New Senate polls show movement to the Republican in Colorado. Nate Silver releases his latest Pollster ratings. And the Ace of Spades HQ aims to change election night vote count reporting. This is HuffPollster for Friday, September 26, 2014.
GOP GAINING IN COLORADO - A new poll from the Democratic firm Public Policy Polling (PPP) conducted for the Americans for Tax Fairness Action Fund found Cory Gardner (R) up 2 points against Sen. Mark Udall (D), the fourth survey in a row to give him at least a nominal lead. Our poll tracking model now shows the always-tight race narrowing to a virtual tie, with about an equal likelihood of either candidate winning, largely because it still gives weight to polling data released earlier in September. [PPP, Colorado chart]
Noise, or a trend? Nate Silver: "All five [Colorado] polls between Aug. 26 and Sept. 11 had [Udall] ahead. All five since have shown him behind. Could this just be a coincidence? Absolutely. If you’re tracking dozens of races for months at a time, you’re going to find a few weird patterns like these — and sometimes they’re just statistical noise...In this case, however, there’s a credible hypothesis to explain the trend toward Gardner. Whereas a few weeks ago, Udall had a heavy advertising advantage in Colorado, more recent ad placements have been almost even, according to data from Echelon Insights, a Republican analytics and consulting firm. Advertising blasts can sometimes produce temporary bounces in the polls; perhaps Udall had one a few weeks ago.
Other polling from yesterday:
Alaska Senate - A Rasmussen survey gave Dan Sullivan (R) a 5-point lead over Sen. Mark Begich, while a Dittman Research (R) poll for the Chamber of Commerce put Sullivan up 6. Earlier in the week, a PPP (D) survey also found Sullivan taking the lead. Begich hasn't been ahead in a nonpartisan poll since July, when a CBS/New York Times/YouGov gave him a significant lead. (A third poll, not yet in our polling model, from Anchorage pollster Marc Hellenthal also has Sullivan up 5.) The Pollster model continues to show a close race, with Sullivan ahead by just over 1 point, and barely more than a 50 percent chance of winning. [Alaska chart, Rasmussen, Dittman, Hellenthal]
Pennsylvania governor - Two polls from Franklin and Marshall College and Magellan (R) fall in line with nearly all the other surveys of the race, showing incumbent Gov. Tom Corbett (R) likely to lose his seat to challenger Tom Wolf (D). Franklin and Marshall gave Wolf a 21-points lead, while the Magellan survey, by far the best for Corbett since January, still showed him down by 9 points. The Pollster model predicts a margin closer to 20 points, and gives Wolf a 98.9 percent chance of winning. [Pennsylvania chart, Franklin and Marshall, Magellan]
NEW FIVETHIRTYEIGHT POLLSTER RATINGS - Nate Silver: "Pollster ratings were one of the founding features of FiveThirtyEight. I was rating pollsters before I was building election models. I was eagerly updating the ratings after every major batch of election results. I rated pollsters while walking two miles uphill … barefoot … in the snow. And then I got a little burned out on them. We last issued a major set of pollster ratings in June 2010 and made only a cursory update before the 2012 elections. What happened? Well, when you publish a set of pollster ratings, people are understandably fixated upon how you’ve rated the individual polling firms: Is Pollster XYZ better than Pollster PDQ?... However, discussions about individual polling firms — there are now more than 300 of them in our database — can sometimes miss the point. I’m more interested in the big-picture questions. Are some pollsters consistently better than others, as measured by how accurately they predict election results? In other words, is pollster performance predictable? And if so, are a pollster’s past results the better predictor — or are its methodological standards more telling? The short answer is that pollster performance is predictable — to some extent. Polling data is noisy and bad pollsters can get lucky. But pollster performance is predictable on the scale of something like the batting averages of Major League Baseball players....the differences in poll accuracy aren’t that large. We estimate that the very best pollsters might be about 1 percentage point more accurate than the average pollster over the long run." 
