It’s one of the most perplexing problems in modern astronomy: Based on multiple observations of stars and galaxies, the universe appears to be collapsing faster than our best models of the universe predict. For years, evidence of this conundrum has been accumulating, prompting some researchers to call it a looming cosmological crisis.

Now, using the Hubble Space Telescope, a group of researchers has compiled a massive new dataset, and they’ve discovered a million-to-one chance that the discrepancy is a statistical fluke. In other words, it’s becoming increasingly likely that there’s some fundamental component of the universe—or some unexpected effect of the known components—that, astronomers have yet to discover.

“The universe seems to throw a lot of surprises at us, which is good because it helps us learn,” says Adam Riess, an astronomer at Johns Hopkins University who led the most recent effort to test the anomaly. The problem is known as the Hubble tension, after the astronomer Edwin Hubble. In 1929, he discovered that the farther a galaxy is from us, the faster it recedes—an observation that paved the way for our current understanding of the universe, which began with the big bang and has been expanding ever since.

Researchers have attempted to calculate the universe’s current rate of expansion in two ways: by measuring distances to nearby stars and mapping a faint glow dating back to the early universe. These two approaches allow us to put our understanding of the universe to the test over more than 13 billion years of cosmic history. The study has also uncovered some key cosmic ingredients, such as “dark energy,” the enigmatic force thought to be driving the universe’s accelerating expansion.

However, these two methods disagree by about 8% on the universe’s current expansion rate. That difference may not seem significant, but if it is real, it means the universe is now expanding faster than even dark energy can explain, implying a flaw in our understanding of the universe.

The findings, which were published last week in The Astrophysical Journal, use specific types of stars and stellar explosions to calculate the distance between us and nearby galaxies. The dataset contains observations of 42 different stellar explosions, which is more than double the size of the next-largest analysis of its kind. According to the team’s findings, the gap between their new analysis and results from early cosmos measurements has reached five sigma, the statistical threshold used in particle physics to confirm the existence of new particles.

Other astronomers see room for possible errors in the data, which means the Hubble tension could still be an artifact. “I don’t know how this large of an error is hiding at this point, and if it is, it’s just something no one has suggested,” says team member and Duke University astronomer Dan Scolnic. “We’ve tested every idea that’s been presented to us, and nothing is working.”

The Hubble tension arises from attempts to measure or predict the Hubble constant, which is the current rate of expansion of the universe. Astronomers can use it to calculate the age of the universe since the big bang.

The cosmic microwave background (CMB), a faint glow that formed when the universe was only 380,000 years old, is one method for calculating the Hubble constant. The CMB has been measured by telescopes such as the European Space Agency’s Planck observatory, providing a detailed snapshot of how matter and energy were distributed in the early universe, as well as the physics that governed them. Using the Lambda Cold Dark Matter model, which predicts many of the universe’s properties with remarkable accuracy, cosmologists can mathematically fast-forward the infant universe as seen in the CMB and predict what today’s Hubble constant should be. According to this method, the universe should be expanding at a rate of approximately 67.36 kilometers per second per megaparsec.

Other teams, on the other hand, measure the Hubble constant by looking at the “local” universe: stars and galaxies that are relatively close to us. This version of the calculation necessitates two types of information: how quickly a galaxy recedes from us and how far away that galaxy is. This, in turn, necessitates the creation of a cosmic distance ladder by astronomers.