作者是最近在媒體上很紅的538.com網主Nate Silver. 這位老兄本來是在搞棒球統計，據說創造了一套模組可以 成功的預測棒球選手的表現。今年三月建了538.com，成功 的預測了民主黨初選的結果，因此一炮而紅。 他在民調解讀跟分析上看來很有一套，最近才因為把左格比 民調批的體無完膚而惹的左格比老兄不太高興。他的網站提供 的資訊還有一個特殊的地方，就是他不僅僅是把民調平均起來 呈現結果，他會把不同的民調做不同的加權(近期的民調加權 比較重)，再參考全國的民調趨勢估計出各州兩位主要候選人可 能的得票率，然後再按照程式模擬民調因為誤差或採樣錯誤而 產生的各種可能結果，來算出各候選人在各州的勝選機率如何。 ncyc大認為Silver的計算給馬侃在FA 33%跟NC 41%對歐巴馬的選 情過度樂觀，是否如此我不清楚，事後也恐怕很難證實這估計是不 是過份有利於歐巴馬。馬侃如果明天早上拿下FA或NC， 你也不能說他不準，畢竟33%跟41%還是很有可能發生。 不過目前針對提前投票的選民的民調顯示歐巴馬在這兩州的提前 投票動員有不小的領先優勢，差距大概是15%以上。目前FA已經有 4,272,280位選民投完票，上屆總票數的55.9%(上屆提前投票的比 例36.1%)，NC有2,597,650位選民完成投票，上屆總票數的73.1%( 上屆提前投票的比例30.8%)。馬侃需要在剩下未投票的選民中獲 得大勝才能扳回局勢。 當然你可以爭論提前投票不過是讓本來就會投你的選民早一點投給你 而已，毫無意義。不過對於打組織戰的地面部隊來講，只要確認選民 早一刻把票投進去，他們就可以把這些選民的名字從動員名單上劃掉 ，不用再浪費時間催促他們投票；在你的對手還在花工夫打電話催己 方陣營的忠實選民投票時，你可以把精力集中在其他意向不那麼堅定 的選票上。這或許會在極接近的選戰中造成影響。從這點來看歐巴馬 在這兩州的勝算應該都至少稍大於50%。反駁的理論還是有，不過不論 哪種理論都只是對現象的主觀詮釋，再過十幾個小時我們會知道結果。 http://www.newsweek.com/id/167186 What to Watch For An hour-by-hour guide to election night. In 2000 and 2004, the outcome of the presidential race was unknown into the wee hours of the morning (and indeed for several weeks thereafter in 2000). This time, it is possible that we will be able to guess the winner of the presidential race relatively early in the evening. Regardless, there will be plenty to watch Tuesday night, particularly for those who can appreciate a good slugfest in the Senate. And lest we count John McCain out, we need only remember the polling disasters that befell states like New Hampshire in this year's primaries. Here, then, is what I will be watching each hour on election night. 6 PM EST. Polls close in portions of Indiana and Kentucky. Traditionally, these are the first states to get called by the networks, spotting the Republicans a quick 19 points in the Electoral College. This year, however, is liable to be a little bit different. Indiana is far more competitive than usual, and is probably the state with the greatest disparity in ground games: the Obama campaign has 42 field offices open there, whereas McCain neglected the state entirely until recently. The responsible thing to do would be for the networks to hold off until at least 7 PM to project Indiana, when polls have closed in Gary and the northwestern part of the state just across the border from Chicago—where Obama hopes to rack up huge margins among black and working-class voters. If for some reason the state is called before 7 PM for John McCain, that probably means we're in for a long night. If, on the other hand, the state is called for Obama in the first hour after the polls close, that could indicate that the force of Obama's field operation has been underestimated, and that McCain is in for a catastrophically poor evening. (Speaking of which, Indiana's equivalent on the Senate side of things might in fact be Kentucky, where Mitch McConnell remains the favorite but where he could be vulnerable in the event of an anti-incumbent wave.) 7 PM EST. Polls close in Virginia and Georgia, as well as most of Florida and most of New Hampshire. Virginia, for my money, is the most important state in this election. If John McCain loses it, his path to victory is exceptionally narrow—he would need to pull out an upset in Pennsylvania, while holding on to Florida and Ohio, and avoiding a sweep out West. Barack Obama has considerably more ways to win without Virginia, but a failure to close out the state would suggest at best a more circuitous route to victory. As Obama remains about five points ahead in most polls of Virginia, what we're really looking for is a quick call on anything before 8 PM that would indicate that the map has indeed changed from 2004, and not in McCain's favor. Georgia and New Hampshire are a bit less essential electorally, but they may tell us the most about whether the polls are off in this election. If there's one state where Obama is likely to overperform his polls, it's in Georgia, where 35 percent of early voters are African-American, and where almost 30 percent of them did not vote in 2004. These are the sorts of voters that may erroneously be screened out by "likely voter" models that rely on past voting history. Obama could not only carry the state, but he might help boost Jim Martin to victory in the U.S. Senate race there—giving the Democrats a plausible path to a 60-seat caucus. On the other hand, if there is any state where the polls might overestimate Obama's numbers, it's in New Hampshire, where nearly the entirely electorate is white and where Obama was famously upset by Hillary Clinton during the primaries. If McCain holds Obama to within about five points in New Hampshire —closer than any current polls—we may need to be worried about some sort of Bradley Effect. 