Improvements - Silver's latest iteration of his ratings addresses some of the complaints HuffPollster had about his last major effort to rate pollster accuracy four years ago: The ratings are, in effect, a meld of both accuracy scores and bonus points awarded for methodology (or inferences about methodology), but his 5,800 word discussion of the details now includes a set of tables based on accuracy ratings alone for the most prolific pollsters. He has also eliminated the penalty points given to internet research firms, and dropped usage of the somewhat misleading term "pollster induced error" for his overall score. [Pollster.com]
What it means to get a 'B' or 'C' - The FiveThirtyEight ratings now emphasize a letter grade rather than the underlying overall score (now dubbed "Predictive Plus-Minus"). Yet many firms get ratings in the B to C range less because their performance is poor than because it is largely indecipherable due to sparse polls and a lot of statistical noise. Silver: "Our finding is that past performance reflects more noise than signal until you have about 30 polls to evaluate, so you should probably go with the firm with the higher methodological standards up to that point...As a final step, we’ve translated each firm’s Predictive Plus-Minus rating into a letter grade, from A+ to F. One purpose of this is to make clear that the vast majority of polling firms cluster somewhere in the middle of the spectrum; about 84 percent of polling firms receive grades in the B or C range. 
Database now open-source Silver: " I’d encourage you to download the database of polls that we’ve used to construct the pollster ratings. We’re making it public for the first time. The database includes (with just a few minor exceptions that I’ll describe below) every poll conducted in the last three weeks of a presidential, U.S. Senate, U.S. House or gubernatorial campaign since 1998, along with polls in the final three weeks of presidential primaries and caucuses since 2000." [538 database]
TRENDE'S THEORY: SENATE RACES WILL BREAK TO GOP - Sean Trende: "A recent piece from Dan McLaughlin at RedState was largely overlooked, but it is hugely important. McLaughlin went through the polls from RCP and constructed averages for mid- to late-September going back to 2002....McLaughlin notes, correctly, that in these years, the races broke almost uniformly toward the “wave” party....Taken together, this tells a story that sticks together reasonably well. From 2002 to 2012, candidates of the president’s party have tended to converge on the president’s job approval. It isn’t an absolute tendency, but it is nevertheless real. The Democrats’ problem is that they seemingly find themselves in a position similar to that of Republicans in 2006: They are in tight races. But so far, they seem unable to move past where the fundamentals suggest they should be able to go: Recall again that their maximum showing has generally been bounded at 47 percent....If this theory is right, we should expect to see these races continue on the basic trajectory we’ve seen over the past few weeks: Democrats holding at their current levels. Eventually, Republicans should begin or continue to improve, as undecided voters engage and make up their minds, and as Republicans narrow the spending battles. Even if this theory is true, it won’t occur in every race, but it will be the general tendency." [RCP]
PROFILING THE 'ACE OF SPADES HQ DECISION DESK' - Ben Smith: "The Decision Desk swings into action on primary election nights, with Brandon Finnigan, a burly 29-year-old, nailed to a busted black leather armchair at one of those cheap beige aluminum desks, halfhearted fake wood top, toward the back of a low-slung house at a truck stop crossroads deep in California’s monotonous Inland Empire….Between calls from drivers broken down in San Diego, and between visits from men there to drop off paperwork, he tracks election results for Ace of Spades HQ, a conservative blog run by an anonymous, combative figure known only as Ace….'I want to fundamentally change how results are reported,' Finnigan told BuzzFeed News. His goal is both to modernize the local election boards and to deflate what he sees as false drama imposed by slow Associated Press calls and desperate television commentators. 'I understand you need an element of suspense and you need something to jibber-jabber about on election night. But you got to jibber-jabber all year. I just want the results.'" [Buzzfeed]
HUFFPOLLSTER VIA EMAIL! - You can receive this daily update every weekday morning via email! Just click here, enter your email address, and and click "sign up." That's all there is to it (and you can unsubscribe anytime).
FRIDAY'S 'OUTLIERS' - Links to the best of news at the intersection of polling, politics and political data:
-Gallup finds American support for daily prayer in schools has dipped slightly. [Gallup]
-A new survey looks at gay and bisexual men's opinions and behaviors regarding HIV/AIDS. [Kaiser Family Foundation]
-Recent polls find Obama's job approval slipping in two deeply blue states. [The Hill]
-Bob Erikson and Chris Wlezien argue that likely voter screens in September can be misleading (and correct an initial misreading of Pew Research data). [WashPost]
-Nate Cohn assesses Kay Hagan's "surprising strength" in North Carolina. [NYTimes]
-Harry Enten contends that whoever is nominated to replace Eric Holder will likely face a tough confirmation hearing. 
-Henry Farrell argues if policymakers had listened to political scientists, the U.S. never would have invaded Iraq. [WashPost]
-Justin Wolfers looks at what the betting markets are saying about a Mitt Romney candidacy in 2016. [NYT]