Nate Silver has been grousing on his Twitter feed lately that graphs like this one are highly misleading, amounting to near malpractice by the media.

What’s wrong with the graph? The numbers there are accurate, aren’t they? Sure, says Silver, but the raw number of new cases in the U.S. doesn’t really tell us whether things are getting better or worse. The other variable we need to know is how many tests are being done. If you test 1,000 people and find 100 infections and then a month later you test 5,000 people and find 100 infections, is the epidemic stable? Or has it actually diminished over those 30 days?

To phrase that differently, how many more infections might we have found on day one if we had tested 5,000 people then instead of 1,000?

That’s been our problem all along, insufficient testing. Maybe the average number of daily cases in the U.S. right now is much higher than it was on April 1. Or maybe it was much higher on April 1 and we only think it’s lower because we weren’t testing nearly as much then. On April 1, the U.S. conducted 107,000 coronavirus tests nationwide. We’ve been over 200,000 daily for the past 10 days and lately at closer to 250,000. Comparing case counts from early April to now is apples and oranges.

In fact, says Silver, by harping on the raw number of new cases and ignoring the rate of positive cases relative to the total number of tests, the media is playing into Trump’s hands. Trump and his allies don’t like hearing how high the number of new daily infections is, which incentivizes them not to pursue more testing. The more tests we do, the more infections we’re going to find, right? Bad optics! But finding more infections is a crucial part of containing the epidemic, along with contact tracing. If the media paid more attention to rates of positive cases relative to testing instead of obsessing about the raw number, the incentive would go the other way. More testing means a lower rate.

It turns out that the positivity rate is shrinking, notes an encouraged Scott Gottlieb. A ray of light through the gloom:

Daily positive cases of covid19 as a percent of total daily tests continues to fall nationally, a very good sign as covid testing expands. Declining positivity could be a leading indicator of an epidemic starting to decline. 10% is still high but it's coming down at a steady pace pic.twitter.com/xVP56dxUmV

— Scott Gottlieb, MD (@ScottGottliebMD) May 7, 2020

Another way of looking at falling covid19 positivity rate across the country. Daily tests are rising, positive cases remaining largely flat, and the positivity rate declining. We're testing more to find same number of cases, suggesting that density of epidemic may be declining. pic.twitter.com/maMYuiG2n8

— Scott Gottlieb, MD (@ScottGottliebMD) May 7, 2020

That’s real improvement. But it’s just one metric. Another metric, one which really is apples-to-apples over time, is deaths. The numbers there are less encouraging:

US daily numbers via @COVID19Tracking:

Newly reported deaths
Today: 2,746
Yesterday: 1,949
One week ago (4/30): 2,138

Newly reported cases:
T: 28K
Y: 25K
4/30: 29K

Newly reported tests:
T: 319K
Y: 242K
4/30: 230K

Positive test rate:
T: 9%
Y: 10%
4/30: 13%

— Nate Silver (@NateSilver538) May 7, 2020

Today’s eye-popping number of 2,746 comes with an asterisk: It includes nearly 1,000 from New York that didn’t happen within the past day but were belatedly confirmed as COVID-related by health authorities within the past day. Even so, yesterday saw nearly 2,000 deaths recorded. The day before that saw nearly 2,600, thanks to some catch-up in weekend reporting. These are big numbers. And they’re not meaningfully improving.

The trend in deaths isn’t looking great compared to other countries either:

It's possible! Lots of other countries are doing it!

We're the only rich country facing death tolls this high, this consistently, for this long. Even the worst-hit EU countries have begun sustained decline; we haven't. Our govt is doing uniquely badly. https://t.co/GG21fEJ24U

— Jeremy Konyndyk (@JeremyKonyndyk) May 6, 2020

Bear in mind that daily deaths have dropped dramatically in New York over the last few weeks. The fact that deaths across the entire U.S. are holding roughly steady with NY in decline means that deaths are increasing elsewhere.

Spend a minute with this visualization at the New Atlantis to get a sense of how hard some states have been hit relative to other leading causes of death. In New York the number of deaths from COVID-19 is so enormous that for the moment it’s outstripped deaths from all other causes. In 10 states (plus D.C.), though, coronavirus deaths now outpace deaths from heart disease, normally the leading cause of fatalities in the U.S. Alzheimer’s disease kills twice as many people in the U.S. every year as flu and pnuemonia combined — and right now COVID-19 is outpacing deaths from that illness in no fewer than 27 states.

It’d be nice if we had a plan to deal with that, but we don’t. So here we are, clinging to glimmers of good news in the data.