Back in 2002, the Web site Real Clear Politics made itself indispensable to politi­cal junkies when it came up with its “poll of polls,” a feature that averaged all the most recent polling of that year’s Senate and gubernatorial campaigns. That way, readers could get a sense of a race at a glance without having to make decisions about whom to trust. “It’s kind of like your wisdom-of-the-crowd notion,” says David Moore, a critic of the industry and a former Gallup pollster. “Sure, you’re going to get a lot of bad ones, but hopefully the bad ones will be just as wrong in one direction as they are in the other and will cancel each other out.”

This year, two sites expanded on R.C.P.’s simple notion. provided aggregate polling data for the most contested states, while, founded by a baseball-stats wunderkind, Nate Silver, employed a sophisticated model that weighted the reliability of different polls as well as factoring in the demographics and voting history of each state. The results were impressive. Silver’s final prediction of 52.3-46.2 Obama over McCain was pretty darn close to 52.8-45.9, the actual final tally. At the state level, fared similarly, with an average margin of error of just 1.6 percent in the heavily polled battleground states.

Has the surprise election result become a quaint historical artifact? In fact, averaging polls only works if the pollsters’ errors really do cancel one another out. This didn’t happen in the New Hampshire Democratic primary, where’s final average had Obama beating Clinton by more than seven points, when he in fact lost to her by two and a half. There’s also the problem that most races don’t produce enough polling to allow the magic of aggregation to do its work. This year was possibly the most polled in human history, but in local races, or most off-year Congressional contests, we’ll still have to settle for the same maddening fluctuations. Which might be good for election addicts who, after all, still need something to obsessively worry about.