7:30 PM EST. Polls close in Ohio and North Carolina. The dynamic to look for in these states involves early voting: more than twice as many people have voted early in North Carolina as did in 2004, and nearly three times as many in Cuyahoga County (Cleveland), Ohio. Recent polling indicates that Obama may have a lead of 20-30 points among early voters in Ohio and a 10-20 point lead in North Carolina. If Republican turnout is at all depressed on Election Day—because of anything from bad weather to low morale—that may be too large a deficit for McCain to make up. 8 PM EST. Polls close in Pennsylvania, Michigan and Missouri. Use caution when interpreting the results from these three states; Missouri in particular is notorious for nearly having been called prematurely both in the 2006 senate race and in this year's Democratic primary. In each state, Barack Obama will rack up huge vote totals in the cities (Philadelphia, Detroit and St. Louis respectively) while trying to hold his own in the rest of the state. If the city numbers come in first, Obama's margins will be exaggerated. If the rural numbers come in first, Obama's prospects will be much better than they appear. But Pennsylvania in particular is the one to watch. If Barack Obama holds onto Pennsylvania—the only state where John McCain seems to have been closing the gap over the last week of the campaign—then winning virtually any red state (Virginia, Florida, Ohio, Colorado, Missouri) would probably clinch the election for him. 9 PM EST. Polls close in Colorado, New Mexico, Wisconsin and Minnesota. Wisconsin and Minnesota should be called fairly quickly for Obama; if they aren't, that's a sign that something has gone truly wrong for the Democratic ticket. New Mexico looks like a safe Obama-state too, but since its vote-counting is notoriously slow, it may take longer to call. The key number to watch in Minnesota should be the difference between Obama's margin of victory and Al Franken's tally in his Senate race against Norm Coleman. If Franken is staying within 5-7 points of Obama as the vote begins to roll in— say, for example, Franken leads by 2 while Obama leads by 7—then Obama's coattails should carry Franken into the Senate. If not, the race may be Norm Coleman's to lose. Colorado, meanwhile, is the last of what I'd characterize as this year's "Big Three" states (the others are Pennsylvania and Virginia). If Pennsylvania and Virginia have split their votes (and Obama hasn't picked up Ohio or Florida), then Obama probably wins if he wins Colorado, and loses if he doesn't. 10 PM EST. Polls close in Nevada, Iowa, Montana and New York. This is the earliest point at which the race might be officially called for Barack Obama—there just aren't enough electoral votes out there, even if he's swept every swing state, to get him to 270 until New York's 31 come in. But assuming that we don't know the outcome of the election by this time, Nevada, where Obama has expanded his lead and where much of the state has already voted, could be Obama's ace in the hole—possibly offsetting a loss in Pennsylvania if paired with other pickups like Colorado and Virginia. The key area to watch in Nevada is Washoe County (Reno), which John Kerry lost by 4 points in 2004 but where the Obama campaign has registered thousands of new voters. If Obama wins Washoe, that means the state—and probably the country— is his. 11 PM EST. Polls close in California, Oregon and Washington. None of these states are in play in the presidential contest this year. The status of the race, however, could have a potential impact on California's Proposition 8, which seeks to strike down same-sex marriage. If Obama appears as though he's headed toward a landslide victory, crestfallen conservatives might not bother heading for the polls to vote for Prop 8. Finally, even if the presidential race has been called by that point, Democrats looking for a little schadenfreude may want to stay up late until the midnight poll close in Alaska, where Ted Stevens is almost certain to be bounced from his Senate seat by Democratic challenger Mark Begich. Should the Democrats pull out an upset in Georgia or Kentucky, it may be Stevens' seat that gets them over the top to a 60-man majority. Nate Silver is the creator of FiveThirtyEight.com, a popular political blog. 本文願意開放自由轉載 -- Obama doesn't need extra votes, especially votes in safe bule states like California or Vermont. Vote Nader, vote your conscience, give those arrogant Democratics who take your votes for granted a warning! --

※ 發信站: 批踢踢實業坊(ptt.cc) ◆ From: 122.127.66.160 ※ 編輯: swallow73 來自: 122.127.66.160 (11/04 20:50) ※ 編輯: swallow73 來自: 122.127.66.160 (11/04 20:58)