was in Snohomish.

Bottom line -- Barreto is soft-pedaling the effect. I suspect he's also taking the trends I've outlined into account, even if he doesn't say so.

N @ 8 -- Yea, even without "late developments", different tranches of ballots from the same county act like distinguishably distinct populations. I've seen the trend change direction and change back again. Adding to the complexity, not everybody processes ballots first-in-first-out every cycle.

I'll follow Barreto's results with interest, and I'm curious how much "secret sauce" (if any) he uses in his mix.

So, I'm still looking for the results of these projections by Matt Barreto. It's about 10:20, and I can't find any evidence of them on the google. Do you know where they are?