From SCVTalk:https://www.santaclarita.com/blog/view.php?blog_entry_id=31085On Tuesday, the City ended their election day count of ballots with just a 46-vote spread in the race for the 3rd council seat between Dante Acosta and Alan Ferdman. On the 15th, they'll finalize the validation and counting of as many as 965 ballots, which may contain as many as 2,895 votes. Following a night of live blogging the election night vote count from council chambers, I wrote the following from a bench outside city hall:11:10PM: FInal thought for the moment: this final count is typically made up of VBM voters who turn in their ballot on election day, more so than provisional poll voters. Not always though. By looking at the results in the VBM versus poll voters tonight, it appears that Dante Acosta does a little better with these voters than Alan Ferdman does. There will be plenty of time to figure out a more sophisticated metric for forecasting, but my hunch right now is a 70% chance Acosta holds on.The next day, I took a closer look at the numbers and I put Ferdman's chances of gaining the 46 votes necessary at almost zero. Maybe he'd gain 8 votes, 12 if he's lucky. My assumption was based good conclusion drawn from a very different election, so my premise is faulty. Late count voters don't necessarily resemble VBM more than poll voters. They don't necessarily resemble either. Ferdman is still facing a longshot. But I'm going to change my clichéd summation from "he's out of it" to "anything can happen." Here's why.The last time we were in a protracted waiting period for election results was four years ago. David Gauny had gained an incredible number of votes on election day. Many saw a late-breaking Gauny trend, certain to continue with the late voters, carrying him onto victory over Frank Ferry. I saw the disparity as something else: an unsophisticated campaign that got very active in the final weeks and probably made the calculated choice to focus on poll voters. Discussions with people close to the campaign later verified this hunch and the results suggested the same: Frank Ferry actually gained votes during the lingering count of late VBMs and provisional voters.2010 gave us a nice, clean example of how the late count can be counterintuitive yet easily explained. But that was due to a peculiar circumstance. Gauny both outperformed (he surprised everyone, nearly winning 3rd place when most figured him to be a distant 5th) and underperformed (if his campaign had mobilized his volunteers earlier, he would have won, for sure). 2012 was a different story, and the only lesson I can draw from it is that the late voters can be a different animal entirely. If we expected the late voters to act like the VBM voters, we would have given Laurie Ender 33 too many votes and shorted Kellar 54 and Boydston 81. If we used the poll voting model we would have given too many votes to Ender (21) and Kellar (6), while Boydston still gets shorted by 43 votes. In other words, a candidate can outperform both their poll and VBM performance in these late counts. Absent any other information, it's really hard to tell if Ferdman will be so lucky.The difference between poll voters and VBM voters is striking. If poll voters elected the City Council, Alan Ferdman would have come in first. Marsha McLean would have come in fifth. But Ferdman isn't like Gauny gaining on Ferry or Boydston breaking away from Ender. He's chasing another challenger, and Dante Acosta came in second with poll voters.Acosta's strength, like Kellar's in 2012, is a balanced appeal. He did pretty well with VBM and 
oll voters. He did just fine in Valencia or Saugus or Newhall or Canyon Country. He's not going to trip. Ferdman, on the other hand, in the most unpredictable candidate. No one had a bigger spread in their VBM share vs. poll vote share. His results from precinct to precinct are all over the place, too. He might be the unofficial mayor of Canyon Country, but he's just another name on the ballot to a disconnected early VBM voter in Valencia. I went through and compared each candidate's performance in each precinct against their overall performance. Ferdman's performance ranged from a 6.6% 7th place finish in northern Valencia to a commanding 21.2% in Placerita Canyon. The standard deviation of his precinct performance was 3.4% contrasted with Acosta's 1.6%. I added that standard deviation to the poll/vbm spread to give each candidate a pseudo-mathematical "Volatility Score."I'm no statistician, but this gives us some idea of how hot (and cold) a candidate could get if the late ballots show a bias towards certain areas or types of voter. Ferdman's strengths are so strong that even if both Acosta and Ferdman got their favorite types of voters in these last 965, Ferdman would gain enough votes to win.It still seems unlikely. We're looking at a small portion of the votes left to spread across 13 candidates. But voters in this late count are unusual by definition. It's more difficult for a provisional voter to vote than other poll voters. Walking in a ballot takes more effort than sticking it in the mail. If one of these campaigns had an 11th-hour push to make sure their supporters voted one way or another, it might be enough to make a significant difference in the outcome.