Most of the more than 4,000 exoplanets astronomers have found across the past few decades come from NASA’s pioneering Kepler mission, which launched in 2009 and ended in late October 2018. But among Kepler’s cavalcade of data, more planets are still waiting to be found—and a new method just turned up the biggest haul yet from the mission’s second, concluding phase, called K2.

The K2 run from 2014 to 2018 was notable for its unique use of the functionality, or lack thereof, of the Kepler space telescope. Essentially a large tube with a single camera, Kepler relied on four reaction wheels (spinning wheels to orient the spacecraft) to point at specific patches of the sky for days or even weeks on end. Such long stares were beneficial for its primary planet-finding technique, known as the transit method, which detects worlds by watching for dips in a star’s light caused by an orbiting planet’s passage in front of it. But when two of Kepler’s reaction wheels failed, one in 2012 and another in 2013, mission planners came up with an ingenious method of using the pressure of the solar wind to act as a makeshift third wheel, allowing observations to continue, albeit with some limitations.

“We had this issue because the K2 mission was working off of two reaction wheels; it rolled a little bit every six hours,” says Susan Mullally of the Space Telescope Science Institute. “And as a result, the light curves have these little arcs that run through them that you have to first remove.”

Various efforts were subsequently made to extract planets from the K2 data. But none have been more successful than one reported in a new paper by Ethan Kruse of NASA’s Goddard Space Flight Center and his colleagues, which was posted on the preprint server last week and accepted for publication in the Astrophysical Journal Supplement Series. Kruse employed an algorithm known as as QATS (for Quasiperiodic Automated Transit Search) and a light-curve-analysis program called EVEREST (for EPIC Variability Extraction and Removal for Exoplanet Science Targets) to better account for the spacecraft’s rolling and other sources of instrumental and astrophysical “noise” in the K2 data. The result was a whopping total of 818 planet candidates—374 of which had never been spotted before—from the first nine of K2’s 20 observation campaigns.

“We were trying to find planets that we knew were missing in other searches,” says Kruse, who spent more than three years on the project as part of his dissertation at the University of Washington. “The main result was that it worked.”

Kruse and his colleagues’ sizable haul includes worlds ranging from hot Jupiters—gas giants that orbit their star incredibly closely—to super-Earths, planets midway in size between our own and Neptune that are very abundant in our galaxy but, oddly, seemingly absent from our solar system. The team also found 87 multiplanet systems, including two new five-planet systems and one new six-planet system, and managed to use transit timing variations—the wobble in a planet’s transit caused by the presence of another world—to spot a previously discovered sub-Neptune. “This is the largest haul for K2 to date from one paper,” says Mullally, who was not involved with the work.

Kruse notes his team’s search was able to find nearly 90 percent of the planets spotted in all previous searches of K2’s first nine observation campaigns, whereas only about half of the planets in his paper had been seen before. Most of the newfound worlds are candidate planets, which means they need follow-up observations to be confirmed. But if they are validated as genuine, the haul of 374 would increase the total number of planets found in the K2 data by about 50 percent—a remarkable result for a single paper. “I think this is really a gold standard in how you want to do astronomical data analysis,” says Benjamin Pope of New York University, who did not contribute to Kruse’s study.

Of the planet candidates found by Kruse and his team, 154 belong to an intriguing group known as reciprocally transiting planets. These are worlds that orbit in the right plane for us to observe their transit around their star, but observers on the planets could also spot Earth orbiting our sun using the same method. “They see us transit, and we see them transit,” says David Kipping of Columbia University, who was not a part of the new paper. Those worlds, Kipping notes, could be prime targets to probe for “technosignatures” of other civilizations. “Inhabitants of such systems have a natural temporal window to attempt communication with us during their times of transit,” he says.

Kruse’s analysis also reinforced some curious trends seen in previous Kepler data. In particular, many planets appear to orbit their stars in a 3:2 resonance with neighboring worlds, meaning one completes three orbits for another’s two—something we see in our own solar system with the moons of Jupiter. “When the periods are related like this, it tells us clues about how they formed,” says Jessie Christiansen of the California Institute of Technology, noting the ratio could hint at regular planet migration in the early stages of planetary systems. “The fact we’re seeing this peak in resonances again and again, to me, is really interesting. It’s not just some fluke of the Kepler data.”

With the Kepler mission over, NASA’s exoplanet efforts have now shifted to a new telescope, the Transiting Exoplanet Survey Satellite (TESS), which was launched in 2018. TESS has already turned up close to 1,000 new candidate planets, many of which were announced this week at the TESS Science Conference at the Massachusetts Institute of Technology, and methods like Kruse’s could help find even more in the future. “You can port all this code directly to TESS with almost no changes,” Pope says. Kruse also plans to apply his technique to the remaining K2 data at some point in the future.

Kruse’s original goal for his paper was to design a single search that could find every transiting planet in a set of data. Although that objective was not quite achieved (Kruse’s analysis probably missed a small number of stragglers still lurking in the first batch of K2 data), Mullally notes that astronomers are eager to find ways to locate every possible planet from any given data set. “There’s no ‘one search to rule them all’ as yet,” she says. “But some of those seem to be performing better than others, and we’re still trying to figure out quite what that parameter space is.” And, for the moment, Kruse’s effort is among the best of the best.